F2008-08-126
Towards a Potential Field Framework for Collision Avoidance Systems
Towards a Potential Field Framework for Collision Avoidance Systems
Collision avoidance belongs to the fundamental behaviors of future cars. The progress in environmental sensor technologies and perception algorithms is one of the corner stones to this end. Another key ability is cognitive information processing to interprete traffic situations, to generate collision avoidance behavior, to decide between possibilities and to perform the appropriate actions on the vehicle maneuver and stability level. Besides the pure technological challenges, it is also important to consider how to design the interaction between the driver and the assistance system.
In this paper we focus on collision avoidance behavior generation. This can e.g. include rule-based and planning-based behavior modules. Collision avoidance approaches so far, like [1] or [2], concentrate on one relevant object. The collision avoidance path is determined geometrically with respect to the relevant object and a control unit guides the car along it and simultaneously brakes the car until standstill. In autonomous driving research, like [3], rule-based approaches are dominant and path planning is considered in the last stage when a collision free driving corridor has already been computed. In all this approaches, the driver is not considered to be part of the control loop. Also, collision avoidance is an isolated function which is initialized in the last possible time instant.
In our approach we go a different way. In [4] a spatially and temporally predictive motion planning algorithm based on potential fields has been proposed, where a dynamic hazard map incorporates future locations of the surrouding objects. Evasion pathes are planned without the need of rules, only based on the information of the dynamic hazard map. This approach automatically includes further modes of assistance such as lane-keeping or lane-changing assistance for a smooth and simple driving mode transition. Depending on the level of assistance, the driver can influence the planned path via haptic feedback through the steering wheel, the brake or throttle pedal. First experiments with test drivers in a fixed-based driving simulator are promising, [5]. Meanwhile, first tests of simple features of this behavior generation approach have been experimetally verfied with a test vehicle, [6].
New in our contribution is the combination of a potential field motion planning behavior with rule-based behavior. We give an overview of our suggested framework, where the need for rule-based decisions is outlined with respect to possible driving strategies and behaviors in cooperation with the driver. Then the system architecture at the information processing level is described, which includes strategical, tactical and reactive components. Contrary to the classical three-layer hierarchy, our motion planning modul is embedded in both the tactical and the reactive level. Rules are stated for behavior decisions and the connection between the rules and the potential field based motion planning is described. Simulation based analysis is presented for particular hazardous traffic scenarios.
References [1] Bender, E.; Darms, M.; Schorn, M.; Stählin,U.; Isermann, R; Winner, H.; Landau, K., [2007], Anti Collision System Proreta - On the Way to the Collision Avoiding Vehicle - Part 1: Basics of the System, ATZ 04/2007 (109), pp. 336-341 [2] Kirchner, A.; Krüger,K.; Mildner,F.; Schmidt An Advanced Collision Avoidance System, ATZ 01/2005 (107), pp. 60-67 [3] Pellkofer, M.; Dickmanns, E.D., [2002], Behavior decision in autonomous vehicles, IEEE Int. Vehicle Symposium, Versaille, France, 17-21 June 2002, pp. 495 - 500 [4] Brandt, T.; Sattel, T.; Wallaschek, J., [2006], On automatic collision avoidance systems, SAE Transactions Journal of Passenger Cars, Electronic and Electric Systems, ISBN 0-7680-1693-2, February 2006, pp. 431-441 [5] Brand, T.; Sattel, T.; Böhm, M., [2007], Combining haptic human-machine interaction with predictive path planning for lanekeeping and collision avoidance maneuvers, IEEE Int. Vehicle Symposium, Istanbul, June 13-15, ISBN: 1-4244-1068-1, pp. 582-587 [6] Sattel,T.; Hesse,T.; Sondermann-Wölke,C., [2008], Experimentelle Untersuchung von Ausweichmanövern mit einer potentialfeldbasierten Assistenzfunktion zur Kollisionsvermeidung, Steuerung und Regelung von Fahrzeugen und Motoren - AUTOREG 2008, Baden-Baden, to appear
Poster presentation: Vehicle safety