“Pedestrian in the loop”: An approach using virtual reality

Abstract

Abstract-A large number of testing procedures have been developed to ensure vehicle safety in common and extreme driving situations. However, these conventional testing procedures are insufficient for testing autonomous vehicles. They have to handle unexpected scenarios with the same or less risk a human driver would take. Currently, safety related systems are not adequately tested, e.g. in collision avoidance scenarios with pedestrians. Examples are the change of pedestrian behaviour caused by interaction, environmental influences and personal aspects, which cannot be tested in real environments. It is proposed to use Virtual Reality techniques. This method can be seen as a new Pedestrian in the Loop testing procedure.

“Pedestrian in the Loop”: An Approach Using Augmented Reality

“…causes are unknown). In [6] a new test environment was introduced as a randomized control experiment with the incorporation of virtual reality technologies. The advantage is that real test persons can be incorporated in an experiment (Pedestrian in the Loop).…”

“…Engineers could incorporate safety critical systems for performance testing in real world scenarios which would help accelerate the transition to autonomous vehicles. Problem of the approach in [6] is that the whole perception of the test person has to be stimulated. Computationally expensive rendering and realistic modelling is not possible or too costly in many cases.…”

“…4). As described in [6] motion capture systems and other technologies can be incorporated to test the interaction, environmental influences and personal aspects in safety critical scenarios. Examples are the position of the pedestrian and the orientation, where the pedestrians looks.…”

“…The data explains the vehicle, why a pedestrian moves on this direction in a certain environment. Other advantages of this approach compared to the virtual reality solution [6] is that real environments can be incorporated and cost intensive computations for visualization of static obstacles (e.g. buildings) are not necessary.…”

“…• Possible and adequate for all environments and situations [1] • Meaningful metrics (e.g. measures for the safety-riskratio) and suitable description forms • Measures for robustness and redundancy for safety reasons • Adequate for testing realistic driving scenarios [4] • Comparison to human performance [5] This document is intended as an extension of [6]. The key point of this study is to collect data of pedestrian behavior, analyze the bevavior in different environments and with different target groups.…”