Robustness under stochastic observations and externally authored contact dynamics

Investigate the behavior of the sampling-verification danger law under process noise, unknown observation-noise laws, planning under noise, and externally authored contact or saturation benchmarks with their own critical events, and establish whether the reported conclusions persist in those settings.

Background

The experiments use deterministic, hand-designed hybrid instruments and measure only bounded observation noise with a known coordinatewise uniform distribution. The paper therefore does not establish whether its gate-miss behavior, verified-but-wrong artifacts, and planner exploitation results persist when the dynamics themselves are stochastic, when the noise distribution is unknown, or when planning occurs under noise.

The authors also identify the need to test an externally authored contact or saturation benchmark with its own critical event. Such a benchmark would assess whether the reported phenomenon is a property of the proposed verification pipeline more broadly rather than an artifact of instruments designed around the danger law. The authors describe this benchmark as the decisive follow-up and note that its analysis should be frozen before outcomes are inspected.