Unmanned aerial vehicle assisted Internet of Vehicles resource allocation method based on deep reinforcement learning

Abstract

The invention provides an unmanned aerial vehicle assisted Internet of Vehicles resource allocation method based on deep reinforcement learning, and the method comprises the steps: achieving the collaborative optimization of vehicle-unmanned aerial vehicle association, spectrum resource allocation and unmanned aerial vehicle motion trail through building a multi-dimensional parameter joint optimization framework, and constructing a system energy efficiency maximization model. According to the method, the distance clustering algorithm and deep reinforcement learning are creatively fused, and the convergence and stability of the training process are remarkably improved through feature space dimensionality reduction. Furthermore, a hybrid execution mechanism combining off-line pre-training and on-line adaptive decision is adopted, and the problem of real-time resource optimization in a dynamic environment is effectively solved. The method is superior to a traditional scheme in key performance indexes such as system energy efficiency, decision real-time performance and stability, and an efficient new resource management normal form is provided for an unmanned aerial vehicle auxiliary vehicle networking system.

Images