Dynamic Task Software Caching-Assisted Computation Offloading for Multi-Access Edge Computing

In multi-access edge computing (MEC), most existing task software caching works focus on statically data at the network edge, which may hardly preserve high reusability due to time-varying user requests in practice. To this end, work considers dynamic MEC server assist users’ execution. Specifically, we formulate a joint update (TSCU) and computation offloading (COMO) problem minimize energy consumption while guaranteeing delay constraints, where limited cache size capability of server, as well demand users are investigated. This is proved be non-deterministic polynomial-time hard, so transform it into two sub-problems according their temporal correlations, i.e., real-time COMO Markov decision process-based TSCU problem. We first model multi-user game propose decentralized algorithm address its Nash equilibrium solution. then double deep Q-network (DDQN)-based method solve policy. reduce complexity convergence time, provide new design for neural (DNN) DDQN, named state coding action aggregation (SCAA). SCAA-DNN, introduce dropout mechanism input layer code activity states. Additionally, output layer, devise two-layer architecture dynamically aggregate actions, able huge state-action space Simulation results show that proposed solution outperforms schemes, saving over 12% energy, converges with fewer training episodes.