robotics
Uncertainty Moves Upstream of the Vehicle Planner
A modular pipeline converted collision probabilities into convex safe sets and reported zero simulated collisions and zero solver failures at 12.7 milliseconds per cycle.
Summary
A modular pipeline converted collision probabilities into convex safe sets and reported zero simulated collisions and zero solver failures at 12.7 milliseconds per cycle.
Instead of embedding all probabilistic risk inside the optimizer, the system propagates interpretable uncertainty through perception and prediction, then hands the planner a convex safe set with a feasible solution each cycle. Monte Carlo intersection trials beat deterministic and chance-constrained baselines, and Clearpath Jackal hardware confirmed real-time execution. The zero figures apply to the reported trials, not public-road safety.
Why it matters
A modular pipeline converted collision probabilities into convex safe sets and reported zero simulated collisions and zero solver failures at 12.7 milliseconds per cycle.
Limits and context
- The zero figures apply to the reported trials, not public-road safety.
Key claims
A modular pipeline converted collision probabilities into convex safe sets and reported zero simulated collisions and zero solver failures at 12.7 milliseconds per cycle.
Qualification: The zero figures apply to the reported trials, not public-road safety.
Evidence: source-2026-10-27-019
Sources
- arXiv preprint 2610.33186, original v1arXiv · primary research
Corrections
No corrections have been recorded for this story.