Corridor conditioned risk modeling improves safety in trajectory planning
CorrRisk-WM: Corridor-Conditioned Risk World Modeling for Safety-Critical Trajectory Planning
Robotics
Summary
Planning safe paths for self-driving cars needs understanding how other agents move and what risks these movements pose. The authors created CorrRisk-WM, a model that predicts how other agents behave in the environment and estimates safety risks for different possible paths the car might take. Their experiments on a large set of real driving scenarios show this model is better at detecting potential collisions or close calls than existing methods. This approach helps cars pick safer routes by considering both movement predictions and the specific geometry of each path.
What this means in practice
- •For autonomous vehicle engineers: Improve on-road local path planning by accurately estimating collision risks along candidate routes using combined environment and trajectory models.
- •For robotics navigation teams: Use corridor-conditioned risk assessments to enhance safe motion planning in dynamic, multi-agent environments beyond just cars.
Authors
Tingyu Guo, Reza Langari
Abstract
Safe local planning requires forecasting surrounding-agent motion and evaluating candidate-specific risks, since identical agent motion can pose different risks to different ego trajectories. We present CorrRisk-WM, a planning-oriented partial world model coupling environment evolution with supervised intrusion and near-miss prediction over bounded candidate-trajectory corridors. A latent environment model recursively predicts agent states and updates agent-agent and agent-map interactions. Each candidate queries the evolving environment through footprint- aware geometry and learned agent-corridor representations. A lightweight recurrent risk module uses temporal context to estimate per-slice hazards; survival aggregation yields first-entry and horizon-level event probabilities. On 29,176 scenarios from 100 Waymo validation shards, CorrRisk-WM achieves intrusion average precision (AP) of 0.8567 and 1-m near-miss first-entry AP of 0.8671. In baseline comparisons, it attains the highest near-miss AP at all three distance thresholds and the lowest observed open-loop collision rate (4.88%), with route progress of 15.35 m. Across three seeds, removing dynamic environment modeling or candidate-conditioned geometric interaction reduces mean intrusion AP from 0.8590 to 0.7624 and 0.7252, respectively. These results support coupling environment evolution with candidate-conditioned geometric reasoning for risk prediction and safety-oriented candidate selection.