Satellite-based ITS Data Offloading & Computation in 6G Networks: A Cooperative Multi-Agent Proximal Policy Optimization DRL with Attention Approach

With the rapid growth of intelligent transportation systems (ITS), there is a growing need to support real-time network applications. However, terrestrial networks are insufficient to support diverse applications for remote airplanes ships, and trains. Meanwhile, satellite networks can be a great supplement to terrestrial networks regarding coverage, flexibility, and availability. Thus, we investigate a novel ITS data offloading and computations services based on satellite networks, in which low-Earth orbit (LEO) and cube satellites are regarded as independent mobile edge computing (MEC) servers, responsible for scheduling the processing of ITS data generated by ITS nodes. We formulate a joint delay and rental price minimization problem for different satellite servers while optimizing offloading task selection, computing, and bandwidth resource allocation, which is mixed-integer non-linear programming (MINLP) and NP-hard. To deal with the problem's complexity, we divide the problem into two stages. Firstly, we proposed a cooperative multi-agent proximal policy optimization (Co-MAPPO) deep reinforcement learning (DRL) with an attention approach for determining intelligent offloading decisions with quick convergence. Secondly, we break down the remaining subproblem into independent subproblems and find their optimal closed-form solutions. Extensive simulations are utilized to validate the proposed approach's effectiveness in comparison to baselines by 8.92% and 3.14% respectively.