Reinforcement-Learning-Based Deadline Constrained Task Offloading Schema for Energy Saving in Vehicular Edge Computing System

In the age of the ever-growing number of tasks generated from the Internet of Things (IoT) devices, one of the most crucial problems with enhancing the Quality of Service in multi-access computing (MEC) is to have a low overdue-task rate (tasks with processing time greater than their deadlines) while minimizing the energy consumed. To properly formulate the task offloading in a vehicular network, we consider both the number of overdue tasks and the total energy consumed by the edge system. We focus not only on maximizing the users' experience by minimizing the percentage of overdue tasks rate but also on saving energy for the service provider. Thus, our Deadline-constrained and Energy-aware problem requires finding a task offloading strategy to reduce the total power consumption of the edge system while still minimizing the number of overdue tasks. Findings from a 2D-street real-data bus traces are also provided for analysis. Furthermore, we develop a schema based on rein-forcement learning techniques named Deadline Constrained and Energy-Aware Task Offloading (DCEAO) to solve the problem. It proves to reduce the base station's power consumption by 38% to 51% while maintaining competitive overdue-task rate to other benchmarks that only focus on minimizing overdue-task rate.