Develop Performance Specifications for Frontal Collision Warning System for Transit buses

Transcription

1 Develop Performance Specifications for Frontal Collision Warning System for Transit buses Wei-Bin Zhang 1, Ron DeLeon 2, Frank Burton 3, Brian McLoed 4, Chinyao Chan 1, Xiqing Wang 1, Scott Johnson 5,and Dan Empey 5 Abstract The U.S. Department of Transportation initiated the Intelligent Vehicle Initiative (IVI) program with a goal to improve safety through the application of advanced technologies. The frontal collision warning function has been identified as one of the key safety improvement measures for the transit vehicle platform. Frontal collision, defined as a bus colliding with a vehicle in front of the bus, is a frequent incident in transit bus operations and the cause of property damage, personal injuries and interruption to operations. A team that includes San Mateo County Transit District (SamTrans), University of California PATH Program (PATH), California Department of Transportation (Caltrans), and Gillig Corporation has been selected by the US DOT to develop and validate performance and technical requirement specifications for a Frontal Collision Warning System (CWS) for transit buses. The project began in January 2000 with a planned duration of two years. SamTrans operates a fleet of 316 buses in one of the most congested areas in the United States, which includes the counties of San Mateo, Santa Clara, and San Francisco. Accident statistics tracked by SamTrans in recent years indicate frontal collisions can result in significant property damage and liability. In addition to frontal collision, passenger falls resulting from emergency braking also contribute to an increased potential for passenger injuries and liability. A frontal collision warning system using advanced sensing and computer technologies can potentially reduce frontal collision accident rate thereby to minimize losses and to reduce operational interruptions. The collision warning system may also help the driver to adequately respond to the hazard with smoother 1 Research Engineer, California PATH Program, University of California at Berkeley, 1301 S. 46 th St., Richmond, CA (510) Chief, Advanced Vehicle Systems, California Department of Transportation, (916) Manager, Communications, San Mateo County Transit District, (650) Vice President, Gillig Cooperation, (510) Development Engineer, California PATH Program, University of California at Berkeley, 1301 S. 46 th St., Richmond, CA

2 maneuvers. Furthermore, information collected through sensors can be recorded for the purpose of accident analysis and for avoiding false claims. The goals of the collision warning system for this project are to (a) address imminent crash warning, (b) provide warnings for smoother maneuvering, and (c) provide warnings when a bus is too close to a forward vehicle. While providing imminent crash warnings is the primary goal of the project, we will also investigate potential extension of the collision warning system for facilitating smooth maneuvers and for added safety features. We ll evaluate the data obtained through the data collection task and will determine whether the secondary goals can be achieved within the scope of work defined under this proposal. This paper discusses about the concept of frontal collision warning system for transit buses, the scope of the frontal collision warning project and reports key findings through data collection and analysis. 1. Introduction Work has been conducted in the research of collision warning system for highway vehicles [1-9]. Commercial systems (such as Eaton Vorad) are also available for specific market applications. A typical frontal collision warning can be described in the following schematic as shown in Figure 1. Noise External interference Vehicle status sensing Collision Warning System Traffic Environment Obstacle detection sensing Information Processing Human Machine Interface Warning algorithm Vehicle movement Driver Figure 1. Schematic of Collision Warning Systems (CWS) 2

3 Our initial study of the traffic environment and traffic data for transit buses indicated that transit buses operate at a very complicated environment. Specifically, many transit routes are distributed in urban environment where significant number of unintended obstacles (such as traffic signals or bus stop structures) can present within the field of view of the sensors. Additionally, traffic behavior around the transit bus can be significantly different than that of the highway. Vehicles travel at much closer spacing and accident data shows that many of the frontal collision are due to the vehicles cutting-in front of the bus. Buses also often make turns at intersections where traffic pattern can be very complicated. These challenges can pose different requirements for collision warning for transit buses than that of the systems for highway applications. The literature review of state-of-art technologies conducted by the project team shows lacking of efforts on FCW for transit buses. Previous studies on collision warning and collision avoidance have been focused on passenger cars and highway applications. The literature review and the initial study indicates that a FCWS for transit bus will have higher level of capabilities than the FCWS that are designed for automobiles and trucks at least in two different ways, including: 1) A sensing arrangement must be designed to accommodate complicated traffic patterns such as head-on traffic at intersections and vehicle cut-in at close spacing and 2) The system must be capable of understand the urban operation environment and distinguish the intended targets (vehicles and pedestrians) from the unintended ones (e.g., traffic signs and trees). The project tram has established an advisory committee which includes 10 transit agencies cross San Francisco Bay Area. The advisory committee has played an important role in identifying the industry needs for the collision warning system and providing technical expertise and data to the project. 2. Project Scope of Work The frontal collision warning for transit buses project incorporates twelve tasks covering several key aspects: - Analysis: Through literature review and data analysis, issues involving bus frontal collision will be addressed. Accident cases will be summarized as collision scenarios. Examples of the scenarios are: (a) collision with a vehicle travelling in the bus forward path, (b) collision with a lane changing vehicle cutting into the bus path, and (c) frontal collision while the bus is making a turn at an intersection. - Data collection: Data acquisition system will be installed on buses to collect realworld data for in-depth understanding of typical scenarios prior to an accident and the response of a driver to the proposed warning device, the conditions under which the CWS must operate and help in the analysis and design of the warning system. 3

4 - Design Collision Warning Scheme and Algorithm: A collision warning system will be developed for validation and verification of the requirement specifications. Emphasis will be placed in the development of the signal processing algorithm that can handle obstacle detection in urban environment and Human Machine Interface that will be effective in collision warning for buses. Field tests will be conducted to verify and validate the performance requirements using the buses equipped with the CWS. - Development of requirement specifications: Preliminary and final requirement specifications will be developed at different stages of the project. This will include some recommendations on the method used to alert the driver of the need for collision avoidance action. The effectiveness of the presentation and the associated distraction will be addressed. The potential for false alarms and consequences will also be included in the specification. The final requirement specifications will include System Performance Specifications, Component Technical Requirements and Interface Requirements. 3. Data Collection and Analysis In order to understand the nature of accidents, the frontal collision team conducted a thorough investigation of the historical accident data involving transit buses. In addition to the data assembled by Samtrans, three additional Bay Area transit agencies also provided accident data from 1995 to The data covers a geographically diverse region with samplings from urban, suburban, and rural bus routes. The accidents were screened, reviewed, and entered into PATH database. The data analysis confirmed that frontal collisions constitute a significant portion of all accidents. Frontal collisions can be broken down into the following categories: collision with a vehicle cutting into the bus path, collision with a vehicle in the bus path, collision with vehicles backing into the bus, ccollision with obstacles other than vehicles, and collision with people. The speed of buses prior to the accident occurrence was generally modest. For incidents near bus stops, traffic lights, or intersections, the speed is usually lower. Among the reviewed accident reports, many incidents involved the bus making a contact with a neighboring vehicle at the front corners at relatively low speeds. There are a significant number of on-board passenger incidents in all transit agencies with many of these incidents associated with bus braking or stopping. The accident data provides a knowledge base for determining the type and frequencies of frontal collision accidents. However, because transit accident data are heavily relying on the recollection of the involved operator, most data may not accurately describe the cause and the time sequence of the accidents. 4

5 In order to better understand the bus operating environment and time sequence of potential accidents, a data collection system is developed. The data collection system includes two radars and a laser radar (lidar) for detecting range and range rate of a frontal obstacle and several ultrasonic sensors for detecting corner obstacles. An additional radar is installed at in rear bumper of the bus facing backwards to detect the movement of the following Corner Sensor Rear-facing Radar Video Radars and Lidar Corner Sensor Figure 1. Sensor Arrangements of the Data Acquisition System 5

6 vehicle in order to understand the impact to the movement of the following vehicle due to the deceleration of the bus. Video sensors are installed to record visual data covering the field of view of the frontal, corner, and rear-facing sensors. The video data will be used in conjunction with the sensors data to analyze the bus operating environment and to understand sensor behaviors. Another video sensor will be facing inward to record passenger movements in order to correlate the vehicle motion and passenger falls. Figure 1 shows sensor arrangements of the data acquisition system. In addition to the obstacle detection sensors, vehicle sensors are also installed to monitor steering angle movement, brake and throttle motion, and vehicle velocity and acceleration. Figure 2 depicts the data acquisition system. The data acquisition was first installed on a 4X4 vehicle for testing and debugging and later was moved to a transit bus. Figure 3. shows a radar, a lidar and two ultrasonic sensors installed under the front bumper of the bus. Velocity Driver side radar Passenger side radar Lidar Left corner sonar Right corner sonar Removable recording media Data Recording and Processing Computer Video Recording Synchronization Acceleratio n Steering angle Throttle movements Brake movements GPS Obstacle detection sensors Front center Left corner Right Rear facing Passenger Vehicle sensing Video Figure 2. Schematic Diagram of the Data Acquisition System Three additional buses will be instrumented with similar data acquisition systems. These buses will be put in regular service on carefully designed routes to collect data. It is understood that the likelihood of the buses instrumented with the data acquisition system run into accidents will be extremely small. However, abundant data collected on these buses will provide accurate description of the relative movement of the buses and the 6

7 surrounding vehicles and imminent crashes. This data can be imported to a vehicle dynamic model so that collision course can be predicted. EV 300 Radar Denso Lidar Ultrasonic Corner Sensors Figure 3 Sensors Installed on a Bus The test vehicle has been driven on highways and local streets to collect data. Initial data indicated that traffic signs and overhead structures on highways provide very little interference to the sensors. However, as predicted, both the radars and the lidar detect significant unintended obstacles. Figures 1 to 3 show a set of data collected by sensors (the driver side radar, the passenger side radar, and the lidar) through a run at an intersection. Each set of data has range, range rate and azimuth. The test vehicle entered into the intersection between 0 and 9 seconds. After a stop (9-12 secs), the vehicle made a right turn (13-17 secs) then continued going straight until reaching the next intersection (18-30 secs) and stopped for the remaining duration. The route the test vehicle turn into has a center dividing island where there are a traffic sign, a tree, and a utility box. From the data one can see that all three sensors detected the traffic sign, the tree and the utility box. Because the range is very close and range rate is relatively large, a warning can potentially be triggered unless the target is either identified as an unintended or not being in the vehicle path. Therefore, a sensing scheme and a target identification algorithm must 7

8 be ' 2)33,*+,4-567,-()7)4 ()*+,-./01 & $ ()*+,84)0,-./0931 <=6;>0?-.@,+1 " 8" 8# $ # 8# :6;,-.31 Figure 1. Passenger Side Radar ()*+,-./01 & $ ()*+,84)0,-./0931 <=6;>0?-.@,+1 " 8" 8# # 8# 8$ :6;,-.31 8

9 Figure 2. Driver Side Radar #" B67)4 ()*+,-.;1 ()*+,84)0,-.;931 B)0,4)C-DE3606E*-.;1 # " # 8# 8$ # 8# :6;,-.31 Figure 3. Lidar incorporated to deal with this scenario. The literature review indicates that this type of scenarios has not dealt with thoroughly in the past. While data is continuously collected, more scenarios will be identified. Sensing schemes and target identification algorithms will be developed under this project. Concluding Remarks The development of requirement specifications for frontal collision warning system is one of the key elements of the federal IVI program. The FCWS team combines expertise from transit industry, bus manufacturer, state DOT and research organization with support from Greater Bay Area transit agencies. To date, work has been focusing on defining industry needs/desires, assessing state-of-art knowledge and technologies, and understanding transit operating environment. In the next a few months, a total of four transit buses will be instrumented. Significant amount of data will be collected and processed. The data analysis will help to identify an effective sensing scheme and the to develop a signal processing and warning algorithm. This final product requirement specification of the frontal collision warning algorithm will be developed by the end of

10 Acknowledgements This project is sponsored by the U.S. Department of Transportation Federal Transit Administration under the National Intelligent Vehicle Initiative Program. The authors of this paper would like to express our appreciation to Brian Cronin of FTA for his guidance and support, Dave Nelson, Paul Kretz, Aaron Steinfeld and Steven Shladover of PATH, Matt Hanson of Caltrans, Chuck Harvey, Jerie Moeller and James Castagno of Samtrans and Gail Ewing of Gillig for their technical assistance and support, and the Advisory Committee members for their active participation of the project. The project team would also like to thank Denso Co. for their contribution of two Lidars for this project. References 1. Frontier Engineering, (1994). IVHS Countermeasures for Rear-End Collisions, Task 1, DOT HS Frontier Engineering, (1995). IVHS Countermeasures for Rear-End Collisions, Peer Review Workshop Summary, DOT Contract DTNH22-93-C-07326,. 3. Burgett, A, et al., (1998). A Collision Warning Algorithm for Rear-End Collisions, 16 th International Technical Conference on Enhanced Safety of Vehicles (ESV), Windsor, Canada, Paper No. 98-S2-P (1998). Notebook of NHTSA Symposium on Rear-End Collision Avoidance including Intelligent Cruise Control, The John Hopkins University, Laurel, Maryland. 5. Fancher, P., et al., (1998). Intelligent Cruise Control Field Operational Test, (Final Report, UMTRI-98-17), prepared by The University of Michigan Transportation Research Institute, for National Highway Traffic Safety Administration. 6. McGehee, D.V., and Dingus, T.A., (1992). The Potential Value of A Front-to- Rear-End Collision Warning System Based on Factors of Driver Behavior, Visual Perception and Brake Reaction Time, Proceedings of the Human Factors Society 36 th Annual Meeting, pp Yoo, H.; Hunter D., and Green, P. (1996) Automotive Collision Warning Effectiveness: A Simulator Comparison of Text vs. Icons, (Technical Report UMTRI-96-29), Ann Arbor, MI: The University of Michigan Transportation Research Institute. 8. Chong, M.; Clauer, T.; and Green, P. (1990). Development of Candidate Symbols for Automobile Functions, (Technical Report UMTRI 90-25), Ann Arbor, MI: The University of Michigan Transportation Research Institute. 9. Hoekstra, E.; Williams, M.; and Green, P. (1993). Development and Driver Understanding of Hazard Warning and Location Symbols for IVSAWS, (Technical Report UMTRI 93-16), Ann Arbor, MI: The University of Michigan Transportation Research Institute. 10