Pedestrian Collision Avoidance System for Scenarios with Occlusions

Abstract

Safe autonomous driving in urban areas requires robust algorithms to avoid collisions with other traffic participants with limited perception ability. Current deployed approaches relying on Autonomous Emergency Braking (AEB) systems are often overly conservative. In this work, we formulate the problem as a partially observable Markov decision process (POMDP), to derive a policy robust to uncertainty in the pedestrian location. We investigate how to integrate such a policy with an AEB system that operates only when a collision is unavoidable. In addition, we propose a rigorous evaluation methodology on a set of well defined scenarios. We show that combining the two approaches provides a robust autonomous braking system that reduces unnecessary braking caused by using the AEB system on its own.

Set-Based Prediction of Traffic Participants Considering Occlusions and Traffic Rules

“…Set-based prediction is also able to consider interaction between traffic participants [69] and formalized traffic rules [14], [70]. The predicted occupancy sets can also be weighted by probabilities [71], [72] d) Occlusion: The risk from occlusions is tackled either by shrinking the field of view over the prediction horizon [73]- [76] or by introducing and predicting individual, potentially present obstacles (aka phantom or virtual objects) [1], [16], [77]- [85]. Early works considering occlusions are motion planners for mobile robots [86], [73]- [75].…”

“…In recent motion planners, a partially observable Markov decision process optimizes the behavior of the ego vehicle such that the collision risk due to occlusions is reduced [81]- [84]. In a pedestrian collision avoidance system, a partially observable Markov decision process propagates the belief states of occluded pedestrians based on reachable sets [85]. The occlusion-aware motion planner in [87] remains collision-free in specific traffic situations for which the authors have manually defined the worst-case.…”

Integration of a Human-aware Risk-based Braking System into an Open-Field Mobile Robot

“…According to the obstacles ' positions, a partially observable Markov decision model braking approach estimates its braking states [3]. Similarly, time-to-brake and time-to-collision indicators can represent a safe evaluation metric [18].…”

Pedestrian Emergence Estimation and Occlusion-Aware Risk Assessment for Urban Autonomous Driving

“…As an alternative to the interaction-aware methods, occlusion-aware [5] risk assessment methods have been proposed. Those alternatives vary from solutions based on partially/mixed observable Markov decision processes (POMDP/MOMDP) [6]- [9] , to solutions based on set-based methods [10]- [15]. Set-based methods and exploitation of behavior of other participants in a rule-based fashion were used [16]- [20].…”