DEVELOPMENT OF A LANE KEEPING SUPPORT SYSTEM FOR HEAVY-TRUCKS

Transcription

1 DEVELOPMENT OF A LANE KEEPING SUPPORT SYSTEM FOR HEAVY-TRUCKS Mauro Montiglio, Stefania Martini, Vincenzo Murdocco Centro Ricerche Fiat ScpA - Strada Torino 50, Orbassano (TO), Italy mauro.montiglio@crf.it, stefania.martini@crf.it, vincenzo.murdocco@crf.it ABSTRACT FIAT Research Center has developed a Haptical Lane Keeping Support system for heavy trucks, aimed to prevent unwanted lane departures. Unlike conventional systems which issue acoustic warnings, this system provides a tactile feedback to the driver, when the vehicle is unintentionally drifting out of the lane. The intervention, in the form of an active torque on the steering wheel, suggests the driver with the corrective steering action to undertake, in order to bring back the vehicle towards the center of the lane. The work has been carried out in the context of research project SAFELANE, within the EU project PReVENT. KEYWORDS: Active Steering Torque, Lane Departure Warning, Lane Vision, ADAS. INTRODUCTION It is known that a relevant cause of road accidents involving heavy commercial vehicles is the unintentional lane departure, due to driver distraction, inattention or drowsiness. A c c o r d i n g t o a c c i d e n t d a t a o v e r t h e l a s t 1 0 y e a r s f r o m G e r m a n r o a d w a y s ( s e e A D A C, ), % o f a l l a c c i d e n t s a r o s e f r o m u n i n t e n d e d l a n e d e p a r t u r e. Intelligent systems that support the driver in ma intaining the lane can contribute to prevent those kind of accidents. The Haptic Lane Keeping Support system developed by Fiat Research Center is intended as a preventive/active safety system, aiming at decreasing the risk of unwanted lane departures by alerting the driver of an unintended movement of the vehicle out of the current lane. The system differs from conventional passive Lane Departure Warning systems, which use only acoustic or visual warning in case of an impending lane departure and still relies on the driver s ability to react to warnings. It is more evolved since it provides an active torque to the driver on the steering wheel, thus suggesting him the right manoeuvre in order to correct the vehicle trajectory towards the lane centre, with the aim to prevent unwanted movement of the vehicle out of a designated traffic lane. On the other hand, the Haptic Lane Keeping Support system described here differs also from fully autonomous vehicles, where the driver is completely removed from the lane keeping task, thus eliminating any possibility of driver errors. 1

2 Pilutti [4] provides an extensive description on lane-keeping systems which starts with a definition of warning, intervention and control functions. Warnings do not directly alter the vehicle trajectory and require that the driver must choose to act on the warning in order for the warning to have any effect. Intervention and control have the ability to directly affect the vehicle trajectory. Intervention has limited authority and is meant to augment driver commands not replace them. C o n t r o l is defined here as automatic control having full authority to steer the vehicle, which effectively removes the driver from the loop. In terms of Pilutti's definition the system presented here falls in the category of intervention systems, since the driver perceives an active torque feeling in the steering wheel, which suggests the driver with the corrective steering action to undertake, in order to bring back the vehicle towards the correct lane pos ition. The system uses of active control to assist the driver in the lane-keeping task, but the main feature of this system is that the driving task still belongs to the driver. In critical situations, like imminent lane departure, the control system interacts with the driver by supporting him, in order to correct the trajectory in an optimal way. SYSTEM CONCEPT The objective is to develop a new generation of on-board adaptive lane-keeping support system, for use in commercial vehicles, on motorways and e x t r a -urban roads. The system reaction in critical lane-departure situations comprises the control of warning devices and an active steering actuator. To perform this, the present Haptic Lane Keeping Support system has to follow some essential requirements: - The system should predict the imminent lane departure prior to crossing a lane marking, so that the driver has enough time to take a corrective action. - The system should not interfere with normal driver s operation, to avoid nuisance. - The system should be highly reliable. - The probability that a proper steering intervention is issued should be sufficiently high. - False warnings and false interventions should be rare. These features require that the system has information about vehicle position and o rientation relative to the road, geometric characteristics of upcoming road segment, vehicle dynamics data. Such information can, for instance, be acquired through a vision based system, radar sensor, GPS. SYSTEM ARCHITECTURE A brief description of main subsystems involved in this Haptic Lane Keeping Support system will be given below. This function has been carried out by equipping the vehicle with a sensing device and with an active the steering unit. The system architecture consists of: - a Lane Vision System represented by a camera system, capable to work out data related to the road geometry (distance of the vehicle with respect to the left/right 2

3 l a n e -marking, curvature of the road ahead, boundary type, number of adjacent lanes) and the vehicle trajectory (heading-angle of the vehicle with respect to the central axis of the lane). The camera is installed inside the cabin, in a wiped area behind the windscreen. The orientation of the camera viewing-angle is along the longitudinal axis of the vehicle, i n order to see the lane in front of the truck and to detect symmetrically left and right lane-markings. - a purpose-built steering-unit based on an Electric Power Steering (EPS) unit, with specific Electronic Control Unit, torque sensor and steering angle sensor. The EPS allows to produce the additional torque applied to the steering column, thus providing the necessary haptical feedback to the driver. As feedback signals, it provides the steering angle and the torque acted by driver. This extra actuator is geared to the input shaft of the Hydraulic Power Steering (HPS) in order to produce the additional torque requested by the System. - an automotive Electronic Control Unit containing Control Algorithms whose objective is to assist the driver to maintain the vehicle inside the lane, using information from the sensorial system. Control strategies process data coming from the vision system and from on-board sensors related to vehicle dynamics; then the algorithms calculate a proper active steering torque, in ord er to support the driver to avoid unwanted lane departures. - a Human Machine Interface to allow the driver to manage the system activation and deactivation and to set the preferred level of active torque; - a CAN bus that makes possible communication among s ensors, steering system and ECU. CONTROL STRATEGY The Haptic Lane Keeping Support system generates a warning in terms of resistant torque on the steering wheel, to alert the driver that the vehicle is unintentionally about to move out of its lane. This active intervention supplies a tactile feedback to the driver and induces him to take a corrective action in order to bring back the vehicle towards the center of the lane. The active torque applied to the steering wheel varies according to the estimated l ook -ahead distance from the borderline and the deviation angle. It results upper limited to avoid excessive intervention levels, which could perturb the tractor-semitrailer vehicle dynamics; the intervention torque can also vary depending on cargo load. A key issue in the development of a Haptic LKS system is the decision when a warning should be issued. The correct timing of warning depends on precise knowledge of the error or deviation that the drive is about to have, related to some parameters, such as lateral displacement, heading angle and vehicle speed. To achieve their goal, control algorithms were structured in the following three basic component groups: - lateral control module; - warning decision module; - human machine interface. The lateral control module is the portion of Haptic Lane Keeping algorithms which retrieves sensor signals from vision system, takes current vehicle state and predicts future vehicle state (lateral position) and calculates the steering command. The vision system provides l ateral displacement from the centreline at some look -ahead distance and the angle between the road 3

4 tangent and heading of the vehicle. If the lateral position of the vehicle, at the estimated look - ahead distance, exceeds an indicated threshold value, a res istant torque is generated. The second part, the warning decision module is the portion of the warning algorithms which elaborates the predicted vehicle state and other data, such as road curvature, lane width, lane boundary type, vehicle acceleration, an d decides whether to trigger an intervention in terms of active steering torque. Finally, the human machine interface is the module that has to display the parameters used for guidance and warning. The whole control system architecture is shown in the figure below: Figure 1 - CRF Haptic Lane Keeping Control System Architecture There are two major problems related to this control system: the intentional curve-cutting and the auto -pilot misuse. The first problem is related to possible false alarms that the system could provide when the driver is crossing slightly the lane marking inside the curve, which is usually referred as a curve cutting. The distinction between intentional lane departures, like an overtake, or unintended drift out of the current lane is a specific aspect considered by the control system. Consequently, during a borderline approaching/crossing if the driver has not switched-on the vehicle s turning indicators, the system evaluates this event as unwanted and applies a resistant torque to the steering wheel to oppose to the lane departure. Many truck drivers tend to touch and cross the lane boundary quite often during normal driving: to reduce the amount of interference for intended lane departures, proper control strategies must be able to restrict the operational field of the system in particular scenarios. When the driver activates the indicators, the function is automatically disabled because a lane change manoeuvre is expected; the driver can also bypass the system at any time turning the steering wheel slightly. But a specific control strategy to manage intentional curve cutting manoeuvres is also necessary in the warning decision module, in order to prevent false interventions: in this case the active torque is deactivated on the inner side of the curve, allowing the driver to negotiate the curve following a trajectory coherent with his driving style. The second problem is related to the possible auto-pilot misuse by the driver: a system that applies a steering wheel torque in order to keep the vehicle in the lane can almost be used as an auto -pilot, as stated by Balzer [3], and the driver can attempt to use it improperly., e.g. he 4

5 can try to drive the truck without his hands on the steering-wheel. To avoid this potential misuse of the system as an auto-pilot, an integral component of the control strategy is a module which detects whether the driver has his/her hands on the steering wheel or not. Here the existing steering torque sensor of the EPS system is used in order lo estimate the amount of torque applied by the driver and therefore to estimate properly the "Hands on steeringwheel" status. When the system recognize that the driver is not keeping his hands on the wheel, the system deactivate the int ervention of the active torque, since the driver must always be kept in the loop and he must be responsible for vehicle driving. Finally, the control system is also able to warn adequately the driver if the vision system is temporarily incapable of managing the road scenario: in this case a dedicated icon in the instrument cluster starts blinking, thus informing the driver about the unavailability of the LKS function. INTEGRATION OF THE ACTIVE STEERING-ACTUATOR ON THE VEHICLE Two different installations have been evaluated during the development phase. The first alternative was to have an Electric Power Steering (EPS) unit connected to the steeringcolumn, i.e. installed inside the cab. The second alternative was an EPS unit directly geared to t h e i n p ut shaft of the Hydraulic Power Steering (HPS) unit, installed below the cab. The installation inside the cab had the advantage of a better protection of the actuator from dust, humidity, etc. but was critical in terms of dimensions, i.e. interference with the driver s knees. Therefore, the installation below the cab was preferred, since the available room between the cab-floor and the chassis was enough for installing an assembly of a conventional HPS unit with the new EPS unit. Figure 2 shows the CAD picture of the active steering unit on the h e a v y-t r u c k. Electric Power Steering unit Hydraulic Power Steering unit Figure 2 - CAD picture of the active steering unit installed on the heavy-truck Figure 3 shows some picture of the EPS steering unit installation on the heavy-t r u c k. 5

6 Figure 3 - EPS steering actuator installation on the heavy-truck ROAD-TESTING OF THE LANE VISION SYSTEM The Lane Vision system currently installed on the prototypal vehicle features a CMOS camera with a real time image processing unit, capturing information from the road geometry (vehicle distance from lane-markings, road curvature ahead, boundary type, number of adjacent lanes) and the vehicle trajectory. After installing the camera system inside the cabin, an extensive testing campaign has been conducted on motorways and extra -urban roads, in different weather and lighting conditions, in order to verify the performance on the vision s ystem and to identify the most critical situations for this state-o f-art system. See next picture showing some scenarios and road -situations which resulted critical for the Lane Detection system and where the system didn t recognize correctly the lane markings. Vision disturbed by vehicle in front Interference of external elements Guard -rail considered as lane -marking Badly erased lane marking Figure 4 - Some examples of critical scenarios for the Lane Detection system 6

7 RESULTS The control strategy developed has been implemented on an IVECO Stralis 480AS heavytruck in tractor-semitrailer configuration. The system has been tested in different road and lighting conditions. Figure 5 - The truck used for developing the Haptic Lane Keeping Support system Extensive road -tests of the system have been performed with several professional drivers, who have tested as well a conventional LDW system, providing acoustic warnings imitating a rumble-strip noise. All the professional driver found the acoustic warnings quite annoying, while everybody preferred the haptical intervention of the EPS unit. Furthermore, these tests indicate that drivers accept a certain amount of nuisance from the haptic intervention of LKS system, while they would not accept a similar amount of nuisance due to acoustic warnings. During test ses sions, some drivers asked for a higher amplitude of the active torque, while somebody asked for a lower setting. Following this requests, a selector was added on the prototypal vehicle to allow the drivers to set a personalized value of the active torque, which remains upper limited, to avoid excessive interventions on the steering unit, which could lead to dangerous lateral dynamics of the heavy-t r u c k. In Figure 6 a typical unintentional drift out of the lane man oeuvre is depicted. At time 10s, the truck travels in the centre of its lane (lane width about 4 m). At time 12s, the truck drift its trajectory towards the right boundary. When the warning decision module recognizes the unwanted movement of the vehicle ou t of the current lane and decides to trigger the alarm, the resistant torque requested, calculated in the lateral control module, is sent to the steering wheel device. It reach about its maximum value, equal to 3 Nm, selected by the driver. 7

8 Figure 6 - An example of Haptic Lane Keeping intervention General results from these test drives have confirmed that the developed LKS system provides good support to the driver, in most of traffic conditions, thus reducing the risk of accidents due to road -departure. ACKNOWLEDGEMENT This work has been carried out in the context of the SAFELANE project within the PReVENT research program, funded by the EU FP6. Special thanks are due to IVECO SpA, which has supported this development and has provided technical, logistical and experimental support to CRF team. REFERENCES 1. Consano L., Murdocco V. Haptical Lane Assistant for Heavy Trucks, Truck and Buses: Solutions of reliability, sustainable environment and transport efficiency, VDI, B o b lingen, J. Pohl and J. Ekmark, Development of a haptic intervention system for unintended lane departure, in Proceedings of the 2003 SAE World Congress, Detroit, MI, USA, March Balzer, D, Safety enhancement by lane observation of Advanced Driver Assistance Systems, VDA Technical Congress, Germany, Pilutti, T E, Lateral Vehicle Co -Pilot to avoid Unintended Roadway Departure, PhD thesis, Dept. Mechanical Engineering, The University of Michigan,