Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces Risk-Sensitive Conjectural Scenario Planning (RCSP), a planning layer that reasons about short-horizon obstacle futures to avoid predictive near-miss commitments in robot navigation. RCSP maintains a belief over local motion conjectures, samples future interactions, and penalizes high-risk outcomes. Experiments in MuJoCo and ROS2/Gazebo show that RCSP reduces dynamic near-miss failures compared to non-adaptive predictors, although it doesn't consistently outperform tuned DWA and TEB on strict benchmark success in DynaBARN/Jackal transfer.
Robots can avoid collisions, but still fail by committing to trajectories that become unsafe; RCSP offers a way to plan with "predictive risk" in mind.
Mobile robots can fail before they collide: a velocity that is safe now may commit the robot to a passage that moving obstacles will soon close. We study this predictive near-miss commitment problem and propose Risk-Sensitive Conjectural Scenario Planning (RCSP), a planning layer that evaluates candidate commands against plausible short-horizon obstacle futures. RCSP maintains a lightweight belief over local motion conjectures, samples future interactions, penalizes high-risk tails, and executes through a local safety check. In controlled MuJoCo bottleneck tasks, the RCSP planner reaches the goal without collisions and yields higher secondary safety and path-quality point estimates than a non-adaptive predictor, with additional latency. In ROS2/Gazebo, adding the local safety layer to a standard Nav2 stack reduces dynamic near-miss failures. On official DynaBARN/Jackal transfer, tuned DWA and TEB remain stronger on strict benchmark success, revealing the boundary of the approach. These simulation results position RCSP as a predictive-risk module that complements existing navigation stacks in dynamic bottleneck regimes.