Situation analysis and decision making for active pedestrian protection using Bayesian networks

One of the major challenges in advanced driver assistance systems is the interpretation of available environment information. It is the foundation for system activation strategies and decision making. Often deterministic motion models are used to predict pedestrian movements, which leads to constricted validity. Investigations have shown that uncertainty is not negligible in pedestrian models due to their high dynamic range. Standard concepts of decision making are not able to deal with uncertain motion models. Decision making gets even more difficult if different emergency maneuvers can be selected, i.e. emergency braking, evasive steering or a combination of both. The benefit of each of these maneuvers depends highly on the future position of the pedestrian. An appropriate maneuver can hardly be selected based on a deterministic pedestrian model. Here, a probabilistic approach to situation analysis based on a pedestrian model with uncertainty is suggested. An emergency maneuver is selected considering the optimal injury risk reduction for the pedestrian.

[1]  H. Winner,et al.  Fahrerassistenzsysteme. Stand der Technik und Ausblick , 2002 .

[2]  Thao Dang,et al.  Handling uncertainties in criticality assessment , 2011, 2011 IEEE Intelligent Vehicles Symposium (IV).

[3]  Wolfgang Rosenstiel,et al.  Maneuver decision for active pedestrian protection under uncertainty , 2013, 16th International IEEE Conference on Intelligent Transportation Systems (ITSC 2013).

[4]  Kuo-Chu Chang,et al.  Hybrid message passing for mixed bayesian networks , 2007, 2007 10th International Conference on Information Fusion.

[5]  Matthias Althoff,et al.  Model-Based Probabilistic Collision Detection in Autonomous Driving , 2009, IEEE Transactions on Intelligent Transportation Systems.

[6]  Rolf Schmidt,et al.  Ein fortgeschrittenes Kollisionsvermeidungssystem , 2005 .

[7]  Dariu Gavrila,et al.  Will the Pedestrian Cross? Probabilistic Path Prediction Based on Learned Motion Features , 2011, DAGM-Symposium.

[8]  Michael Brost,et al.  Action Recognition and Prediction for Driver Assistance Systems Using Dynamic Belief Networks , 2002, Agent Technologies, Infrastructures, Tools, and Applications for E-Services.

[9]  Steffen L. Lauritzen,et al.  Stable local computation with conditional Gaussian distributions , 2001, Stat. Comput..

[10]  Jörg Henning Schneider.  Modellierung und Erkennung von Fahrsituationen und Fahrmanövern für sicherheitsrelevante Fahrerassistenzsysteme , 2009 .

[11]  Robert G. Cowell,et al.  Local Propagation in Conditional Gaussian Bayesian Networks , 2005, J. Mach. Learn. Res..

[12]  Stephan Zecha,et al.  Optimaler Fussgaengerschutz durch situationsgerechte Einschaetzung der Fussgaengerbewegung , 2008 .