A Power Allocation Scheme for MIMO-NOMA and D2D Vehicular Edge Computing Based on Decentralized DRL

Area of Science:

• Vehicular Edge Computing (VEC)
• Wireless Communications
• Artificial Intelligence

Background:

• Vehicular edge computing (VEC) involves task processing locally or on mobile edge computing (MEC) servers.
• Task offloading decisions depend on vehicle-to-infrastructure (V2I) and vehicle-to-vehicle (V2V) communication status.
• Complex VEC environments feature uncertain MIMO-NOMA V2I channel conditions and random task arrivals.

Purpose of the Study:

• To address the complexity of task offloading in VEC systems with uncertain communication environments.
• To propose an effective power allocation scheme for optimizing VEC performance.
• To enhance decision-making in dynamic V2I and V2V communication scenarios.

Main Methods:

• Utilized device-to-device (D2D)-based V2V communication.
• Employed multiple-input multiple-output and nonorthogonal multiple access (MIMO-NOMA)-based V2I communication.
• Developed a power allocation scheme using decentralized deep reinforcement learning (DRL) and the deep deterministic policy gradient (DDPG) algorithm for continuous action spaces.

Main Results:

• The DRL-based power allocation scheme effectively manages task offloading in VEC.
• The proposed DDPG algorithm achieved optimal policy acquisition in continuous action spaces.
• Experimental results showed superior performance compared to existing greedy strategies.

Conclusions:

• The DRL and DDPG approach offers a robust solution for power allocation in VEC.
• The proposed scheme significantly reduces power consumption and improves rewards in complex vehicular networks.
• This research advances intelligent resource management in vehicular edge computing.