Unmanned vehicle planning method and device based on deep reinforcement learning

Abstract

The invention provides an unmanned vehicle planning method and device based on deep reinforcement learning. The method comprises the following steps: obtaining to-be-tested network performance information and to-be-tested observation state information; inputting the to-be-tested network performance information and the to-be-tested observation state information into an unmanned vehicle scheduling prediction model to obtain a prediction action of the unmanned carrying vehicle output by the unmanned vehicle scheduling prediction model; wherein the unmanned vehicle scheduling prediction model is obtained by selecting a second preset number of training samples from an experience playback pool for optimization training after the number of the training samples in the experience playback pool reaches a first preset number, the training sample in the experience playback pool is obtained according to a historical observation state of a corresponding time step number, historical network performance information, a historical action selected based on the historical observation state, and a next historical observation state and a historical reward obtained based on the historical network performance information after the selected historical action is executed. The state change can be captured in time, and the accuracy of motion prediction is improved.

Images