Edge computing in mobile networks enables energy- and time-efficient processing of advanced computational tasks beyond end-user devices. With the emergence of new applications, such as those required for autonomous driving, these tasks are often complex, consisting of interdependent sub-tasks. Additionally, many of these applications rely on neural network processing, such as for object or obstacle detection, which places significant demands on the computing power of end-user devices, like autonomous vehicles, if processed solely on-board.
A viable approach to reduce the computing burden on vehicles is to offload selected computational tasks to distributed edge resources in the form of Multi-access/Mobile Edge Computing (MEC), complemented by in-vehicle pre-processing. Our objective is to design and experimentally verify effective solutions for the joint allocation of computing and mobile communication resources to support the distributed processing of complex applications, including those based on neural networks.
To ensure the availability and reliability of resources when needed, we extend this approach by incorporating mid-term pre-allocation (booking) of communication and computing resources, with time horizons ranging from seconds to hours. The outcomes of this project will significantly reduce the cost of in-vehicle computing hardware while also shortening the application processing time.
