A Vehicle-Assisted Computation Offloading Algorithm Based on Proximal Policy Optimization in Vehicle Edge Networks.

Area of Science:

• Computer Science
• Artificial Intelligence
• Networking

Background:

• The Internet of Vehicles (IoV) faces challenges with increasing device connectivity and task processing demands.
• Traditional Vehicle Edge Networks (VEN) struggle to manage computational resources efficiently for numerous smart devices.

Purpose of the Study:

• To propose a novel vehicle-assisted computation offloading algorithm (VCOPPO) for User Equipment (UE) tasks in IoV environments.
• To address the challenges of large-scale device access and real-time task processing in VEN.

Main Methods:

• Formulated a non-convex optimization problem considering VEN utility and task processing delay, with constraints on energy and transmission rate.
• Developed the VCOPPO algorithm, integrating dynamic parked vehicle incentives and computational resource allocation.
• Employed a stochastic policy within VCOPPO to determine optimal computation offloading decisions and resource allocation schemes.

Main Results:

• VCOPPO demonstrated superior performance in network reward and task processing delay compared to Dueling DQN, DQN, and Q-learning.
• Achieved significant improvements: 31%, 18%, and 91% increase in network reward.
• Reduced task processing delay by 78%, 63%, and 74% respectively against benchmark algorithms.

Conclusions:

• VCOPPO effectively schedules and allocates computational resources in IoV networks.
• The proposed algorithm offers a robust solution for enhancing VEN performance amidst growing connectivity.
• VCOPPO significantly outperforms existing methods in optimizing network rewards and minimizing task latency.