Q-Learning Based Task Offloading and Resource Allocation Scheme for Internet of Vehicles

In this paper, the task offloading and resource allocation problem for the Internet of Vehicles (IoV) is investigated. In our considered offloading scheme, a Bayesian classifier is first adopted to classify the task according to its different requirements in latency and energy consumption. Based on the classification results, each vehicle user equipment (VUE) then selects the corresponding offloading mode. More specifically, if the VUE has higher requirements for energy consumption, the task will be carried out at other vehicles through the vehicle to vehicle (V2V) offloading mode. Otherwise, it will choose to offload the task through mobile edge computing (MEC) offloading mode. To achieve a trade-off between latency requirement and energy consumption in the task executing process through offloading decision, we formulate the offloading and resource allocation scheme as a mixed-integer non-linear problem. To achieve an approximate solution, a Q-learning based solution is proposed. Simulation results demonstrate that the proposed scheme has better performance in terms of higher system throughput, lower latency, and lower energy consumption compared with the existing schemes.