Interception-Driven Inverse Reachability for Engagement Zone Construction

Grant Stagg, Cameron K. Peterson, Alexander Von Moll, Isaac Weintraub

In contested environments, autonomous vehicles may need to plan around adversarial pursuers whose launch locations are unknown. This paper presents an interception-driven inverse-reachability framework for inferring a feasible pursuer launch region directly from observed interception events for a single pursuer. Each interception induces a geometric constraint on the unknown launch location, and intersecting these constraints yields a bounded set guaranteed to contain the true origin under maximum-capability assumptions. Mapping this inferred set through the pursuer reachable region produces deterministic engagement zones with an explicit worst-case safety interpretation. A probabilistic extension models uncertainty in the pursuer launch location and yields graded engagement-risk fields for risk-aware planning. To accelerate localization, we introduce an information-driven planner for sacrificial agents that selects trajectories to maximize expected contraction of the feasible launch region. Monte Carlo simulations show that the proposed framework rapidly reduces launch-location uncertainty and enables substantially shorter safe trajectories after only a small number of sacrificial deployments.