Abstract
Ad Hoc Networks with directional antennas are usually applied to disaster relief, battlefields and other special environments, where multidimensional resource allocation is an important issue for formation control applications. In this paper, the multidimensional resource allocation problem included time-frequency resource blocks and transmission power in directional ad hoc networks is modelled as non convex programming problem. Additionally, the interference caused by the time-frequency reuse capability of using directional antennas is considered. Then, we propose a distributed resource allocation framework for multiple frequencies time division multiple access (MF - TDMA) directional wireless ad hoc networks based on multi-agent deep reinforcement learning, to solve the nonconvex programming problem. Simulation results indicate that the proposed allocation framework achieves time- frequency block resource reuse and significantly enhances network capacity compared to traditional algorithms.