Semantic-Aware Path Planning by Using Dynamic Weighted‎Dijkstra for Autonomous Driving

In the current urban road environment, it is difficult to balance path planning, semantic perception, and energy consumption management. The existing algorithms respond slowly to sudden congestion, traffic lights, and pedestrian activities, and ‎lack energy constraint considerations. This paper builds a closed-loop autonomous driving system covering path planning, ‎positioning fusion, and energy perception control on the MATLAB simulation platform. The path planning module uses ‎built-in traffic signals, speed limit areas, and pedestrian activity areas as dynamic weights, and generates smooth feasible ‎trajectories through the improved dynamic weighted Dijkstra algorithm combined with B-spline interpolation to ensure ‎efficient response in various scenarios. The positioning fusion module adopts extended Kalman filtering technology to ‎integrate GPS, wheel speed odometer, front-mounted camera, and LiDAR data, and simulates signal loss and Gaussian ‎noise interference to ensure high-precision positioning under harsh conditions. The energy perception control module ‎just the acceleration and steering strategies in real time based on simplified vehicle dynamics and residual energy budget, ‎and supports dry/wet road braking simulation, to achieve endurance optimization under the premise of ensuring ‎safety. The study shows that the proposed method can simultaneously improve safety and energy efficiency in dynamic ‎traffic and multi-source noise interference environments‎.