Robust Computation Offloading and Trajectory Optimization for Multi-UAV-Assisted MEC: A Multiagent DRL Approach

For multiple Unmanned-Aerial-Vehicles (UAVs) assisted Mobile Edge Computing\n(MEC) networks, we study the problem of combined computation and communication\nfor user equipments deployed with multi-type tasks. Specifically, we consider\nthat the MEC network encompasses both communication and computation\nuncertainties, where the partial channel state information and the inaccurate\nestimation of task complexity are only available. We introduce a robust design\naccounting for these uncertainties and minimize the total weighted energy\nconsumption by jointly optimizing UAV trajectory, task partition, as well as\nthe computation and communication resource allocation in the multi-UAV\nscenario. The formulated problem is challenging to solve with the coupled\noptimization variables and the high uncertainties. To overcome this issue, we\nreformulate a multi-agent Markov decision process and propose a multi-agent\nproximal policy optimization with Beta distribution framework to achieve a\nflexible learning policy. Numerical results demonstrate the effectiveness and\nrobustness of the proposed algorithm for the multi-UAV-assisted MEC network,\nwhich outperforms the representative benchmarks of the deep reinforcement\nlearning and heuristic algorithms.\n