Harmonisation Platform 2 Test Targets. Part A: Car Targets

Transcription

1 Harmonisation Platform 2 Test Targets Part A: Car Targets Prepared by Paul Lemmen (Humanetics) Jonas Ekström (Volvo) Colin Grover (Thatcham) Oliver Bartels (BASt) Patrick Seiniger (BASt) Benjamin Marx (Daimler) Version 7 Jun Tushida (Toyota) Maminirina Ranovona (Toyota) Johann Stoll (AUDI) Christian Domsch (BMW) Christof Gauss (ADAC) Issue date 14 May 2012 HP2 Targets 1/52

2 Summary This is the working document for HP2 on the targets for testing of Advanced Emergency Brake Systems (AEBS). It is meant to report development and evaluation activities on the targets to the PNCAP group. Information is collected from vfss, AEB, ASSESS and ADAC. A first draft was provided early October. This is a second draft with updated data collected over the October December 2011 timeframe. The current Part A is dealing with targets representing cars. An identical document, Part B, on pedestrian targets is under preparation for delivery April The document provides some background information on test scenarios and sensor technologies, an overview of requirements as specified by the different projects as well as performance information obtained from testing. In addition information on operational aspects of the test set-ups and cost related info like efficiency of testing (tests per day + test engineers needed) and possibilities for implementation on tracks are provided. As evaluations of targets and test set-ups are still ongoing by the involved projects a final version of the document is expected April May HP2 Targets 2/52

3 Contents 1 Introduction Overview of test scenarios Overview of key sensor technology Radar Camera PMD-Sensor LIDAR Sensor fusion Requirements and means of compliance Dimensions Physical characteristics Radar Camera PMD Sensor LIDAR Summary detectability requirements and means of compliance Functional requirements Test speeds and manoeuvres Positioning accuracy Impact severity and crash forgivingness Aerodynamics Durability Summary functional requirements Performance information ASSESS target (ASSESSOR) Radar Cross Section ASSESSOR target Camera PMD and LIDAR Functional requirements vfss ADAC AEB Summary Operational aspect Test set-ups General description Use of steering robot Driver braking reactions HP2 Targets 3/52

4 6.2 Test data collection and accuracy (VuT / target) Next steps Conclusions / summary Appendix A Description of Test set-ups ASSESS Kart System BASt TNO pre-crash test system (PCTS) ADAC Target System HP2 Targets 4/52

5 1 Introduction Advanced Emergency Brake Systems (AEBS) use remote exterior sensors like radar or camera to detect an imminent crash. Depending on the system a warning may be provided to the driver, brakes pre-charged and/or partial or full braking applied automatically applied to minimize the impact. These actions may be combined with activation of restraints like pretensioning of seat belts or raise of the bonnet (pedestrian safety). Because of their potential in crash avoidance and injury mitigation Euro NCAP intends to include assessment of AEBS in future protocols. Procedures will be defined by the PNCAP group using information from a number of projects: - Advanced Forward-Looking Safety Systems (vfss) Cooperation between OEMs, research and insurance groups world-wide developing test and assessment methods for forward facing safety systems related to accidents with pedestrians and cars. vfss also develops and applies methods on system effectiveness. - Advanced Emergency Brake systems (AEB) Cooperation between insurance organisations Thatcham and IIHS with support from research groups, a supplier and two OEMs. Aims and goals identical to vfss. - Assessment of Integrated Vehicle Safety Systems (ASSESS) EU FP7 Project consortium of OEM s, suppliers, test houses, research organisations and universities. Total 14 partners. Research on test methods for car car accidents (no pedestrians) considering driver behavioural aspects (warning), pre-crash performance evaluation, crash performance evaluation and system effectiveness. - Allgemeiner Deutscher Automobil-Club (ADAC) ADAC defined an evaluation method for AEBS considering the warning and autonomous braking actions to inform consumers on the system performance. The method was applied to various systems offered to the marked and reported in the media. To streamline input from the various projects so-called Harmonisation Platforms (HP s) have been established. Goal is to exchange information on key subjects, thereby generating a clear overview of similarities and differences on the approaches and results. The projects will run independently but via the HP s they are well informed of mutual developments. Three HP s have been established: - HP1 Test scenarios - HP2 Test targets - HP3 Effectiveness analysis This document is the working document for HP2 on the test objects, so-called targets, used in evaluation testing of AEBS for car-car collisions. The targets have to fulfil a wide set of requirements related to sensor-fidelity (mimic a real car for relevant sensors), dynamics (reproduce real world scenarios), crashability / crash forgivingness (avoid damage during testing), durability, handling, etc. All projects involved are either developing or evaluating test targets. Results are reported in this working document. An identical document will be provided for pedestrian detection systems. The document starts with information on test scenarios (chapter 2) and sensor technologies (chapter 3). These chapters are meant to provide background to the requirements definitions of the target. Next the requirements as set by each project are summarised (chapter 4) followed by performance evaluations of various targets (chapter 5). Operational aspects of test set-ups with the targets are provided in chapter 6. This includes a description of the set-up and an indication of the test efficiency. Information on the data collection and required measurement accuracy is given in chapter 6.2. HP2 Targets 5/52

6 As the projects are still continuing their activities an overview of future work and relevant output for this HP2 document is summarised in Chapter 7. As such this document should be regarded as a draft version to be updated by April May HP2 Targets 6/52

7 2 Overview of test scenarios Table 1 and Table 2 on the next pages summarize test scenarios as defined by the different projects. Table 1 groups the information per project while Table 2 gives results per scenario type. The scenarios form the basis for specifying requirements related to dynamics, crashability, etc. Detailed info on the definition of the scenarios including real world relevance is to be provided by HP1. Remarks: - AEB, vfss and ASSESS did extensive accident surveys to retrieve the scenarios. Examples of test speeds as identified in ASSESS are given in Figure 1 through Figure 3. (Note that further info on the scenarios will be provided in the report of HP1). In the ASSESS project the final test configurations result from input of the accident surveys and testability of the scenarios. For example initial testing showed that current test tools cannot handle impact speeds above 50 km/h without introducing damage to the VuT. Therefore scenarios that can be tested depend on the performance of the VuT. - The ASSESS project includes driver reactions in terms of braking after warning. Reaction times are derived from simulator tests. The test set-up requires the introduction of braking at given points in time in the vehicle under test. This does affect the test set-up, see chapter 6, but not the target and propulsion system. Therefore these scenarios are not listed separately in Table 1 and Table 2. - Apart from to rear-end impacts the ASSESS project considers cut-in, crossing and frontal impact scenarios. These are not included here as the document is focussing on rear-end impacts. It is noted though that some of the initial requirements as defined in ASSESS relate to these scenarios. - In the ADAC set-up stationary target tests are conducted with increasing speed. Testing is aborted (meaning no higher speeds tested) at the moment when the system fails. For the rear-end scenarios the maximum speed of the target is 80 km/h as specified by ASSESS. The maximum deceleration is 7 m/s² again specified by ASSESS. vfss assumes a maximum deceleration of 6.2 m/s². The maximum impact speed (in case the VuT is not reacting) is 80 km/h for ASSESS. However, as stated above 50 km/h seems technically feasible at this stage. For costs reasons it might be necessary to limit the number of tests and thus choose relevant testing speeds. Considering general knowledge, without going in detail of accidentology (hence to be confirmed by HP1), it could be useful to distinguish between low severity (representing damage and low end of injury severity), mid severity (representing most types of injuries) and high severity (severe injuries and fatalities). In the rear end test scenarios, these three groups could be represented by approaching speeds of: - below 20 km/h - between 20 km/h and 40 km/h - above 40 km/h up to 60 km/h So despite the large number of tests provided by some of the projects the system performance could be confirmed by 3 tests with approaching speeds of 20 km/h, 40 km/h and 60 km/h. It should be noted though that the 60 km/h impact speed is probably too sever and will likely lead to damages on target and car. In case of damages to the car the performance of the sensor(s) could be affected. Therefore the test at 60 km/h should only be done if it is sure that the tested car will decelerate to a speed with lower risk for any damage. This leads to a reduced test matrix which is covering wide areas of the tests proposed to HP2 by its members: HP2 Targets 7/52

8 Driving speed km/h GIDAS A2 SV GIDAS A2 TV 0 50th 75th 95th Figure 1 Driving speeds (percentiles) for decelerating lead vehicle obtained from GIDAS [ref] Figure 2 Driving speeds (percentiles) for constant speed difference scenario obtained from GIDAS (left) and relative (closing) speed for these scenarios [ref] Figure 3 Driving speeds (percentiles) for stationary target obtained from GIDAS (left) and OTS (right) [ref] HP2 Targets 8/52

9 Table 1 Test scenarios grouped per project Scenario Vehicle under Test Target vehicle Max V at impact Overlap Stationary target V 0 = 25 km/h V 0 = V c = 0 V max = 25 Full Stationary target V 0 = 50 km/h V 0 = V c = 0 km/h V max = 50 km/h Full Constant speed V 0 = 90 km/h V 0 = V c = 50 km/h V max = 40 km/h Full Decelerating target V 0 = 50 km/h at distance to target d = 15m (-0+3) V 0 = 50 km/h; b v = 6,2 (+-0.1) m/s²; V c =?? km/h V max = 50 km/h Full Stationary target V 0 = 10 km/h V 0 = V c = 0 km/h V max = 10 km/h Full Stationary target V 0 = 20 km/h V 0 = V c = 0 km/h V max = 20 km/h Full Stationary target V 0 = 30 km/h V 0 = V c = 0 km/h V max = 30 km/h Full Stationary target V 0 = 40 km/h V 0 = V c = 0 km/h V max = 40 km/h Full Stationary target V 0 = 50 km/h V 0 = V c = 0 km/h V max = 50 km/h Full Stationary target V 0 = 60 km/h V 0 = V c = 0 km/h V max = 60 km/h Full Note: Speed increments in the above tests reduce to 5 km/h once collision has occurred Stationary target under increasing angle between 0 and 45 V 0 = 10 km/h V 0 = V c = 0 km/h V max = 10 km/h Full but target under angle Constant speed V 0 = 30 km/h V 0 = V c = 20 km/h V max = 10 km/h Full Constant speed V 0 = 40 km/h V 0 = V c = 20 km/h V max = 20 km/h Full Constant speed V 0 = 50 km/h V 0 = V c = 20 km/h V max = 30 km/h Full Constant speed V 0 = 60 km/h V 0 = V c = 20 km/h V max = 40 km/h Full Note: Speed increments in the above tests reduce to 5 km/h once collision has occurred Decelerating target V 0 = 50 km/h at distance to target d = 12m V 0 = 50 km/h; b v = 2 m/s²; V c =?? km/h V max =?? km/h Full Decelerating target V 0 = 50 km/h at distance to target d = 12m V 0 = 50 km/h; b v = 6 m/s²; V c =?? km/h V max =?? km/h Full Decelerating target V 0 = 50 km/h at distance to target d = 40m V 0 = 50 km/h; brakes to stationery for approx 0.6 and 1.6s before impact V c = 0 km/h V max = 50 km/h Full Stationary target V 0 = 20 km/h V 0 = V c = 0 km/h V max = 20 km/h Full Stationary target V 0 = 30 km/h V 0 = V c = 0 km/h V max = 30 km/h Full Stationary target V 0 = 40 km/h V 0 = V c = 0 km/h V max = 40 km/h Full Stationary target V 0 = 70 km/h V 0 = V c = 0 km/h V max = 70 km/h Full Constant speed V 0 = 50 km/h V 0 = V c = 20 km/h V max = 30 km/h Full Constant speed V 0 = 100 km/h V 0 = V c = 60 km/h V max = 40 km/h Full Constant speed (brake assist test) V 0 = 100 km/h Driver shall brake V 0 = V c = 20 km/h V max = 50 km/h Full HP2 Targets 9/52

10 insufficiently 1 second after first warning Decelerating target V 0 = 60 km/h at distance to target d = 40 m V 0 = 60 km/h; b v = 3 m/s²; V c = 5 km/h V max = 55 km/h Full Decelerating to stop target V 0 = 50 km/h at distance to target d = 120 m V 0 = 40 km/h; b v = 3 m/s²; V c = 0 km/h V max = 50 km/h Full Stationary target V 0 = 50 km/h V 0 = V c = 0 km/h V max = 50 km/h Full Stationary target V 0 = 50 km/h V 0 = V c = 0 km/h V max = 50 km/h 50% Stationary target V 0 = 80 km/h V 0 = V c = 0 km/h V max = 80 km/h Full Constant speed V 0 = 50 km/h V 0 = V c = 10 km/h V max = 40 km/h Full Constant speed V 0 = 50 km/h V 0 = V c = 10 km/h V max = 40 km/h 50% Constant speed V 0 = 100 km/h V 0 = V c = 20 km/h V max = 80 km/h Full Decelerating target V 0 = 50 km/h at distance to target d = 15m V 0 = 50 km/h; b v = 4 m/s²; V c = 12 km/h V max = 38 km/h Full Decelerating to stop target V 0 = 50 km/h at distance to target d = 14m V 0 = 50 km/h; b v = 7 m/s²; V c = 0 km/h V max = 50 km/h Full Decelerating target V 0 = 50 km/h at distance to target d = 14m V 0 = 80 km/h; b v = 4 m/s²; V c = 12 km/h V max = 68 km/h Full Decelerating to stop target V 0 = 80 km/h at distance to target d = 45m V 0 = 80 km/h; b v = 7 m/s²; V c = 0 km/h V max = 80 km/h Full HP2 Targets 10/52

11 Table 2 Scenarios grouped per type Stationary target Project Vehicle under Test Target vehicle Max V at impact Overlap AEB V 0 = 10 km/h V 0 = V c = 0 km/h V max = 10 km/h Full AEB, ADAC V 0 = 20 km/h V 0 = V c = 0 km/h V max = 20 km/h Full vfss V 0 = 25 km/h V 0 = V c = 0 V max = 25 Full AEB, ADAC V 0 = 30 km/h V 0 = V c = 0 km/h V max = 30 km/h Full AEB, ADAC V 0 = 40 km/h V 0 = V c = 0 km/h V max = 40 km/h Full vfss, AEB, ASSESS V 0 = 50 km/h V 0 = V c = 0 km/h V max = 50 km/h Full ASSESS V 0 = 50 km/h V 0 = V c = 0 km/h V max = 50 km/h 50% AEB V 0 = 60 km/h V 0 = V c = 0 km/h V max = 60 km/h Full ADAC V 0 = 70 km/h V 0 = V c = 0 km/h V max = 70 km/h Full ASSESS V 0 = 80 km/h V 0 = V c = 0 km/h V max = 80 km/h Full Constant speed AEB V 0 = 30 km/h V 0 = V c = 20 km/h V max = 10 km/h Full AEB V 0 = 40 km/h V 0 = V c = 20 km/h V max = 20 km/h Full AEB, ADAC V 0 = 50 km/h V 0 = V c = 20 km/h V max = 30 km/h Full ASSESS V 0 = 50 km/h V 0 = V c = 10 km/h V max = 40 km/h Full ASSESS V 0 = 50 km/h V 0 = V c = 10 km/h V max = 40 km/h 50% AEB V 0 = 60 km/h V 0 = V c = 20 km/h V max = 40 km/h Full vfss V 0 = 90 km/h V 0 = V c = 50 km/h V max = 40 km/h Full ADAC V 0 = 100 km/h V 0 = V c = 60 km/h V max = 40 km/h Full ADAC (brake assist test) V 0 = 100 km/h Driver shall brake insufficiently 1 second after first warning V 0 = V c = 20 km/h V max = 50 km/h Full ASSESS V 0 = 100 km/h V 0 = V c = 20 km/h V max = 80 km/h Full Decelerating target AEB V 0 = 50 km/h at distance to target d = 12m V 0 = 50 km/h; b v = 2 m/s²; V c =?? km/h V max =?? km/h Full AEB V 0 = 50 km/h at distance to target d = 12m V 0 = 50 km/h; b v = 6 m/s²; V c =?? km/h V max =?? km/h Full AEB V 0 = 50 km/h at distance to target d = 40m V 0 = 50 km/h; brakes to stationery for approx 0.6 and 1.6s before impact V c = 0 km/h V max = 50 km/h Full vfss V 0 = 50 km/h at distance to target d = 15m (-0+3) V 0 = 50 km/h; b v = 6,2 (+-0.1) m/s²; V c =?? km/h V max = 50 km/h Full ADAC V 0 = 60 km/h at distance to target d = 40 m V 0 = 60 km/h; b v = 3 m/s²; V c = 5 km/h V max = 55 km/h Full ASSESS V 0 = 50 km/h at distance to target d = 15m V 0 = 50 km/h; b v = 4 m/s²; V c = 12 km/h V max = 38 km/h Full ASSESS V 0 = 50 km/h at distance to target d = 14m V 0 = 80 km/h; b v = 4 m/s²; V c = 12 km/h V max = 68 km/h Full HP2 Targets 11/52

12 Decelerating target to stop ADAC V 0 = 50 km/h at distance to target d = 120 m V 0 = 40 km/h; b v = 3 m/s²; V c = 0 km/h V max = 50 km/h Full ASSESS V 0 = 50 km/h at distance to target d = 14m V 0 = 50 km/h; b v = 7 m/s²; V c = 0 km/h V max = 50 km/h Full ASSESS V 0 = 80 km/h at distance to target d = 45m V 0 = 80 km/h; b v = 7 m/s²; V c = 0 km/h V max = 80 km/h Full Stationary target under increasing angle between 0 and 45 AEB V 0 = 10 km/h V 0 = V c = 0 km/h V max = 10 km/h Full HP2 Targets 12/52

13 3 Overview of key sensor technology A sensor is a device that measures a physical quantity and converts it into a signal which can be read by an observer or by an instrument. AEBS use remote exterior sensors to detect dangerous traffic situations. Sensors (most commonly) used in AEBS include radar (24 and 77 GHz), camera and laser scanner. A short description of these sensors in relation to requirements for the test targets is provided below. 3.1 Radar Radar is an object-detection system which uses pulses of radio waves in way to measure the location of nearby objects or relative speed of moving or fixed objects. The detected object will return part of the energy of the received radar wave. Depending on the following characteristics, it is possible to distinguish automotive radar sensors between: short-range radars, long-range radars and multimode radars. Short-Range radars (SRR) use the frequency band of 24GHz. They present a range up to ~40-50m and a wide observation angle. They can achieve a range resolution of ~15cm and high range accuracy. Hence they can determine the exact position of potential obstacles. Long-Range radars (LRR) use a frequency of 77GHz. Unlike the previous, this type of radar is used to detect objects up to 200m but with a small observation angle of ~±8. This is sufficient for long range applications but its performance drops for targets very close to the vehicle (<20m), resulting in a lower of range measurement quality and angular measurement accuracy. Multimode radar (MMR) is electronically scanning radar, combines a wide Field of View (FOV) at mid-range with long-range coverage to provide two measurement modes simultaneously. This allow to use the same device to enables ACC but also other systems as forward collision warning with pre-crash sensing, brake support. By use of one defined radar corner reflector the received energy will be high and the RADAR will detect the object at long range. This high received energy could be a problem for discrimination (distinguish two objects close to each other). Use of a single corner reflector instead of two or more is an issue at short distance since the target is "outside path". Therefore it's important that the characteristic of the test object matches the real object (Car, pedestrian etc.). Preferable is to NOT use corner reflectors at all. 3.2 Camera Camera sensors are an increasingly important part of active safety systems. They sense lane markings, obstacles and distances. There are two main sensing principals to be mentioned when discussing active safety camera sensors and they are CMOS and CCD. There are many ways of comparing the performance of camera sensors. These methods could include looking at the cameras dynamic range, its signal-to-noise ratio, its low-light sensitivity. The image from the camera is usually processed by some vision algorithm. The purpose is to recognize the object. The algorithm is trained on real objects and therefore it's important that the characteristic of the test object matches the real object (Car, pedestrian etc.). Performance requirements of the test objects (targets) depend strongly on the algorithms used. The most basic requirement relates to the overall dimensions which should represent that of a car. Some algorithms only use contour lines whereas more advanced use information from various details on the vehicle like wheels, rear-lights, interior parts to reduce the number of false warnings. As camera systems are getting more and more advanced the test targets should allow for adding on relevant features to keep up with sensor developments. HP2 Targets 13/52

14 3.3 PMD-Sensor A Photonic Mixing Device, PMD sensor is an optical sensor whose operating principle is based on the Time of Flight (ToF) in the near infrared range (850 nm (nano meter)). The outgoing signal should be reflected properly from some key elements like rear lights and licence plate. Target requirements relate to reflection properties and tautness of the surface at these parts. A further description of this sensor will be provided in the final version. 3.4 LIDAR LIDAR (LIght Detection And Ranging) is a technique used for remote sensing and measures the distance to objects by using laser pulses. The common sensing principle for lidars is the use of time of flight (TOF) technology where a laser pulse is emitted and the reflected signal is detected. The time delay between transmission and reception is measured and the distance can then be calculated due to the proportionality between TOF and distance. The performance of lidars decreases with low visibility such as rain or snow or if the sensor gets blocked by e.g. dirt. The LIDAR sensor uses reflection parts on the object sensitive for IR. For car's this is mainly reflectors, the light housing, and the licence plate. The test object must therefore be equipped with these reflecting parts. However, too big reflectors could saturate the LIDAR receiver with malfunction as a consequence. Therefore it's important that the characteristic of the test object matches the real object (Car, pedestrian etc.). Target requirements relate to reflection properties and tautness of the surface at these parts. 3.5 Sensor fusion Sensor fusion is the combining of sensory data or data derived from sensory data from disparate sources such that the resulting information is in some sense better than would be possible when these sources were used individually. The term better in this case can mean more accurate, more complete, or more dependable, or refer to the result of an emerging view, such as stereoscopic vision (calculation of depth information by combining twodimensional images from two cameras at slightly different viewpoints). Figure 4 Radar sensor concealed behind front bumper (Daimler), camera system as used by Volvo and Lidar sensor (Hella) typically mounted in bumper area. HP2 Targets 14/52

15 4 Requirements and means of compliance This chapter starts with dimensional and physical (sensor related) requirements of the (soft) target. Next requirements for the combined target propulsion system are provided. 4.1 Dimensions Requirements Dimension and mass requirements as collected from the different projects are summarized in the table below. All projects require a 3-D surrogate with dimensions of a mid-sized car. Typical overall dimensions as collected by ASSESS are included. However, for the scenarios under consideration only the rear-end is to be represented (Width x Height). For this part requirements on the depth follow from fit to propulsion systems, crash forgivingness (soft target) and physical characteristics sensors (e.g. shadow needed for camera). Therefore no specific requirements are provided on that dimension. Table 3 Dimensional requirements Description ASSESS vfss AEB ADAC Shape 3-D also for rear-end only testing 3-D focusing on rear-end view 3D 3D currently focusing on rear-end Overall size Depth Rearend part Sized to the popular mid size family car (L x W x H): - Opel Astra 4209 x 1804 x 1488 mm - VW Golf V 4209 x 1804 x 1488 mm Not specified. Follows from fit to propulsion system, crash forgivingness and sensor requirements. Mid size car Not specified Medium sized hatchback or saloon, as per ASSESS Similar to mid size car Not specified Compliance evaluation Compliance of test targets to these dimensional requirements is easily checked by comparing outer dimensions with values specified. 4.2 Physical characteristics The test targets need to mimic physical properties of a passenger car for relevant / available remote exterior sensors. Requirements for each of the sensors described in chapter 3 are provided below followed by a summary section and means of compliance. It is noted here that the German AK3 group is currently discussing key characteristics / requirements needed for safe detection by sensors which are currently used in AEB systems or expected to be on the market in the near future. Input is collected from German OEM s and suppliers. A draft document from the AK3 group is expected early January If the information is public and when considered relevant by the HP2 partners the results will be included in the final version of this report Radar Characteristics for radar sensors are probably the most complex to be realised in the target. Radar Cross Section (RCS) depends on many parameters like distance to the target, view angle (both horizontal and vertical), multipath reflections from elements in the lower surface, influence of local details like sharp edges, etc. Last but not least radar physics are way more complex and less straightforward to understand and interpreted than other sensors like camera. HP2 Targets 15/52

16 In the ASSESS project RCS requirements were derived from: 1) Measurements on mid-size cars 2) Expert input Measurements on cars: A 360 RCS profile was established from measurements on three mid-sized cars (Opel Astra, VW Golf and Nissan Almera). See Figure 5. Note that these measurements were done in a 2-D horizontal plane. No info related to yaw or pitch rotations could be collected in the set-up used. In addition to this ASSESS requires adequate representation of RCS characteristics versus range. See for example Figure 6. Although no specific requirements are defined in terms of corridors the response of the target is to be compared with that of mid-size cars. Judgement of the response is done by radar expert opinions. Detailed measurements by Toyota comparing reflective power of a real car with that obtained from targets (corner reflector and ASESS target) revealed additional requirements in terms of scatter of the response over time and overall shape of the reflection power as function of the beam angle. Scatter may occur due to fluttering of the surface or due to overall dynamics of the target and should be reduced to a minimum in view of repeatability and reproducibility of the test set-up. Although no specific requirements in terms of allowable scatter can be defined the provisions should be taken to have a taut surface minimizing the scatter. The overall shape of the reflection power is used in data interpretation and object classification algorithms and therefore it should be represented correctly. Figure 7 gives examples of the Toyota measurements collected in stationary tests at 5, 30 and 70 m using a measurement period of 100 sec per scan. From tests in ASSESS it was found that the main reflective surface should be curved to obtain the correct shape of the reflection power. A target minimal curvature of approximately 2,5 m is needed. Vertical curvature is preferred as well in view of overall dynamics but of secondary importance. The curvature requirement is also set by AEB and recommended by various vfss partners like Daimler. It also reduces variations in the test outcome due to variations in offset and yaw angle alignment. Some OEM s like AUDI specified to have reflective power from vehicle parts away from the back like rear end axle. It is well known that reflections are obtained from various vehicle parts including parts at an offset from the rear-end. Detailed measurements in ASSESS have shown this as well. Depending on the positioning of the sensor in the front-end this may become more or less relevant (the lower positioned the more relevant). However, correct representation of reflection versus depth characteristics is regarded to be too complex at this stage and not required by most of the radar experts. The example of the AUDI radar indicates that specific requirements should be considered when testing a vehicle. The HP2 group recommends that the car manufacturer forwards and justifies specific requests for addon elements to the test target. Expert input: Further requirements from the ASSESS as well as other projects include avoidance of triple reflectors (corner reflectors), shielding of the rear surface down to the ground to avoid multiple path reflections, shielding of the propulsion system. Triple reflectors (corner reflectors) represent ideal reflectors, which reflect incident rays ideally and punctually. Thus a spatial expansion as obtained from the vehicle rear-end is missing. As a consequence advanced radar systems will not correlate the reflections of a triple reflector with those of the rear-end of vehicle. Moreover multipath reflections inside the reflectors may result in strong interference leading to loss of signals. Finally the larger triple reflectors suffer from additional running time of the beams inside the reflectors and thereby incorrect computation of the distance to the collision surface. For these reasons these reflectors should be avoided if possible. However, reflectors might be needed to introduce specific characteristics and reflective sources at a distance from the rear end to represent parts like rear axle. See f.i. ADAC target. For this reason some sensor suppliers and other experts indicate that some reflectors should be allowed in the target but avoided if possible. HP2 Targets 16/52

17 When considering testing in partial overlap or with angled approaches, the use of corner reflectors should be avoided. Indeed, under such conditions, the reflection profile of the target is likely to differ from the one of an actual car. In particular, the centre of the target (i.e the actual offset) can be estimated with some errors by the sensors, and some uncontrollable variation may appear in the reflection power of the target. Therefore, the use of corner reflector should be strictly limited to full overlap scenario with straight approaches. Figure 5 Radar cross section results for 360 tests of 3 mid-sized passenger cars using 24 GHz sensor (top) and resulting corridors (bottom left). Corridors for 77 GHz sensor (bottom right). For rear end parts only 180 degrees values relevant. Figure 6 Example of dependency of RCS as function of range: measurements on Mercedes S class collected with for 24 GHz sensor HP2 Targets 17/52

18 5m-100% overlap 30m-100% Overlap 70m-100% Overlap Reflection power(dbm) AURIS ASSESSOR Reflection power(dbm) AURIS ASSESSOR Reflection power (dbm) AURIS ASSESSOR Figure 7 Reflection power for Toyota Auris, Corner reflector and ASSESS target as function of beam angle obtained with 77 GHz radar. Measurement data taken over 100 sec per scanner step at distances of 5m (left), 30m (middle) and 70 m (right) Camera For the camera the visual difference between the car target and a real standard car of the compact class should be as small as possible. All projects provided, largely overlapping, requirements for this sensor. The most extensive set, including some requirements related to the development of the system, was provided by AUDI: - Target size: comparable to mid size car as specified in table Free space between chassis and road (app. 17 cm) and realistic shadow must be visible - Licence plate must be visible - Plain colour of chassis without any texture - Representation of wheels and tire shape from both behind and each side - Straight object borders (especially the horizontal ones) - Realistic representation of car lamps - Active lights at dusk resp. night - Active brake light for dynamic target - High vertical symmetry of test object - Realistic representation of rear windshield (partly transparent, minimum grey) As stated some of these requirements, like active lights at dusk, relate to the development of the system. Regarding the dimensions, positioning and reflective properties of licence plates and lights information can be found in various standards like 70/222/EEC, German national standard DIN 74069, ECE-R 3 and ECE-R 48. Typical values for the dimensions of the licence plate are length 520 mm Х Height 120 mm. From the consumer testing point of view ADAC requests usage a light colour while others, AEB and ASSESS, recommend use of low contrast colours like silver to challenge the camera system. In addition to this list AEB partner Volvo specifies the representation of interior parts like headrests behind the windshield. Figure 8 summarizes requirements as defined by Volvo. Table 4 summarizes requirements related to the different sensors including the optical sensors. As can be seen from the information most projects define somewhat more general requirements like Should represent mid-size passenger car with relevant optical features included from vfss compared to the very specific requirements set by AUDI. HP2 Targets 18/52

19 Figure 8 Vision requirements as defined by Volvo PMD Sensor As for camera and radar the PMD sensor requires correct representation of the dimensions of the vehicle. The surface should have proper reflective properties in PMD relevant IR band (around 850 nm wavelength) of 50% 1 as derived from measurements on real cars and without specular reflections. Rear window should be represented. In particular elements like licence plate and rear lights should be represented correctly either by having these elements explicitly included (not recommended for reasons of crash forgivingness as specified below) or by use of reflective foil with comparable reflectivity in PMD relevant IR band. The reflector foil surface area must be oriented in towards the VuT direction of the following vehicle. Also the foil should remain as flat as possible. Fluttering of the target surface should be avoided for this purpose. Also the foil should be inspected on a regular basis for damages like kinks, bends or surface scratches resulting from impacts. Regular exchange of the foil is necessary / recommended LIDAR Also the LIDAR requires correct representation of the dimensions of the vehicle. The surface should have proper reflective properties and rear window should be represented. As for the PMD elements like licence plate and rear lights should be represented correctly either by having these elements explicitly included (not recommended for reasons of crash forgivingness as specified below) or by use of reflective foil with comparable reflectivity Summary detectability requirements and means of compliance Table 4 summarizes the physical requirements in relation to relevant sensor types. As indicated requirements for the radar are still under discussion hence updates to be expected for the final version of this report. 1 For PMD sensors the reflectivity is indicated as a relative value where the 95% definition correlates with a white DIN A4 paper HP2 Targets 19/52

20 Compliance evaluation Due to their different nature each project is evaluating the compatibility with these requirements in a different way. ASSESS is developing test tools including the target. An iterative approach is applied where updates are installed and evaluated against requirements to fine tune the performance. Performance evaluation is done on different levels depending on the sensor type: 1) Correlations between response and requirements as defined for the radar sensor. As no definite corridor values were defined for items like scatter evaluation is often done on the basis of expert opinions. 2) For the PMD sensor reflective properties of surrogate materials was evaluated to match with measurements on real cars. 3) For the camera sensor basic evaluation was applied where it was checked if required features like rear lights, wheels etc. were present. In addition to this a method is being explored in which pictures of targets are processed by camera algorithms and compared with pictures of cars. 4) Extensive track testing provides input on parameters like tautness of the surface (resulting in test variations). vfss on the other hand evaluates targets from members and third parties by organising testing events during which participants provide subjective evaluations. Each participant is runs various tests with the different targets and reports on the findings and experiences without any detailed measurements. The information of all participants is collected and condensed into a summary report. Recommendations for improvements are provided to suppliers of targets. AEB is testing along the same line as vfss in a smaller set-up with lesser number of targets and lesser number of vehicles. Measurement data as well as collected sensor data are being analysed to identify shortcomings of the targets and provide recommendations for improvements. For the ADAC target RCS measurements have been conducted by several manufacturers. A further compliance check for the radar and other sensors was done when the first comparison test of six vehicles was finalised by providing the test results to the manufacturers for plausible checking. There were no complains. 4.3 Functional requirements The test targets will be used in combination with various propulsion systems. Some examples are provided in Figure 9. Because of the direct interaction with the propulsion system it is difficult to define functional requirements for the target itself. Therefore the following requirements apply to target in combination with propulsion system. Figure 9 Examples of propulsion systems (a) Rabbit vehicle IDIADA; (b) AB Dynamics soft crash set-up Daimler and (c) BASt kart HP2 Targets 20/52

21 Table 4 Physical requirements target Radar Camera PMD LIDAR ASSESS vfss AEB ADAC - Should represent midsize -Appropriate 360 passenger car. profile - Curvature under - no concentrated investigation to refine reflection source details - Shielding of internal - Use of reflectors is to be parts considered as exception - Curved reflective (hence need to be surface to reduce yaw forwarded and justified by angle dependency OEM) - No reference data available - Appropriate 360 radar profile - Curved surface of reflective material with radius of at least 2,5 m to obtain correct shape of reflection over beams - Correct RCS versus range characteristics - RCS Stability to be realized by taut surface - Avoid use of corner reflectors if possible - Shield rear surface down to ground to avoid multiple path reflections - Colour to correspond with most common colour of the vehicles on the roads but low contrast colour like silver or gray preferred - Wheel simulations - Ground clearance of 17 cm for shadow - Licence plate - Lightning including brake lights - No requirements set by ASSESS partners as this sensor was not used in ASSESS. Therefore requirements as defined by AUDI assumed. - Target size: comparable to mid size car - Licence plate and rear lights to be represented with correct reflective properties - Rear window to be represented - Should represent midsize passenger car with relevant optical features included. - Target size: comparable to mid size car as specified in table Licence plate and rear lights to be represented with correct reflective properties in PMD relevant IR band (around 850 nm wavelength) - Rear window to be represented - Taut surface finish required to prevent fluttering - Use of low contrast colour (e.g. silver or gray) for body surface - lights and licence plate - strong horizontal contrast edges, - transparent glass area with interior illustration - under body shadow - all features to be symmetrical - Similar to camera detectability: reflection from licence plates, lights and reflectors - Taut surface finish required to minimise deflection at speed - Should be representative to real world car. No reference data available however. - The target should be detected by all cars in the test - From consumer perspective light colour - Shadow under the target in necessary - Licence plate, rear window, rear lights should be simulated - Reflective elements necessary - Reflective elements necessary Although protocols are yet to be developed it may generally be assumed that the propulsion system with target will be operating on a flat road surface under good / reasonable weather conditions (meaning no limited wind, dry (no rain, snow, hail), temperature range between 7 C and 35 C). The road condition is expected to be good HP2 Targets 21/52

22 without crack or holes. This is important for sliding contact between the target and the road surface which occurs during and after crashes Test speeds and manoeuvres Speed, acceleration and deceleration requirements follow from the scenarios as defined in chapter 3. Resulting requirements are summarized in Table Positioning accuracy To ensure that a system will not pass or fail based on variations in the kinematics between VuT and target / propulsion system test accuracies were is defined by most projects. See Table 5. Values defined are based on experience; no specific sensitivity testing was conducted to retrieve / verify these numbers. The ASSESS project will conduct repeats of various tests to further study Repeatability and Reproducibility of different test set-ups Impact severity and crash forgivingness The combined target and propulsion system should in principle be capable of sustaining impacts from VuT s of maximum 50 km/h speed difference. At this speed no damage should be introduced to the VuT. Although test matrices defined by ASSESS and ADAC include scenarios with higher impact speed, testing has shown that 50 km/h is at the upper range of what can be achieved for impact speeds when using realistic target / carrier systems Aerodynamics The target / propulsion system is meant to reproduce traffic scenarios at speeds up to 50 (vfss); 60 (ADAC); and 80 (ASSESS) km/h as specified in chapter 3. The surface finish should remain taut for PMD reflective surfaces and, probably to a lesser extent, for camera and radar systems. In case of rabbit vehicle static tilt will occur due to wind drag. This should be less than 5 (number obtained from customer specifications to various test labs, no further rational in terms of measurement data provided) Durability The target is exposed to various external loads including impact forces from the VuT, abrasion loads when over the asphalt, low level aerodynamic loads, etc. Especially the impact and abrasion loads may damage the target and provisions should be taken for an acceptable lifetime of the target. Defining requirements on durability is difficult. In general good reparability is given more importance than durability. Restoring damage during a test sequence is of key importance to complete the sequence. Also, once failed, spare parts should be available Summary functional requirements Table 5 summarizes the functional requirements. Compliance evaluation Maximum speed and deceleration characteristics evaluation of the target / propulsion system is straightforward. Regarding the velocity and positioning accuracies recording of these parameters for the VuT and Target / Propulsion system is required with given measurement accuracy (see table 12, chapter 6). Full validation testing has to be conducted within a given calibration interval (e.g. one year) per test system. This will NOT be needed for each test run. If possible, self-diagnostics of the measurement system will be used for on-site accuracy (applicable to most DGPS / IMU systems). If that is not possible, mounting of the test system to VuT and installation of all stationary parts of the test system will be HP2 Targets 22/52

23 documented and at least one verification test run will be conducted (e.g. stationary displacement measured during contact VuT / target which means displacement in x = 0). For the crash related items it is anticipated that beyond the specified impact speeds the VuT as well as the target / propulsion system may be damaged. Testing should be started with low risk scenarios (with minimal v max ) and proceeded with scenarios with generally increasing risk as proposed by AEB and ADAC. For the aerodynamics stability requirement it is recommended evaluate compliance via test runs with VuT and target travelling at identical constant speed at given distance. Sensor recordings in VuT to be evaluated for increasing speed. HP2 Targets 23/52

24 Table 5 Functional requirements Description ASSESS vfss AEB ADAC Maximum speed target 80 km/h 50 km/h 50 km/h 100 km/h Maximum deceleration target Test velocity accuracy Lateral position accuracy w.r.t. CL of test Longitudinal position accuracy between VuT and target at start of braking in lead vehicle decel scenario 7 m/s^2 6,2 m/s^2 6 m/s^2 5 m/s^2 ± 0.5 km/h ±5% Input from Colin ±1 km/h at whole speed range ± 0.20 m ±0.25 Input from Colin ±0.20 m ± 0.50 m Accuracy of -0 / +3 at 15 m distance resulting from decelerating lead vehicle scenario Input from Colin ± 1 m Deceleration acc. ± 0.2 m/s 2 ± 0.1 m/s 2 ± 0.5 m/s 2 Fit of target to propulsion system Aerodynamic stability Crash capabilities Crash forgivingness Durability The system shall fit on the propulsion systems available in ASSESS: - IDIADA Rabbit vehicle with crane - BAST Kart - AB Dynamics driving box Daimler - up to a speed of 80 km/h - Static tilt under constant velocity less than 5 The system should not be damaged in crashes with passenger cars up to v of xx km/h for rear shunt tests. Crash forgiving up to v of 50 km/h for rear shunt tests. The value of 50 km/h is based on feasibility testing. In practise crash forgivingness up to 80 km/h was requested but this is not to be achievable. Needed but no specific requirement could be defined due to wide variety of tests. Presence of repair kit is specified though. Carrier should not interfere with physical properties of target Up to 50 km/h 60 km/h stationary and preferably also moving 60 km/h (see above) Needed but no specific requirement available. Possibilities for repair during tests is more important. - Intermediate fit of rear end to rabbit vehicle set-up (see Figure 9 (a)). - Ultimately ABD Soft Crash Target Vehicle. Target speed up to 50km/h for current front to rear shunts test proposal. Future developments may require greater up to 80 or 100km/h Crash forgiving up to v =?? km/h for rear end shunts Not defined The system shall fit on the propulsion system used at ADAC Target speed up to 100 km/h The system should not be damaged in crashes with pass. cars up to v of 50 km/h for rear shunt tests. Crash forgiving up to v = 50 km/h for rear impact of target Not defined HP2 Targets 24/52

25 5 Performance information As stated in section each project evaluated compatibility with requirements in a different way. In ASSESS performance evaluation is done on the basis of correlation between response and requirements as well as performance in track tests. Only the ASSESSOR target, developed in the project is evaluated. vfss did subjective evaluations of a wide range of targets by a wide range partners based on track test results. AEB is evaluating along the same line as vfss in a smaller set-up with lesser number of targets and lesser number of vehicles. AEB does use measurement and detailed sensor data tough to support the analysis. The next sections provide performance information on the different targets from the different project. First results of the ASSESS project are presented followed by vfss, ADAC and AEB. 5.1 ASSESS target (ASSESSOR) Radar Cross Section ASSESSOR target The compliance of the full 3-D ASSESSOR target with corridors established from 360 degrees measurements on 3 mid size cars is shown Figure 16. After fine tuning of the radar reflective material results are largely within corridors over the entire circumference. At 180 (rear-end facing) results are at the centre of the corridor for the 24 GHz sensor and at the lower bound for the 77 GHz. Note that these data relate to measurements at a distance of 18.5 m. Correlation of range versus RCS information for a 24 GHz sensor (Daimler) is shown in Figure 11 (24 GHz system). Again reasonable to good correlation is obtained. However, measurements by Toyota for a 77 GHz sensor at several distances highlighting that the level of correlation could be quite inconsistent depending on distance. Figure 12 shows that the ASSESSOR peak reflection is smaller than the reflection of an actual car at a distance of 10m. In addition, detailed observation of RCS horizontal profile (Figure 13) in a stationary set-up shows that the ASSESSOR is not consistently matching the characteristics of a car for this sensor. While the correlation is perfect at 30 m, it provides unclear centre at shorter distances and unstable reflection at longer distances, most likely caused by fluttering of the surface. Studies with various reflective surfaces showed that a curved surface with radius r 2,5 m provides good correlation with real cars. The curvature will also reduce test variations occurring from minor misalignment in overlap and yaw between target and VuT. As the ASSESSOR can be used in different set-ups using various propulsion / carrier systems the RCS in different set-ups was evaluated using a 77 GHz sensor. Figure 14 shows the RCS as function of distance for a reference car (red curves), the ASSESSOR standalone (left), ASSESSOR mounted on the BASt kart (centre) and on the ADAC rig (right). From these data it is observed that the RCS varies with the test set-up. This is explained by the fact that the target does not shield the entire propulsion system / carrier. As the target is somewhat above the ground multi path reflections underneath the target may occur. Moreover both the BASt kart and especially the ADAC carrier had metal parts extending above the target causing variations at larger distances. The target or target / carrier combinations should be improved in this respect to reduce variations. Both the introduction of a tauter surface and the curvature would be beneficial for other sensors like PMD / LIDAR as well. A summary of the performance of the ASSESS target is provided in Table 9. HP2 Targets 25/52

26 Figure 10 Correlation of 3-D ASSESSOR radar cross section wih corridors established from 3 mid-sized passenger cars. For rear ends only 180 degrees values are of interest. Figure 11 RCS measurement result versus range for Mercedes S class (left) and ASSESSOR rear end (right) using 24 GHz radar Figure 12 RCS measurement result versus range for Toyota Auris and ASSESSOR rear end using 77 GHz radar. HP2 Targets 26/52

27 5m-100% overlap 30m-100% Overlap 70m-100% Overlap Reflection power(dbm) AURIS ASSESSOR Reflection power(dbm) AURIS ASSESSOR Reflection power (dbm) AURIS ASSESSOR Figure 13 Reflection power for Toyota Auris, Corner reflector and ASSESS target as function of beam angle obtained with 77 GHz radar. Measurement data taken over 100 sec per scanner step at distances of 5m (left), 30m (middle) and 70 m (right) RCS in db 0-5 RCS in db -5 RCS in db Entfernung in m Entfernung in m Entfernung in m Figure 14 Influence of propulsion system on RCS of ASSESSOR. Red curves represent measurements on reference car, blue curves are data for ASSESSOR as standalone (left), attached to BASt kart (centre) and to ADAC rig (right). Data collected with 77 GHz sensor Camera None of the vehicles tested so-far in the ASSESS project uses these sensors. Testing by vfss and AEB partners provided some performance info. In general the overall dimensions of the ASSESSOR rear-end match with typical dimensions of a mid-size car. Licence plate, rear-end lights (active) are present as add-on items. Since these were not always applied in testing by third parties (see Figure 15) it was recommended to have these items rigidly attached to the ASSESSOR). Simulation of wheels is needed but these can be easily added. Windshield is present. Regarding the surface it was recommended to have a low contrast colour surface without any texture. For a more objective evaluation ASSESS is exploring the use of camera algorithms evaluating pictures from the target in comparison to pictures of real cars. This method will be applied in future versions of the ASSESSOR. A summary of the performance of the ASSESS target for camera s is provided in Table 10. Figure 15 ASSESS Target (left) + configurations used during test events (letters added, tail and braking lights removed or placed on other positions) HP2 Targets 27/52

28 5.1.3 PMD and LIDAR For the PMD performance a sample of the licence plate and rear lights tail reflector sticker material was measured by vfss partner AUDI and compared to real licence plates/tail reflectors. The measurements showed that the material is accepted as substitute. During the testing it was found that upon repeated impacts the reflective properties of the sticker material degrade. As a consequence stickers should be replaced after about 10 to 20 tests. A summary of the performance of the ASSESS target is provided in Table Functional requirements Functional requirements as described in section largely relate to the combination of propulsion system (carrier) and target. Most of these requirements are therefore dealt with in chapter 6. However, some items directly related to the soft target itself are included in this section. These relate to fit of the target to different carriers, aerodynamic stability, crash capability and crash forgivingness. The ASSESSOR target was designed to fit at various propulsion systems as used in the ASSESS project. This includes (see chapter 6) the BASt kart, the IDIADA rabbit vehicle and the AB Dynamics drive box. In addition some first preliminary testing in the ADAC set-up was done. The target can be mounted on all carriers but crash capabilities and crash forgivingness appear to depend on the combination of carrier target. External loads applied upon impact result in lifting of carrier target systems for those systems where the Centre of Gravity is below the line of force acting between VuT and target. For instance in the current BASt kart ASSESSOR configurations impacts should be restricted to 40 km/h to avoid lifting. This can be improved by adjustments to the target as well as the carrier (note that the BAST carrier was already optimised for this purpose). Identical observations were made in other configurations. The crash forgivingness of the ASSESSOR is somewhere between 30 and 50 km/h. Again this depends on the combination of carrier and target. The heavier the carrier the lower the allowed impact speeds to avoid damage to the VuT. For the systems tested in ASSESS speeds up to 40 km/h were fine although damage could occur to some soft parts of the vehicle at speed of 30 km/h. Further improvements to be expected by fine tuning the vent holes. Regarding the durability issues were found in the ASSESSOR especially the connection between air-filled tubes and vented box suffered upon serious impacts. The first ASSESSOR failed in this mode after 200+ tests during a severe impact test. The failed target could be repaired by the manufacturer but this takes several days out of service. Design solutions for the improvement of the tube vented box interface have been proposed and will be implemented in the next version. Also during testing the tubes may start to leak. However, this could easily be resolved using a repair kit without significant delay in testing. A summary of the performance of the ASSESS target is provided in Table vfss As indicated the vfss project organises round robin tests with partners, evaluating various targets. Figure 16 gives an impression of the first testing event held in Table 6 presents the results of the second event held mid Performance evaluation of the various targets was done via a questionnaire in which all participants provided findings and opinions. As indicated before the ASSESSOR target was tested without camera features (see also Figure 15) resulting in a bad qualification for this sensor. HP2 Targets 28/52

29 Figure 16 Impression of vehicles and targets used during the first round robin event organised by vfss Table 6 Outcome of vfss target evaluations from June 2011 event NHTSA- Balloon Car Balloon LKA stationary targets ADAC Toyota Target ASSESS 2010 classic (without camera features) ASSESS 2D (without camera features) ASSESS curved (without camera features) moving targets ASSESS 2D (without camera features) ADAC Overall Assessment of radar characteristic Overall Assessment of visual characteristic Overall Assessment of evaluation of system performance evaluation Overall results Based on the results in table 6 vfss organised another round robin in October 2011 during which the ASSESSOR, an updated version of the ADAC target (Version 2), the LKA balloon and a target prepared by Thatcham were evaluated. During this event the LKA balloon and the ADAC V2 target were found to perform well as can be seen from the results in Table 7. The ASSESSOR target showed degraded RCS performance which is explained by the fact that the target was tested repeatedly during the period before the vfss event. This should be improved in a final version. Table 8 gives an overview of the rating for the suitability of the target and test set-up for use in the test procedures as proposed by vfss. Again the ADAC and LKA (indicated as MB Balloon) set-up perform well in comparison to ASSESS and others. Table 7 Outcome of vfss target evaluations from October 2011 event HP2 Targets 29/52

30 Table 8 Test procedures and target suitability for test procedures from October 2011 event 5.3 ADAC ADAC is using their test target in comparative testing of AEB systems in vehicles (see report: Comparative test of advanced emergency braking systems). See Figure 17 below. For the target RCS measurements have been conducted by several manufacturers. The measurements have been conducted with the latest version of the ADAC Target system. A typical example is depicted in Figure 17 which shows the good correlation with a real car. Figure 17 Impression of vehicles and ADAC target V1 (improved version available) considered in AEB comparative testing [Ref. to Comparative test of advanced emergency braking systems ] --- BMW 5 Series --- ADAC Target V2 Figure 18 RCS of ADAC Target V2 in comparison to BMW 5 Series HP2 Targets 30/52

31 A further compliance check for the radar and other sensors was done when the first comparison test of six vehicles was finalised by providing the test results to the manufacturers for plausible checking. Manufacturers got info on measures for timings of functions in their system from ADAC and agreed on the outcome. The ADAC target V2 is suitable for radar sensors with 77 GHz and 24 GHz. One small corner reflector contributes to a centred RCS. Radar experts attending some test events preferred this setup with one corner reflector which is not too dominant. 5.4 AEB Thatcham conducted tests with 5 cars and three different targets in a rabbit vehicle set-up. Targets were compared by running tests at an approach speed of ~20 km/h starting at a distance of 61 m. Reference tests against a real car were made as well. Results are plotted as the system outputs confidence level of an object based on radar and visual attributes. The score is given on a scale of 0-5 with 5 being the higher confidence (green). Score of 0 indicates insufficient visual detail to confirm the object (red). Preliminary results are provided in Figure 19 [source The AEB test results are currently being analysed at detailed level by Volvo to explain differences between the different targets, especially differences between the various configurations of the ASSESS target (in particular: are these differences caused by the radar or the camera performance). When discussing the AEB results it was noted by several experts that the camera response is likely to be affected by the vertical support beams which are black in one case and white / gray in other cases. Camera systems may interpret the target / beam system as a truck rather than a passenger car which may be in conflict with the radar observations. Also the use of rear lights should be more consistent for camera recognition. Just like the example of Figure 14 on the RCS variation of the ASSESSOR in different setups this shows that subtle differences in a set-up can affect sensor detections and thereby system performance. Figure 19 Performance of test targets as collected by AEB presented in terms of confidence levels (5 is high, 0 is low) for test with closing speed of ~20 km/h. [ HP2 Targets 31/52

32 Figure 20 Illustration of differences in set-up: vertical support beam variation in between tests / targets as well as rear light variations [ 5.5 Summary Table 9 through Table 11 summarise performance info on the different targets as described above. Only those targets which are under consideration by one or more projects are included. These are the ASSESSOR (ASSESS target), ADAC balloon and the Laudenklos (LKA) balloon. Further information is collected during test programs in October / November (vfss, ASSESS) and from further data analysis (detailed sensor information collected by AEB). As stated the ASSESS target will be updated In view of the sensitivities to different test set-ups and modifications observed in the previous sections HP2 might take a further effort to specific in detail the elements needed to mimic passenger cars for the various sensing systems. This involves for instance detailed definition of the reflective material (material type, locations applied, size of reflective area, replacement instructions after testing) for PMD and LIDAR, detailed definition of reflective elements for radar (on the basis of AK3.11 activities), etc. HP2 Targets 32/52

33 Table 9 Performance evaluation for radar Requirement ASSESS target ADAC Target LKA balloon - Appropriate radar profile rear end Yes, checked with detailed measurements. Also 360 RCS installed. Reflection profile could be improved at different distances. Yes, checked with measurements Currently under evaluation. More info to follow from vfss testing Oct Appropriate radar profile 360 Yes Not yet available Currently only rear - Preferably no concentrated reflection source (corner reflector) but based on recommendation from suppliers allowed to some extend. - Curved surface of reflective material with radius of at least 2,5 m to obtain correct shape of reflection over beams and reduce yaw angle dependency - Shield rear surface down to ground to avoid multiple path reflections - Correct RCS versus range characteristics - RCS Stability to be realised by taut surface OK no corner reflectors included Current version largely flat. Feb 2012 new version with indicated radius Tests of RCS in three different set-ups show sensitivity. This is (partially) because the target did not cover the propulsion system entirely at the upper parts. To be improved. OK, evaluation data available for 24 GHz and 77 GHz radars Checked and to be improved in the next version One 100mm corner reflector included. (see comment in 0) Curved bumper included - Radius 2,65 m. Equipped with radar reflection foil (height: 15 cm) Not necessary as it is implemented only on ADAC rig OK, evaluation data for 24 GHz and 77 GHz Stability tested in many tests. Functionality of AEBS repeatable end Curved shape due to internal pressure. Radius not exactly known but in range of 2.5 m. OK Currently under evaluation. More info to follow from vfss testing Oct 2011 No measurements available but internal pressure in balloon expected to give stability. HP2 Targets 33/52

34 - Colour to correspond with most common colour of the vehicles on the roads but low contrast colour like silver or gray preferred. Table 10 Performance evaluation for camera Requirement ASSESS target ADAC Target LKA balloon Currently white but can be changed - Wheel simulations From rear view wheels are not well simulated. To be positioned underneath vehicle with correct width - Ground clearance of 17 cm for OK shadow Rear simulated with realistic photo of yellow car. Colour of the car can be changed. OK, simulated by realistic photograph OK 19 cm ground clearance, realistic shadow simulation no problem for camera systems - Licence plate OK Real licence plate - OK - Lightning including brake lights OK but currently using active lights as add-on. Need fixed marking as back-up in case active lights are not installed - Target size: comparable to mid size car - Plain colour of chassis without any texture - Straight object borders (especially the horizontal ones) - Active lights at dusk resp. night (secondary importance) - Active brake light for dynamic target (secondary importance) - High vertical symmetry of test object - Realistic representation of rear windshield (partly transparent, minimum grey) Passive rear lights on photograph. Active lights can be installed No information, to follow from vfss testing Oct 2011 No information, to follow from vfss testing Oct 2011 No information, to follow from vfss testing Oct 2011 No information, to follow from vfss testing Oct 2011 No information, to follow from vfss testing Oct 2011 OK OK No information, to follow from vfss testing Oct 2011 OK Plain colour possible, but best performance with realistic photograph No information, to follow from vfss testing Oct 2011 OK OK No information, to follow from vfss testing Oct 2011 Available Can be installed No information, to follow from vfss testing Oct 2011 Available Can be installed No information, to follow from vfss testing Oct 2011 OK OK No information, to follow from vfss testing Oct 2011 OK Grey colour possible, but latest tests by OEM require particular absorbent material. No information, to follow from vfss testing Oct 2011 HP2 Targets 34/52

35 Table 11 Performance evaluation for Lidar Requirement ASSESS target ADAC Target LKA balloon - Target size: comparable to mid size car as specified in table 4.1 OK OK No information, to follow from vfss testing Oct Licence plate and back lights to be represented with correct reflective properties in PMD relevant IR band (around 850 nm wavelength) - Taut surface finish required to prevent fluttering OK. Reflectivity of reflector stickers compared with licence plate, tail lights etc. and found to be correct Not directly checked for LIDAR but improvements included in next version related to radar OK Reflector stickers and licence plate suitable for PMD OK no fluttering also at high speeds No information, to follow from vfss testing Oct 2011 No information, to follow from vfss testing Oct 2011 Table 12 Functional performance Requirement ASSESS target ADAC balloon LKA balloon Fit of target to propulsion system Fits on IDIADA Rabbit vehicle, BAST Kart, AB Dynamics drive box Fits on ADAC propulsion system Fits on specific pulling system Aerodynamic stability All in rear end only configuration on 3-D set-up - Tested up to a speeds of 80 km/h No static tilt in configurations tested due to types of carriers used OK, tested up to 100 km/h Crash capabilities Crash tests conducted at 60 km/h. In track tests impacts by VuT up to 50 km/h sustained. However, lifting of carrier observed due to fact that CoG and impact force are not aligned. Performance therefore related to carrier / target combination. Crash forgivingness Evaluated in crash tests up to 60 km/h. No damage observed but operational performance depending on carrier used. In ASSESS track testing max impact speed with BASt kart and IDIADA rabbi found to be about 40 km/h, which is lower than the requested 50 km/h. To be improved in next version Durability 1 st target sustained 200 tests among which various severe impact tests (crash tests up to 60 km/h) before sending to manufacturer for damage repair. In addition minor damages during testing that could be repaired on site. However, RCS found to degrade during repeated impact testing / transport of target. To be improved. Crash tests conducted at up to 60 km/h. In track tests impacts by VuT up to 50 km/h sustained. Crash forgiving up to 50 km/h speed difference Some hundred tests have been conducted with different speeds without significant damage. Up to max. 50 km/h Up to max. 50 km/h Large number of tests conducted (> 1000) with impacts. No (substantial) damage reported HP2 Targets 35/52

36 6 Operational aspect This chapter starts with a short overview of different test set-ups used in ASSESS, vfss, ADAC and AEB. The overview includes info on test efficiency (how many tests per day). Next some specific items related to the use of steering robots and the possibilities of addressing driver braking reactions are considered. 6.1 Test set-ups General description Appendix A gives a short description of a number of test set-ups as currently installed in a number of test tracks in Europe. A summary overview is provided in Table 16 at the end of this chapter. The table includes info on the target / carrier system, way of control of the VuT and target, and compliance with requirements like positioning accuracy. Also info on test efficiency, type of scenarios (looking to other scenarios than rear-ends only) and the ease of implementation on other tracks are included. Figure 21 below provide compliance information of the BASt kart set-up w.r.t. velocity, positioning and acceleration accuracies. Cumulative results over 20 tests (mix of stationary, constant speed and decelerating scenarios) show that requirements on positioning accuracy, velocities etc. are met in at least 50% of the tests except for the initial distance and deceleration level of the kart in the decelerating lead vehicle scenarios. These can easily be improved by introducing direct distance measurement like laser and controlled braking level in the kart. Figure 22 shows a typical example of accuracies achieved for the IDIADA set-up indicating that a lateral position accuracy of ±2 cm can be achieved by using steering robots v Krt v OV v req All experiments, n OV = 20, n Krt = y y req y suggested All experiments, n = 20 cumulative distribution v suggested cumulative distribution figures_reproducibility.m - 24-Jun :03: v in km/h All experiments, n = 12 1 cumulative distribution Offset at min(t 0.1 impact,t TTC=min ) in % Offset req in % figures_reproducibility.m - 24-Jun :03: Offset = y 1.4/100 in % cummulative distribution 0.1 figures_reproducibility.m - 24-Jun :03: y = 0.5 ( y - y ) in m max min All scenarios A1, including repetitions of A1A3, n = 8 1 p = 0.49 h = 0 TTC < TTC median = 2.21s 0.9 TTC >= TTC median = 2.21s Small corridor width should lead to higher warning TTC values, 0.2 however this is not the case. Test procedure seems to be robust against 0.1 corridor variations. A ranksum test does not identify diffrent distributions. figure_corridor_warningttc.m - 24-Jun :09: Corridor width in m HP2 Targets 36/52

37 1 All experiments, n = 4 1 All experiments, n = x x req 0.9 d²x Krt /dx² d²x Krt,req /dx² cumulative distribution cumulative distribution figures_reproducibility.m - 24-Jun :03: x in m figures_reproducibility.m - 24-Jun :03: d²x Krt /dx² in m Figure 21 Example of preliminary reproducibility (procedure and tools under improvement): cumulative info from BASt kart set-up comparing position and velocity values with requirements as defined in ASSESS. Lower graphs show distance between VuT and kart at moment of braking in lead vehicle deceleration scenario and kart deceleration level. Figure 22 Typical example of lateral position accuracy obtained with driving robot. Data from ASSESS test by IDIADA Use of steering robot Driving robots are frequently used in testing of AEB systems. They allow accurate reproducing of test scenarios in terms of positions, velocities and accelerations. It is possible to equip the vehicle with steering, brake, accelerator and gear box robots, in order to automate the driving actions to the vehicle. Depending on the needs of each scenario (in terms of manoeuvres and tolerances), different combinations of driving robots can be used. In order to limit the number of hardware used during the tests, the following configuration is proposed by several test houses: - VUT equipped with steering robot (to keep lateral range) and braking robot (to reproduce driver reaction after warning signal is triggered). - Propulsion system equipped with steering control (to keep lateral range), accelerator control (to keep a certain speed) and braking control (to reproduce deceleration in scenarios where leading vehicle brakes with a certain deceleration) With this configuration the only task out of automated control is the longitudinal range and the speed of the VUT, which is easily implemented by a trained driver. The accuracy provided by these systems can achieve ±2 cm for lateral range (see Figure 22) and ±5 cm for longitudinal range. It should be stated though that there is a conflict between the precision requirements for target and VuT track-holding and possible VuT steering reactions. The test aims at simulating an inactive/passive driver, i.e. a driver who does not actively steer for some time before the collision. If a driver were to correct steering during the test, the systems might recognise him/her as an alert driver and react differently depending on the amplitude and frequency (speed) of the steering action. Also continuous steering actions may result in small yaw motions that may affect sensor readings. Figure 23 shows steering wheel angle and rate corresponding to the test sequence of Figure 22. The amplitudes from the control action are very limited, both in terms of angle (bellow ±1.5 º) and rate (bellow ±15 º/s). It has been checked with several test houses, OEMs and suppliers that these values should not disable the actuation of an AEB systems. HP2 Targets 37/52

38 RangePosLateral2 [m] SR Angle [ ] Timer_1 [s] Timer_1 [s] Figure 23 Steering wheel angle and rate corresponding to the test sequence shown in Figure 22 As an alternative to the steering robot a trained test driver may be considered. As it concerns longitudinal scenarios is seems quite well feasible. In such case the driver should get instructions to represent an inalert driver and hence no steering action should be applied during the last seconds of the test. In all cases it is recommended that the actuation of the driver over the steering wheel (in terms of angle and rate) and tolerances of longitudinal and lateral ranges are constantly monitored and kept below a certain limit Driver braking reactions An important item in the test set-up is the inclusion of driver warnings. Studies by ASSESS [Ranovona et. al., 2011, ASSESS D3.2 Report on the experimental study results of the evaluation of behavioural aspects] and other projects have shown the relevance of the warning, hence means must be available for evaluation. Typical driver reaction times on warnings have been derived in various studies. See e.g. Table 13 [Burg/Moser, Handbuch der Verkehrsunfallre ATZ Verlag, 2007]. In the ASSESS higher reaction times with a 50 %- ile value of 1.67 sec were found. As stated in the ASSESS D3.2 report these values should be considered only as an example and not as a generic model because of the dependency on the distraction and the scenario considered. Table 13 Driver reaction times on warnings (Burg/Moser, Handbuch der Verkehrsunfallre ATZ Verlag, 2007) HP2 Targets 38/52

39 The ASSESS project experiments use a braking robot (plus optionally an accelerator robot) to reproduce driver brake actuation. Each vehicle s warning signal (preferably audio signal) is detected by a signal processing system (simplest case: one single frequency detector integrated circuit). Driving robots are actuated after a specific delay time. The level of pedal force was also derived from the simulator studies and can be applied in the robot. ADAC evaluates the driver warning separately from the autonomous vehicle actions in terms of alert cascade (meaning the system should provide an increasing level of alertness) Assessment of the alert cascade includes the activation time(s) and whether or not the alarm is noticeable and effective. Figure 24 shows an example with a cascade that is considered to be effective: Initially the alarm consists of an acoustic and visual warning on the instrument cluster and (optional) head-up display. If the driver fails to react, the car makes a little jerk to alert the driver. The dual-stage partial braking reduces speed relatively early, giving the driver time to react. Right before the crash, the car initiates autonomous full braking. 6.2 Test data collection and accuracy (VuT / target) Certain variables need to be recorded during the test runs. On the one hand this concerns data needed for performance evaluation like velocity and deceleration of the vehicle under test. On the other hand it includes info needed to evaluate the test itself like position of the VuT and the target to give an indication of the alignment in lateral direction. An overview of test data to be recorded is included in Table 14. This table also includes requirements on measurement accuracies as set by the different projects as well as sample rates. The sample rate depends on the expected dynamics for each variable. Velocities and positions are composed of more or less low frequencies allowing for interpolation and upsampling for points of time in-between discrete times. All sampling rate requirements in Table 14 result from the requirement that TTC should be relatively accurate. TTC is calculated from positional and velocity measurements between the two vehicles. The TTC accuracy is a result the individual accuracies in this measurement chain - thus, individual accuracies should be relatively high. As can be seen sampling rates of 100 Hz (10ms) for all trigger events (including brake robot actuation) and time synchronization are needed. This will result in a longitudinal positional accuracy of the trigger events of 0.1 m (for difference velocity of 10 m/s). Positional measurement sampling rate of each vehicle can be maximum 20 Hz if time synchronization is appropriate. In that case, however, positions can be interpolated because of the relatively low dynamic behaviour. DPGS does not deliver position updates with more than 20 Hz. Figure 24 Example of alert cascade as considered by ADAC Info from AEB on driver reaction to be included. Caution should be taken on the fact that radar sensors due to their measurement principle have a relatively high cycle time and thus low update rate. HP2 Targets 39/52

40 Note that synchronized time is required in all cases where two individual measurement systems on each vehicle are used. This is of course not required if only one vehicle performs measurements, however it is expected that it will be demanding to achieve appropriate accuracy of positional and velocity measurements of the remote vehicle. In order to ensure more or less constant conditions during all tests in all test laboratories, it will also be necessary to record those variables that need to be held constant or that are given values for the test setup, e.g. lateral distance between the vehicles, time of driver reaction, if possible, warning time, manual brake actuation to ensure the human driver did not interfere etc. Table 15 shows values that were defined within the ASSESS project. As indicated in section these precision requirements for target and VuT track-holding might be in conflict with the VuT steering reactions being interpreted as alert driver by the system and thereby suppressing autonomous system activation. The indicated track-holding requirement may be too exacting and should be revised/adapted in the pre-trial phase. If synthetic driver reactions will be applied by means of driving robots, the signal processing chain between warning pickup and driver reaction should have no significant delay. If two independent measurement units are used, there needs to be a precise method for offline (post-processing) synchronisation, e.g. GPS time that is recorded on both units.. HP2 Targets 40/52

41 Table 14 Test data to be recorded Measured function Accuracy Sampling rate Comments Actual TV Speed over synchronized time Actual VuT Speed over synchronized time Actual offset over synchronized time 0.1 km/h 20 Hz (if time synchronisation has an accuracy of at least 10 ms) 0.1 km/h 20 Hz (if time synchronisation has an accuracy of at least 10 ms) 0.03 m 20 Hz (if time synchronisation has an accuracy of at least 10 ms) IMU with DGPS / solo DGPS, Radar or Laser, PMD on VuT if necessary accuracy has been *PROVEN*, Datron Correvit if time synchronisation has been *PROVEN* IMU with DGPS or solo DGPS / Datron Correvit if time synchronisation has been *PROVEN* IMU with DGPS / solo DGPS / Laser Scanner / PMD / Camera System Actual braking force by braking robot over synchronized time 1% maximum value 100 Hz Brake actuator output Actual distance over synchronized time 0.03 m 20 Hz (if time synchronisation has an accuracy of at least 10 ms) IMU with DGPS / solo DGPS, Radar or Laser, PMD on VuT if necessary accuracy has been *PROVEN* Synchronized time of impact 0.01 s Not applicable for time-stamp recording / 100 Hz for continuous recording of trigger signal Synchronized time of warning 0.01 s Not applicable for time-stamp recording / 100 Hz for continuous recording of trigger signal Touch sensor (trigger) on at least one of the two vehicles (two touch sensors could provide time synchronisation if GPS time is not available for that task Signal processing system (audio or optical, depending on kind of warning) Synchronized time at Pretensioner trigger Synchronized time of Human driver brake actuation (if any) TV longitudinal acceleration over synchronized time VuT longitudinal acceleration over synchronized time 0.01 s Not applicable for time-stamp recording / 100 Hz for continuous recording of trigger signal 0.01 s Not applicable for time-stamp recording / 100 Hz for continuous recording of trigger signal 0.1 m/s² 100 Hz to allow filtering 0.1 m/s² 100 Hz to allow filtering Some kind of trigger (ASSESS: measuring fuse in regular fuse box for seat belt pretensioners) Touch sensor (trigger) on brake pedal) On-board acceleration sensors / IMU On-board acceleration sensors / IMU HP2 Targets 41/52

42 Table 15 Proposed test accuracy (limits for test execution) [ASSESS D4.2] Parameter Controllability Measurement accuracy Test Velocity ± 1.0 km/h ± 0.1 km/h Distance (longitudinal) ± 0.50 m ± 0.03 m Distance (lateral) ± 0.20 m ± 0.03 m Acceleration / Deceleration ± 0.5 m/s 2 ± 0.1 m/s 2 HP2 Targets 42/52

43 BASt Kart IDIADA Rabbit Vehicle Picture Target / carrier system and way of control Target attached to remotely controlled Kart. Kart is equipped with highprecission DGPS and IMU measurement system. Kart velocity is controlled by onboard microcontroller (deceleration and following distance to be controlled by on board microcontroller. Kart lateral dynamics controlled by operator (Who also has the option to According to intervene in all phases, if measurements, the needed for security). whole setup achieves Kart measurements the accuracy transmitted to VuT for realtime specifications. display. Target suspended from crane or trolley with swiffels. Both attached to SUV, used as propulsion system. Target with full autonomous driving capabilities: steering robot (position in the lane) accelerator robot (speed of the target) brake robot (deceleration of the Table 16 Summary of test set-ups VuT control Target Compliance with requirements Various but mainly tested with ASSESSOR VuT controlled by human driver (lateral dynamics and initial acceleration phase until experiment starts), by driving robots (longitudinal dynamics). Human driver has the option to intervene in all driving phases. VuT with some full autonomous driving capabilities: steering robot (position in the lane) brake robot (brake reaction after the warning is issued) Relative positions and speeds of VuT and Target controlled in real time Various but tests in ASSESS conducted with ASSESSOR Maximum speed: proven to be above 80 km/h Maximum deceleration: proven to be 7 m/s² Lateral position accuracy: proven to be < m in at least 50% of all tests Longitudinal position accuracy: will be improved for next test series Other Fitment of full 3D ASSESSOR possible after October Aerodynamic stability Crashability of the system: maximum of 40 km/h speed difference at impact is survivable without significant damage to either vehicle Crash forgivingness: see above. According to ASSESS requirements: Max. speed 100 km/h Maximum decel. 7 m/s2 Lat. position acc. + 2 cm Long. Pos. acc. + 5 cm Other Fitment of full ASSESSOR 3D Aerodynamic stability Crashability of the system: maximum of Test efficiency Type of scenarios Implementation on other tracks 2 to 3 persons needed Easily possible for set up (including VuT instrumentation) and operation. With this capacity about 2 days needed for preparation and execution of full set of tests VuT and Rabbit vehicle instrumentation: 5 days De instrumentation: 1.5 days Execution of all rearend scenarios: 4 days Execution of all cut in scenarios: 2 days Rear end scenarios and with full 3 D target also cut in manouvres. Methods for inclusion of driver reactions based on accousic warning installed. According to ASSESS: Scenarios: Rear end with full 3 D target Cut in maneuvers with full 3 D target Reactions: No driver reaction Driver reaction after warning (acoustic or Easily possible. Compatible Differential GPS antenna needed. Thatcham runs identical set up HP2 Targets 43/52

44 TNO VEHIL target) Target mounted on PCTS system consisting of trolley, which is guided and driven The VuT is placed on the chassis dyno in a controlled test back and forth with a guided environment. Reactions rope, like can be found in of the chassis dyno are crash labs. The rope itself is coupled back to the driven by a motor which is main controller which hardware in the loop uses this to alter the coupled to the chassis dyno setpoints of the PCTS. at which the VuT is placed As such relative positions and speeds of VuT and Target controlled in real time Tests planned with ASSESSOR for Oct Nov timeframe 50 km/h of closing Data analysis and speed at impact reporting: 4 days Crash forgivingness: strongly depending on the VuT Max speed : 80 km/hr relative speed between VUT and test target Max decel: 10 m/s2 Lat. Pos. accuracy: first estimation: +/ 0.1 m Long. Pos. accuracy: first estimation: 0.03 m The above numbers are indicative. Final numbers based on testing experience will be available after a first test series planned for November. vibrational warnings and coupled with the brake robot in the VuT) Initial set up (2011) Not possible rear end and frontal (high speed) scenarios. Later versions include lateral control of target allowing for cut inn and crossing scenarios. ADAC Rail/carriage system that can be towed by any available car Use is being made of test drivers for both steering and braking. ADAC Target v1: Modified balloon car ADAD Target v2: Further modifications, optimized RCS and camera detectability Other Crashability of the system and crash forgivingness: maximum of 40 km/h of closing speed at impact due to crash backup system. If target has a velocity of 80 km/hr it will be automatically braked to 40 km/hr with 10m/s2 before it impact into the crash backup system. Maximum speed up to 100 km/h Crash forgiving up to 50 km/h Maximum deceleration 7 m/s² Lateral position accuracy +/ 0,2 m Longitudinal position accuracy controlled by the VuT 2 persons needed for Rear end scenarios set up and operation. Installation time ca. 1h. Typical testing of 1 car within 1 day. After an impact only visual checks necessary (5min) Easily possible HP2 Targets 44/52

45 AB Dynamics drive box LKA setup Info will be provided in final version of report Target build around Central Drive Box which uses an electric motor with onboard batteries to propel the vehicle and houses the control system, which can accurately guide the vehicle along a pre programmed course at a defined speed. The controller uses position feedback from a GPScorrected inertial navigation system to ensure that highprecision guidance is achieved. VuT with some full autonomous driving capabilities: steering robot (position in the lane) brake robot (brake reaction after the warning is issued) Relative positions and speeds of VuT and Target controlled in real time Daimler soft crash and 3 D ASSESSOR Maximum Speed: 70 km/h Maximum deceleration: unknown Path Following Accuracy: Dependent upon motion pack type [2 cm (1 SD RMS) typical maximum] Commercially available system Thatcham rabbit vehicle Identical to IDIADA set up HP2 Targets 45/52

46 7 Next steps As discussed during a PNCAP meeting January 2012 the ADAC target was selected for further testing. It was recommended to ADAC to further cooperate with vfss, AEB, suppliers and individual OEM s to further fine tune and evaluate the target. Final version to be achieved mid HP2 Targets 46/52

47 8 Conclusions / summary Information on test targets and test configurations for evaluation of Advanced Emergency Brake Systems (AEBS) has been collected from vfss, AEB, ASSESS and ADAC. The current Part A is dealing with targets representing cars. An identical document related to pedestrians is expected end The document provides requirements as specified by the different projects and gives performance information collected so far. Also addition information on operational aspects of the test set-ups as used in various laboratories is provided as well. The information is provided in a neutral way without indicating specific preferences for any type of target or test set-up. Decisions on this are to be made in the PNCAP group on the basis of the collected information. To facilitate the reader performance information is summarized in various tables in chapters 5 and 6. It should be noted here that subtle differences in test set-ups can result in differences in system performance. Examples for radar and camera were provided in chapter 5. This should be taken into account when deciding on allowing for a single set-up only or multiple set-ups. HP2 Targets 47/52

48 Appendix A Description of Test set-ups ASSESS Kart System BASt Required equipment and mounting in VuT The test setup involves two vehicles, one is the VuT (Vehicle under Test), the other is the propulsion system. Both are equipped with position and velocity measurement systems of high accuracy (in the case of ASSESS preliminary testing, this was a unit that combines differential GPS and inertial measurement (DGPS-IMU) and achieves an accuracy of around 4 cm). All measurement data of the propulsion system is transmitted to the VuT via Wi-Fi and thus available for instant evaluation of each test run. The propulsion system s speed and deceleration are held constant by a PID controller on board. For safety reasons, both controllers need to be activated by the propulsion system operator. Lateral movement of the propulsion system is always controlled manually. One key feature of the ASSESS test setup is that driver reactions can be applied to the VuT s system depending on the time of warning. Acoustic and optic driver warnings are processed by the appropriate devices like automatic fast frequency detections system for acoustic warning, and passed to braking and accelerator robot systems in the VuT. After a given delay time corresponding to driver reaction times derived from drive simulator tests with volunteers, the appropriate reactions are applied to accelerator and brake pedal. The position measurement system and Wi-Fi bridge is easily installed in the trunk of the VuT, some antennas need to be placed on the roof, and an additional computer display and keyboard will be placed in the front (mounted to the windshield via vacuum cups). Preparation of the VuT takes about 1 to 2 hours. If driver reactions should be applied to the VuT, braking and (perhaps) accelerator robots need to be mounted. This will take another 1 hour. Implementation on other tracks Given the dynamics of the cart, the VuT and the ASSESS test scenarios the required length of the track is approximately 300 m. The required width is about 20 m. These number result from the scenarios with both vehicles travelling at the same initial speed and the propulsion system braking suddenly with a give deceleration, see above. The kart system is highly mobile all necessary equipment fits into a regular transport van like the Mercedes Sprinter, Fiat Ducato etc. Unloading, mounting and initial checks including accuracy checks and dry runs require up to 2 hours, demounting and loading including deequipment of the vehicle under test requires up to one hour. Weather and environmental conditions In principle the kart set-up allows for testing in rainy and other bad conditions. However, as the braking performance of the VuT is influenced by weather conditions it is recommended to test on dry roads. Test efficiency As indicated the Kart set-up needs operation by two persons. Based on experience gained so far, a valid and checked set of ASSESS rear-end test scenarios can be recorded typically in 4 hours. Together with the preparation of the VuT and the propulsion system the rear-end test sequence as specified by ASSESS (including driver reaction runs) can be conducted within one and a half working day for two persons. Efforts are summarized below. HP2 Targets 48/52

49 This does not include data analysis and documentation. Staff effort for preparation of VuT 3 hours, 2 persons (6 hours total) 1 hour braking and accelerator robot, 1 hour DGPS IMU, 1 hour measuring positions, setting configurations, final checks Staff effort for testing on test track 1 day, 2 persons 2 hours unloading, mounting, initial checks 4 hours testing 1 hour demounting, loading TNO pre-crash test system (PCTS) Required equipment and mounting in VuT The test setup involves one vehicle, the VuT (Vehicle under Test), and a propulsion system called the pre-crash test system (PCTS). The VuT is placed on the chassis dyno in a controlled test environment, being the test facility called VeHIL which is situated in Helmond, The Netherlands. This VeHIL facility is a large hall of 200x40 meter in which the total test setup is placed. On the PCTS a target can be mounted, e.g. the assessor. The target is mounted on the trolley, which is guided and driven back and forth with a guided rope, like can be found in crash labs. The rope itself is driven by a motor which is hardware in the loop coupled to the chassis dyno. So the reactions of the chassis dyno are coupled back to the main controller which uses this to alter the setpoints of the PCTS. The chassis dyno, where the VuT is placed on, is equipped with a position and velocity measurement system. The PCTS is equipped with a position and velocity measurement system too (encoder). All measurement data of the PCTS and the VuT is transmitted to the control room via a hardwired connection and is thus available for instant evaluation. One key feature of the ASSESS test setup is that driver reactions can be applied to the VuT s system depending on the time of warning. Acoustic and optic driver warnings are processed by the appropriate devices like automatic fast frequency detections system for acoustic warning, and passed to braking and accelerator robot systems in the VuT. After a given delay time corresponding to driver reaction times derived from drive simulator tests with volunteers, the appropriate reactions are applied to accelerator and brake pedal. On the VuT itself, no extra measurement systems have to be added to measure the position and the velocity, because this is done by using the chassis dyno. The chassis dyno has to be set upped for each car by tuning the road load parameters. These can be determined by performing a coast down analysis outdoor and on the chassis dyno. By comparing the results it is assured that the dynamic performance of the car on the chassis dyno is comparable to the dynamic performance on the road. Both frontal / rear-end scenarios with and without overlap can be tested. Some setup time is needed between these different offset tests to change the PCTS. Estimated is that this will be 1 hour (to be verified). Implementation on other test-facilities The PCTS cannot be installed directly in other facilities. In that sense it is like a crash facility which cannot be moved instantaneously. Of course the concept itself can be rolled out over multiple facilities. Weather and environmental conditions HP2 Targets 49/52

50 The PCTS and chassis dyno together in the VeHIL facility assure that there is a controlled environment in which a test can be repeated time after time under the same conditions. This assures testing without being dependent on weather conditions. Repeatability Because the system runs in a controlled environmental and is fully computer controlled the repeatability should be comparable to the repeatability as currently is expected from a crash lab facility. The repeatability has to be proven via the ASSESS tests later this year. VuT damage risk The target that has to be tested, e.g. assessor, can be mounted on the PCTS. The PCTS itself can assure that the impact between the VuT and the assessor is always limited. This is done by using predefined safety zones in which the PCTS starts decelerating to assure that the impact is controlled and the impact speed remains below a predefined threshold. Next to that, in front of the VuT a damper will be placed that can minimize the impact energy on the VuT. Required equipment and mounting in VuT The test setup involves one vehicle, the VuT (Vehicle under Test), and a propulsion system called the pre-crash test system (PCTS). The VuT is placed on the chassis dyno in a controlled test environment, being the test facility called VeHIL which is situated in Helmond, The Netherlands. This VeHIL facility is a large hall of 200x40 meter in which the total test setup is placed. On the PCTS a target can be mounted, e.g. the assessor. The target is mounted on the trolley, which is guided and driven back and forth with a guided rope, like can be found in crash labs. The rope itself is driven by a motor which is hardware in the loop coupled to the chassis dyno. So the reactions of the chassis dyno are coupled back to the main controller which uses this to alter the setpoints of the PCTS. The chassis dyno, where the VuT is placed on, is equipped with a position and velocity measurement system. The PCTS is equipped with a position and velocity measurement system too (encoder). All measurement data of the PCTS and the VuT is transmitted to the control room via a hardwired connection and is thus available for instant evaluation. One key feature of the ASSESS test setup is that driver reactions can be applied to the VuT s system depending on the time of warning. Acoustic and optic driver warnings are processed by the appropriate devices like automatic fast frequency detections system for acoustic warning, and passed to braking and accelerator robot systems in the VuT. After a given delay time corresponding to driver reaction times derived from drive simulator tests with volunteers, the appropriate reactions are applied to accelerator and brake pedal. On the VuT itself, no extra measurement systems have to be added to measure the position and the velocity, because this is done by using the chassis dyno. The chassis dyno has to be set upped for each car by tuning the road load parameters. These can be determined by performing a coast down analysis outdoor and on the chassis dyno. By comparing the results it is assured that the dynamic performance of the car on the chassis dyno is comparable to the dynamic performance on the road. Both frontal / rear-end scenarios with and without overlap can be tested. Some setup time is needed between these different offset tests to change the PCTS. Estimated is that this will be 1 hour (to be verified). Implementation on other test-facilities The PCTS cannot be installed directly in other facilities. In that sense it is like a crash facility which cannot be moved instantaneously. Of course the concept itself can be rolled out over multiple facilities. HP2 Targets 50/52

51 Weather and environmental conditions The PCTS and chassis dyno together in the VeHIL facility assure that there is a controlled environment in which a test can be repeated time after time under the same conditions. This assures testing without being dependent on weather conditions. Repeatability Because the system runs in a controlled environmental and is fully computer controlled the repeatability should be comparable to the repeatability as currently is expected from a crash lab facility. The repeatability has to be proven via the ASSESS tests later this year. VuT damage risk The target that has to be tested, e.g. assessor, can be mounted on the PCTS. The PCTS itself can assure that the impact between the VuT and the assessor is always limited. This is done by using predefined safety zones in which the PCTS starts decelerating to assure that the impact is controlled and the impact speed remains below a predefined threshold. Next to that, in front of the VuT a damper will be placed that can minimize the impact energy on the VuT. ADAC Target System The test setup involves two vehicles, one is the Ego-vehicle (vehicle under test), the other (CO-vehicle) is any towing vehicle that can tow the ADAC Target System connected to its towing bar. The Ego-vehicle and the ADAC-Target can be equipped with position and velocity measurement systems of high accuracy. ADAC Target V1 The ADAC Target System consists of a target mounted on a carriage that can run along a rail system. The setup is towed behind a car or mini bus to obtain moving target test conditions. In the case that the subject car crashes into the target object, the carriage is accelerated along the rail system to get the same speed as the subject car at the other end of the rail (at the towing vehicle) the carriage is stopped with a spring damper system. The carriage used on the rail system consists of a guide system that engages with the rail with metal U-sections and rollers. For stability two wheels that run on the road are added. The target object is supported on a platform and backup with a vertical plate. The ADAC Target System can be used as stationary object as well as movable target. For stationary use the system is used without a towing vehicle. The propulsion system s speed and deceleration are held constant by the driver of the Egovehicle. A speed limiter can assist to hold constant speed. Lateral movement of the propulsion system is controlled manually. Brake or accelerator robot systems can be applied to Ego- and Co-vehicles to ensure accuracies. HP2 Targets 51/52

52 The ADAC Target system can be transported in a regular transport van like the Mercedes Sprinter, Fiat Ducato etc. ADAC Target V2 (improved camera detectability and RCS) Implementation on other tracks The ADAC Target System can easily used on other tracks. The requirement for the track length is mainly influenced by the test scenarios. It should be minimum 300 m. The target system is not sensitive to minor damages on the track surface. Weather and environmental conditions The ADAC Target System can be used in rainy and other bad conditions. However, as the braking performance of the VuT is influenced by weather conditions it is recommended to test on dry roads. Test efficiency The operation of the ADAC Target System needs two persons. Typical testing of one vehicle can be conducted within one day. The installation time for the ADAC target is less than one hour. The target system is available all day without any restrictions. Tests can be conducted without waiting times. After an impact the system should only be visually checked and the target must be fixed at the end of the rail system again. This takes max. 5 minutes. Service The service costs are very low. Only the rollers have to be checked and exchanged from time to time. This can be done by any mechanic. HP2 Targets 52/52