Digital twin-assisted resource allocation framework based on edge collaboration for vehicular edge computing

Abstract

Vehicular Edge Computing (VEC) supports latency-sensitive and computation-intensive vehicular applications by providing caching and computing services in vehicle proximity. This reduces congestion and transmission latency. However, VEC faces implementation challenges due to high vehicle mobility and unpredictable network dynamics. These challenges pose difficulties to network resource allocation. Most existing VEC network resource management solutions consider edge–cloud collaboration and ignore collaborative computing between edge nodes. A reasonable collaboration between Roadside Units (RSUs) or small cells eNodeB can improve VEC network performance. Our proposed framework aims to improve VEC network performance by integrating Digital Twin (DT) technology which creates virtual replicas of network nodes to estimate, predict, and evaluate their real-time conditions. A DT is constructed centrally to maintain and simulate VEC network, thus enabling edge nodes collaboration and real-time resources information availability. We employ channel state information (CSI) for RSUs selection, and vehicles communicate with RSUs through a non-orthogonal multiple access (NOMA) protocol. We aim to maximize the VEC system computation rate and minimize task completion delay by jointly optimizing offloading decisions, subchannel allocation, and RSU association. In view of the resulting optimization problem complexity (NP-hard), we model it as a Markov Decision Process (MDP) and apply Advantage Actor–Critic (A2C) algorithm to solve it. Validated via simulations, our scheme shows superiority to the benchmarks in reducing task completion delay and improving VEC system computation rates.