Method and system for edge server selection for in-vehicle media content distribution with deep reinforcement learning

A remote computing device comprising a controller programmed to:

• receive training data comprising a plurality of vehicle trajectories and a request to download content from a cloud server associated with each vehicle trajectory, each vehicle trajectory traveling through an area covered by a plurality of edge servers;
• determine a location, a transmission range, and a bandwidth capacity of each of the plurality of edge servers based on a map associated with the plurality of edge servers;
• for each vehicle trajectory, determine which of the plurality of edge servers will be within range of the vehicle at a plurality of time steps during the vehicle trajectory based on speed limits at different points along the vehicle trajectory and the transmission range of each of the plurality of edge servers;
• for each vehicle trajectory, determine a graph associated with the vehicle trajectory, the graph indicating a state comprising possible connections between a vehicle and the plurality of edge servers at a plurality of time steps during the vehicle trajectory based on the determination as to which of the plurality of edge servers will be within range of the vehicle at the plurality of time steps, each edge server comprising a node of the graph, and each possible connection between the vehicle and an edge server comprising an edge of the graph;
• input the state associated with each vehicle trajectory into a convolutional neural network configured to output an action, the action comprising a partition of the content associated with the vehicle trajectory associated with the state across the plurality of edge servers;
• for each vehicle trajectory, determine a latency of transmission of the content from the cloud server to the vehicle based on the action determined by the convolutional neural network;
• input each state and each associated action determined by the convolutional neural network into a graph neural network configured to train the convolutional neural network to minimize the latency of transmission of the content from the cloud server to the vehicle via the plurality of edge servers using reinforcement learning with the determined latency used as a reward;
• and
• train the convolutional neural network based on an output of the graph neural network.