VEC Collaborative Task Offloading and Resource Allocation Based on Deep Reinforcement Learning Under Parking Assistance

With the emergence of autonomous vehicles, meeting the vehicle’s computing needs for computationally intensive and latency-sensitive tasks has become a challenge. Cellular Vehicle-to-Everything (C-V2X), an essential Internet of Vehicles technology, is expected to be enhanced and strengthened in the 6G era to improve road traffic safety and realize intelligent transportation. However, when too many computationally intensive and latency-sensitive tasks of autonomous vehicles are offloaded to the MEC server, the server can become overloaded, unable to meet the offloading demands of numerous vehicles. Inspired by the idle parked cars on both sides of the road, this paper proposes utilizing idle vehicles to assist VEC servers in offloading computing tasks, thereby increasing resource capacity and expanding communication range. Therefore, this paper firstly utilizes roadside parked vehicles with idle computing resources as a task offloading platform to develop a mobile edge computing task offloading strategy based on roadside parked vehicle collaboration. Secondly, a more flexible offloading solution is proposed based on comprehensive consideration of offloading decisions and resource allocation in a multi-user and multi-server parked vehicle-assisted MEC environment. Subsequently, to ensure service quality for end users, we consider weighted total delay cost and energy consumption as optimization objectives. The optimization problem is further formulated as a Markov decision process, and a joint computing offloading and resource allocation optimization algorithm based on deep reinforcement learning is proposed to minimize the total delay and energy consumption of vehicle users. Finally, experimental results validate the algorithm. Compared with other benchmark solutions, the proposed scheme improves system performance by 78%, 77.3%, 72.7%, and 71.1%, respectively. The proposed scheme reduces the total system cost during task offloading and enhances system performance.