Chance-Constrained and Risk-Bounded Hierarchical Path Planning for Autonomous Driving in Safety-Critical Scenarios

Safety-critical path planning is a core capability for autonomous driving systems operating in dense traffic, adverse weather, and rare interaction scenarios. Existing research has made substantial progress in scenario-based validation, safety-critical scenario generation, and simulation-driven assurance, yet many studies focus either on test scenario production or on nominal planning performance rather than a unified risk-bounded planning architecture. This paper presents a safety-critical path planning framework that integrates semantic environment understanding, uncertainty-aware motion prediction, and risk-bounded trajectory optimization. The method combines a global risk-aware graph search, a local chance-constrained trajectory optimizer, and an emergency safety shield that enforces minimum separation and collision-imminence constraints. In contrast to purely shortest-path or comfort-oriented planners, the proposed framework explicitly optimizes expected travel efficiency under bounded tail risk. We also define a composite criticality score that fuses time-to-collision, predicted occupancy overlap, semantic hazard priors, and perception uncertainty. A literature review is provided to position the method within scenario-based validation, critical scenario generation, safety metrics, and autonomous driving assurance. Illustrative experiments on urban intersection, ramp merging, and adverse-weather scenarios show that the proposed method reduces collision rate and high-risk exposure at modest efficiency cost. All numerical results in this manuscript are reference values intended for drafting and demonstration rather than submission-ready claims.