Joint Optimization Time-Slotted Computing Offloading and V2X Resource Allocation by Reinforcement Learning
针对动态车辆编队中计算卸载与资源分配难题，提出一种基于规则的强化学习框架，联合优化时隙任务卸载与V2X资源分配，在满足能耗约束下降低计算复杂度，适用于车联网场景。
This article addresses the challenges of computation offloading and resource allocation in dynamic vehicular platoons, where high-speed mobility and sparse roadside unit (RSU) deployment lead to intermittent connectivity and complex task scheduling delays. This article proposes a rule-based reinforcement learning (RL) framework that jointly optimizes time-slotted task offloading and vehicle-to-everything (V2X) resource allocation while strictly adhering to energy consumption constraints, which reduces computational complexity while ensuring optimization accuracy. The framework integrates domain-specific rules—such as leader vehicle platoon leader (PL) capacity limits, RSU computation thresholds, and energy budgets—into the RL decision-making process to ensure feasible actions across four offloading scenarios: local computation, direct RSU offloading, relay-based RSU offloading, and PL offloading. The problem is formulated as a mixed-integer nonlinear programming (MINLP) model and decomposed into vehicle-level mode selection and computing unit-level resource allocation. Extensive simulations demonstrate the algorithm’s robustness in dynamic environments.