Learning to Navigate Connected Autonomous Cars for Long-Term Communication Coverage

Autonomous cars employed as mobile base stations could provide communication networks in network-congested areas. In this paper, the authors leverage emerging deep reinforcement learning (DRL) techniques for enabling autonomous cars control and present a novel and highly effective DRL-based control framework called DRL-C3.

Autonomous cars employed as mobile base stations (MBSs) have strong capabilities to provide temporal communication coverage, especially in extreme conditions such as on a crowded street or shopping mall where the existing communication infrastructure cannot support the massive amount of concurrent connection requests that overcapacitates neighboring base stations. In this case, base stations at the network edge may have extra capacity, and thus, autonomous cars within the area need to connect with each other as a multihop network to route the message all the way to the edge base station. As shown in Figure 1, to perform long-term long-distance tasks, autonomous cars with a higher degree of autonomy, or “intelligence,” will be integrated into a “team” (or network) cooperating with each other. Figure 1.

Considered autonomous car control scenario.