Load-balanced Task Allocation for Covid-19 Close Contact Detection in Heterogeneous MEC Networks

ABSTRACT

In this paper, we investigate the close contact detection for COVID-19 patients based on the heterogeneous mobile edge computing (MEC) framework. Collecting the spatial-temporal data of a large number of mobile users, the base stations equipped with MEC servers organize these data via the R-tree structure. The cloud center (CC) aggregates the spatial-temporal data from all MEC servers. Considering the mobility of users as well as various positions of MEC servers, the CC then partitions and assigns the close contact detection tasks to different servers for faster processing. Aiming to minimize the system latency, we propose a Deep Deterministic Policy Gradient-based task and resource allocation scheme, where the computing loads are balanced among different servers. Simulation results show that a minimum system latency is reached while maintaining the load balance among all servers. Up to 37% detection accuracy enhancement is achieved compared with an existing task allocation scheme without load balance.