Mining unmanned electric locomotive speed control method based on improved epsilon-greedy strategy and reinforcement learning

The invention discloses a mining unmanned electric locomotive speed control method based on an improved epsilon-greedy strategy and reinforcement learning, and the method comprises the steps: enabling a mining unmanned electric locomotive to serve as an intelligent agent, and setting the motion of the intelligent agent as an integer within the range of the optimal driving/braking torque applied to an axle by the electric locomotive; the method comprises the following steps: respectively designing actions, states and reward functions of an intelligent agent, and completing the construction of a main body algorithm structure for controlling the speed of the mining unmanned electric locomotive by taking real-time state information of the electric locomotive as input and taking driving/braking torque applied to an axle and a reward value obtained by the intelligent agent as output by utilizing a Q-Learning algorithm. The mining electric locomotive is ensured to autonomously complete behaviors such as safe car fol