Energy Minimization in UAV-Aided Networks: Actor-Critic Learning for Constrained Scheduling Optimization

In unmanned aerial vehicle (UAV) applications, the UAV's limited energy supply and storage have triggered development of intelligent energy-conserving scheduling solutions. this paper, we investigate minimization for UAV-aided communication networks by jointly optimizing data-transmission UAV hovering time. The formulated problem is combinatorial non-convex with bilinear constraints. To tackle problem, firstly, provide an optimal algorithm (OPT) a golden section search heuristic (GSS-HEU). Both solutions are served as offline performance benchmarks which might not be suitable online operations. Towards end, from deep reinforcement learning (DRL) perspective, propose actor-critic-based stochastic (AC-DSOS) develop set approaches to confine action space. Compared conventional RL/DRL, novelty AC-DSOS lies in handling two major issues, i.e., exponentially-increased space infeasible actions. Numerical results show that able feasible solutions, save around 25-30% compared AC-DRL algorithms. developed GSS-HEU, consumes 10% higher but reduces computational time second-level millisecond-level.