Learning Material-Aware Hamiltonian Risk Fields for Safe Navigation

2026-05-07

AI Summary

This work addresses the problem of unsafe or ineffective maneuvers in risk-aware navigation caused by indiscriminate obstacle avoidance. It proposes a selectively activated port-Hamiltonian navigation strategy that constructs a risk field through a contextual energy term, generating guidance forces only when feasible low-risk paths exist and suppressing lateral forces otherwise to prevent illusory avoidance behavior. The risk field features a falsifiable selective signature, with its activation mechanism derived from the gradient structure of contextual energy rather than empirical tuning, and integrates Conditional Value-at-Risk (CVaR) optimization to emphasize tail-end critical risks. By fusing semantic maps with real-world terrain perception, the method significantly improves performance across multiple benchmarks: success rate in delayed escape tasks rises from 0.480 to 0.810, spurious activation on real terrain drops to 0.114, catastrophic failures on static maps decrease from 0.60 to 0.10, and collision rates in high-speed scenarios are reduced to zero.

Abstract

Risk-aware navigation should be selective: a policy should expose evasive degrees of freedom only when the local scene admits a lower-risk feasible maneuver, and suppress them when no safer alternative exists. We show that adding one context-energy term to a port-Hamiltonian navigation policy produces a learned force channel with exactly this falsifiable signature. When the local risk field contains a feasible lower-risk direction, the induced context force activates toward it; when the apparent escape is blocked or not yet available, a route-aware gate suppresses lateral force rather than hallucinating an unsafe maneuver. A CVaR tail-risk objective focuses gradient updates on rare but consequential risk transitions. We validate the selectivity signature across four settings. In the primary delayed-required-escape benchmark, route-aware CVaR reduces premature force activation from 0.950 to 0.180 versus DWA while raising success from 0.480 to 0.810 with zero replans. On real off-road terrain (RELLIS-3D), route-aware enrichment achieves correct activation rate 0.837 and false activation rate 0.114, compared to 0.378/0.752 for scalar risk gradients. On static semantic maps (DFC2018), enrichment reduces catastrophic failure from 0.60 to 0.10 and oscillation by 90.7% while preserving path efficiency. In highway traffic, collisions drop from 100% to 0% when a lane escape is feasible; when no escape exists, the policy suppresses the lateral maneuver. The selectivity property follows from the gradient structure of the context energy rather than from training-time tuning.

Problem

• risk-aware navigation
• selective evasion
• feasible maneuver
• tail-risk
• collision avoidance

Innovation

• Hamiltonian risk fields
• risk-aware navigation
• context-energy term
• CVaR optimization
• selective force activation