Abstract

This dissertation develops methods for planning safe and efficient trajectories for autonomous vehicles operating in environments containing adversarial and environmental threats with uncertain parameters. The central objective is to quantify and control the risk of adverse events--such as interception or detection--when threat characteristics are imperfectly known. To address this challenge, probabilistic weapon engagement zone (PEZ) models are introduced to estimate the likelihood of violating adversarial engagement constraints under uncertainty in pursuer state and capabilities. Multiple computational strategies are developed for PEZ evaluation, including first-order sensitivity-based approximations, Monte Carlo estimation, and neural-network surrogate models. These approaches enable rapid engagement-probability assessment and allow risk constraints to be embedded directly within trajectory optimization and path-planning frameworks. The engagement modeling framework is extended to turn-rate-limited pursuers through the formulation of curve--straight probabilistic engagement zones, capturing realistic kinematic constraints while preserving computational tractability. For scenarios characterized by bounded rather than probabilistic uncertainty, set-based engagement zone formulations are derived to provide conservative safety guarantees. Interception outcomes from low-priority sacrificial agents are leveraged to infer feasible regions for unknown pursuer launch locations, producing engagement models consistent with observed data. Learning-based techniques are further developed to update threat parameters from interception events and to design informative sacrificial-agent trajectories that accelerate inference while safeguarding high-priority assets. The proposed methods are validated through simulation studies demonstrating reduced engagement risk and improved mission performance relative to deterministic and nominal planning strategies. Finally, the framework is applied to radar detection avoidance and environmental hazard estimation, illustrating that the underlying risk-aware planning and learning principles generalize beyond weapon engagement scenarios.

Degree

PhD

College and Department

Ira A. Fulton College of Engineering; Electrical and Computer Engineering

Rights

https://lib.byu.edu/about/copyright/