Joint Computation Offloading and Power Allocation Strategy in NOMA-Based Dynamic MEC Network Assisted by RIS

Mobile edge computing (MEC) holds great promise as an effective solution that empowers resource-constrained intelligent applications to transfer computation-intensive operations to neighboring edge servers. To realize its potential, reconfigurable intelligent surface (RIS) provides high spectral and energy efficiency, can effectively reduce the computation costs. This paper discusses a joint computation offloading and power allocation strategy for non-orthogonal multiple access (NOMA) based MEC scenario assisted by RIS. First, we formulate a cost minimization problem which considers task buffering delay, power consumption, and task queue length constraints. Then, we use Lyapunov optimization method to transform the task queue length constraint into a queue stability problem. Finally, a Markov decision process (MDP) model and a double deep Q-network (DDQN) based computation offloading and power allocation (DCOPA) algorithm are proposed to obtain optimal computation offloading strategy and achieve the system's cost objective. Through simulation results, it is demonstrated that the proposed method can effectively reduce the computation offloading delay and has a lower computation cost than other methods. The algorithm is effective in both convergence and long-term performance.