By integrating the Deep Q-Network, a model-free reinforcement learning system, with the dynamic programming optimization method, the efficiency of mining operations can be enhanced, thereby reducing waiting times and energy consumption.
Dispatched 10 homogeneous autonomous vehicles in a closed mining area by using the intelligent system in real time. Enhanced the full-time autonomous operation capability to reach more than 120% of the manual by EV trucks.
A multi-objective speed trajectory optimization method using dynamic programming was proposed considering travel time, energy consumption, and battery lifetime. And it is verified on the fuel and electric vehicle models.
Based on ROS2 in the Ubuntu22.04 system, built nodes for perception, localization, path planning, and MPC tracking.