Driving under influence: Robust controller migration for MEC-enabled platooning

Connected cars are becoming more common. With the development of multi-access edge computing (MEC) for low-latency applications, it will be possible to manage the cooperative adaptive cruise control (CACC, also known as platooning) of such vehicles from the edge of cellular networks. In this paper, we present a controller that manages platooning from the network edge by adapting to varying network conditions. We incorporate a mechanism in the controller that allows vehicles to switch to automated cruise control when delays exceed safety thresholds, and back to platooning when the delays are sufficiently low to support it. We also formulate the problem of maintaining a low-latency connection in the presence of high mobility through migration, and propose a Q-Learning algorithm to solve this problem. We finally propose an Asynchronous Shared Learning scheme that enables multiple migration agents to cooperate, in order to expedite the convergence of migration policies. Compared to state-of-the-art migration techniques, our scheme exhibits better compliance of vehicle speed and spacing values to preset targets, and ameliorates statistical dispersion.

Automated summary

This paper presents a controller for managing platooning of connected cars from the network edge, adapting to varying network conditions. It also proposes a Q-Learning algorithm to maintain low-latency connection in high mobility scenarios. The introduced scheme exhibits better compliance with vehicle speed and spacing targets compared to existing techniques.