Safety-constrained RL paper validates reachability-checked navigation on Clearpath Jackal

Evidence notes
- The paper combines CVaR-constrained reinforcement learning with post-training reachability verification for mobile-robot navigation.
- The authors validate sim-to-real transfer on a physical Clearpath Jackal robot after evaluating tail-risk behavior beyond average cumulative cost.
Company context
Clearpath Robotics builds unmanned ground vehicles and mobile robot platforms for research, development and field robotics, including Jackal, Husky and related robot bases.