Autonomous vehicle navigation around human pedestrians remains a challenge due to the potential for complex interactions and feedback loops between the agents. As a small step towards better understanding of these interactions, this Methods Paper presents a new empirical protocol based on tracking real humans in a controlled lab environment, which is able to make inferences about the human’s preferences for interaction (how they trade off the cost of their time against the cost of a collision). Knowledge of such preferences if collected in more realistic environments could then be used by future AVs to predict and control for pedestrian behaviour. This study is intended as a work-in-progress report on methods working towards real-time and less controlled experiments, demonstrating successful use of several key components required by such systems, but in its more controlled setting. This suggests that these components could be extended to more realistic situations and results in an ongoing research programme.
[1]
Natasha Merat,et al.
When Should the Chicken Cross the Road? - Game Theory for Autonomous Vehicle - Human Interactions
,
2018,
VEHITS.
[2]
Huosheng Hu,et al.
Computationally efficient solutions for tracking people with a mobile robot: an experimental evaluation of Bayesian filters
,
2010,
Auton. Robots.
[3]
Natasha Merat,et al.
Empirical game theory of pedestrian interaction for autonomous vehicles
,
2018
.
[4]
Zhi Yan,et al.
Online learning for human classification in 3D LiDAR-based tracking
,
2017,
2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).