Intelligent and Collaborative Computing Offloading and Resource Management in Satellite-Cloud-MEC Integrated IoVs

In this paper, we investigate collaborative computing offloading and multi-dimensional resource slicing/allocation within low Earth orbit (LEO) constellation-assisted Internet of Vehicles (IoV) networks. To support the increasing resource demand for delay- and computation-intensive tasks, we develop an IoV system that leverages both terrestrial multi-access edge computing (MEC) servers and core network cloud servers as service providers, enabling collaborative access via terrestrial and non-terrestrial networks. Specifically, we formulate an optimization problem to achieve efficient collaborative computing offloading with guaranteed quality of service in the considered IoV system. Given the challenges of heterogeneous timescales, mixed-integer optimization variables, and dynamic task arrivals, we decompose the formulated problem into two subproblems and design a two-timescale hierarchical Markov decision process (HMDP) framework for subproblem transformation. We then propose two hierarchical hybrid actor-critic (HHAC) algorithms: hierarchical hybrid deep deterministic policy gradient (HHDDPG) and hierarchical hybrid proximal policy optimization (HHPPO), to solve the HMDP-transformed subproblems efficiently. Extensive simulation results demonstrate that our proposed HHAC algorithms achieve high average satisfaction and low average delay compared to four benchmark methods.