MPSO: An Optimization Algorithm for Task Offloading in Cloud-Edge Aggregated Computing Scenarios for Autonomous Driving

With the development of cloud computing and edge computing technologies, these technologies have come to play a crucial role in the field of autonomous driving. The autonomous driving sector faces unresolved issues, with one key problem being the handling of latency-sensitive applications within vehicles. Cloud computing and edge computing provide a solution by segmenting unresolved computing tasks and offloading them to different computing nodes, effectively addressing the challenges of high concurrency through distributed computing. While the academic literature addresses computation offloading issues, it often focuses on static scenarios and does not fully leverage the advantages of cloud computing and edge computing. To address these challenges, a multivariate particle swarm optimization (MPSO) algorithm tailored for the cloud-edge aggregated computing environment in the autonomous driving domain is proposed. The algorithm, grounded in real-world scenarios, considers factors that may impact computation latency, abstracts them into quantifiable attributes, and determines the priority of each task. Tasks are then assigned to optimal computing nodes to achieve a balance between computation time and waiting time, resulting in the shortest total average weighted computation latency time for all tasks. To validate the effectiveness of the algorithm, experiments were conducted using the self-designed CETO-Sim simulation platform. The algorithm's results were compared with those of simulated annealing, traditional particle swarm optimization, purely local computation, and purely cloud-based computation. Additionally, comparisons with traditional algorithms were considered in terms of iteration count and result stability. The results indicate that the MPSO algorithm not only achieves optimal computation offloading strategies within specified time constraints when addressing computation offloading issues in the autonomous driving domain but also exhibits high stability. Furthermore, the algorithm determines the processing location for each computing task, demonstrating significant practical value.