Autonomous Automotives on the Edge

With autonomous automotives routinely leveraging computationally intensive tasks to enable robust navigation, a number of design challenges have emerged. Whilst processing tasks is traditionally carried out on the vehicles, the emergence of Multi-Access Edge Computing (MEC) has paved the way to transfer such tasks to be executed not just locally on the vehicles but also offloads such tasks to powerful MEC servers co-located with cellular base-stations. In a vehicular MEC environment, a computation offloading policy determines which tasks are to be executed locally on the vehicles and which tasks are to be transferred to MEC servers for further processing. In recent years, a number of offloading policies in have been delineated considering several optimization objectives. However, the uncertainty associated with observability metrics due to the high stochasticity of the network environment has been less explored. In this paper, we highlight the impact of this uncertainty on timeliness guarantees for safe autonomy and propose a quantitative model checking approach towards offloading tasks to MEC servers. We believe our article can motivate new research directions towards offloading in the vehicular MEC context.