Multi-Agent-Deep-Reinforcement-Learning-Enabled Offloading Scheme for Energy Minimization in Vehicle-to-Everything Communication Systems

Abstract

Offloading computation-intensive tasks to mobile edge computing (MEC) servers, such as road-side units (RSUs) and a base station (BS), can enhance the computation capacities of the vehicle-to-everything (V2X) communication system. In this work, we study an MEC-assisted multi-vehicle V2X communication system in which multi-antenna RSUs with liner receivers and a multi-antenna BS with a zero-forcing (ZF) receiver work as MEC servers jointly to offload the tasks of the vehicles. To control the energy consumption and ensure the delay requirement of the V2X communication system, an energy consumption minimization problem under a delay constraint is formulated. The multi-agent deep reinforcement learning (MADRL) algorithm is proposed to solve the non-convex energy optimization problem, which can train vehicles to select the beneficial server association, transmit power and offloading ratio intelligently according to the reward function related to the delay and energy consumption. The improved K-nearest neighbors (KNN) algorithm is proposed to assign vehicles to the specific RSU, which can reduce the action space dimensions and the complexity of the MADRL algorithm. Numerical simulation results show that the proposed scheme can decrease energy consumption while satisfying the delay constraint. When the RSUs adopt the indirect transmission mode and are equipped with matched-filter (MF) receivers, the proposed joint optimization scheme can decrease the energy consumption by 56.90% and 65.52% compared to the maximum transmit power and full offloading schemes, respectively. When the RSUs are equipped with ZF receivers, the proposed scheme can decrease the energy consumption by 36.8% compared to the MF receivers.

Cloud–Fog Collaborative Computing Based Task Offloading Strategy in Internet of Vehicles

“…In general, we set both α and β to 0.5 to obtain the task offloading strategy that balances system delay and energy consumption. Equation (25) indicates that all tasks need to be offloaded, and each subtask can only be offloaded at one location. Equation (26) represents the constraints on the weight coefficients of system delay and energy consumption.…”

“…When the chromosome is encoded as {2, 0, 3, 1, 0, 4, 2, 3}, the tasks M 2 and M 5 are offloaded to the source vehicle, the task M 6 is offloaded to the cloud server, and the remaining tasks are offloaded to the same serial number of fog servers. (16) if Recursively run the algorithm; (25) end if (26) end if (27)…”