Implementation of D* Lite Algorithm for Dynamic Pathfinding in a Street Environment

This project implements the D* Lite algorithm for dynamic pathfinding in a simulated urban street environment featuring realistic autonomous vehicle constraints. The AV, modeled as a 23 grid-aligned car, navigates forward through a narrow 10100 grid with both static and probabilistically generated dynamic obstacles, such as pedestrians crossing at designated intervals. The simulation integrates a forward safety buffer of three grid cells, equivalent to the vehicles length to ensure safety, which allows the AV to anticipate and react to potential threats before they intersect its path. The system computes planning and safety masks at each timestep, which enable real-time halting and re-routing behavior in response to environmental changes. While the safety radius improves collision avoidance, D* Lite remains a geometric planner without predictive capabilities. Rare collisions can still occur due to dynamic obstacle motion between sensing and actuation. Results of this research demonstrate that the algorithm supports low-latency decision-making and efficient re-planning under uncertainty. The project validates the algorithms practical utility for simulating AV navigation in simplified yet dynamic scenarios.