Semi-Explicit Solutions to the Prying-Pedestrian Surveillance-Evasion Differential Game and Extensions to Two Pursuers

arXiv:2607.21087 2026 Architecture 1 ideas extracted · analyzed Aug 30, 2026

What the math gives to ML

The paper provides a constructive geometric solution to a two-pursuer/one-evader differential game by reducing optimal motion to a small set of turn-straight branches and selecting the branch with minimum completion time. Its transferable asset is a reachable-set planning layer: tangent points to circular exclusion or surveillance regions can be computed analytically, while discrete branch switching handles competing pursuers. This can be embedded as a safety and planning module around a neural policy, providing hard geometric constraints and an explicit branch-switching signal for training.

Ideas from this paper

Mechanism confirmed, baseline not beaten 2026

Tangent-Branch Neural Evasion Layer

Wrap a neural multi-agent policy with an analytic planner that generates turn-straight trajectories tangent to pursuer surveillance disks, then selects the branch with the smallest predicted completion time. The network supplies high-level preferences or residual corrections, while the geometric layer prevents unnecessarily entering exclusion regions and exposes an explicit branch-switching signal for training.

Useful7/10
Difficulty5/10
Novelty7/10
Paper: Semi-Explicit Solutions to the Prying-Pedestrian Surveillance-Evasion Differential Game and Extensions to Two Pursuers arXiv:2607.21087