Occlusion-aware trajectory search improves hidden vehicle detection
Worst-Case Hidden-Vehicle Trajectory Search in Spatiotemporal Occlusion Regions
RoboticsCryptography and Security
Summary
When self-driving cars cannot see other vehicles or pedestrians due to obstacles, they face uncertainty about what might be hidden. The authors developed a method that looks back through several frames of video to better guess the possible paths of hidden vehicles. Their method finds the worst-case hidden vehicle trajectories that could cause problems, helping the car plan safer moves. Tests on real driving data showed it reduced uncertain hidden objects significantly and found scenarios that could lead to collisions.
What this means in practice
- •For autonomous vehicle engineers: Improve path planning by accounting for worst-case hidden vehicle trajectories within occlusion regions.
- •For robotics software developers: Use multi-frame occlusion analysis to enhance environment understanding and safety in mobile robots navigating dynamic settings.
Tested on one dataset.
Authors
Ruichen Tan, Zengxiang Lei, Satish Ukkusuri
Abstract
Occlusion creates fundamental uncertainty in autonomous driving. Existing methods often propagate frame-wise hypotheses or optimize ego behavior against prescribed hidden-agent predictions, leaving the worst history-consistent interaction unexplored. We introduce History-Conditioned Minimax Trajectory Search (HC-MTS), which combines temporal occlusion reasoning with response-aware search. First, HC-MTS constructs finite hidden-state modes, each certified by a backward witness satisfying multi-frame visibility, occupancy, semantic-map support, and class-specific kinematic constraints. It then solves a bilevel minimax problem: an inner finite oracle maximizes the ego driving score over destination attainment and ride comfort, while the outer search selects the legal hidden-vehicle trajectory that minimizes this best-response value. Across eight Waymo Open Motion Dataset scenarios, increasing the visibility-memory horizon from K=1 to K=20 reduces the mean per-scenario vehicle, pedestrian, and total retained hidden-seed counts by 18.12%, 21.67%, and 18.45%, respectively. HC-MTS identifies six avoidable counterexamples, while no legal collision-producing attacker is found in the remaining two scenes within the finite search budget.