Autonomous Driving Planning Method for Complex Uncertain Environments
Date:
A report on an efficient GPU-based implementation method for the fusion of localization and perception uncertainties was presented. By meticulously decomposing the propagation process and alternating between GPU and CPU usage, a computational efficiency improvement of over 6 times was achieved. In combination with planning methods, a more efficient planning process was implemented. Compared to uncertainty-aware approaches such as CC-MPC, the effectiveness of the method proposed in this report was demonstrated.