Reinforcement Learning-based Scheduling for UAV-assisted Ad-hoc Computing-aware Network

Edge computing can greatly reduce the high data transmission latency associated with cloud computing, offering user equipment (UE) faster and more efficient computing power. Many studies have explored the use of Unmanned Aerial Vehicles (UAVs) as mobile computing nodes, but these often depend on central nodes like base stations or satellites. To overcome this limitation, we propose a UAV-assisted Ad-hoc Computing-aware Network, where each UAV functions as an independent decision-making node. This network can rapidly establish a reliable communication and computing framework, providing computational support to the area. Given that UAVs within the network can form multi-hop groups and offload computation, we introduce the Network-based Deep Deterministic Policy Gradient (N-DDPG) algorithm. This algorithm optimizes the overall computational capacity of the drone network by considering the perceived network topology. Simulations show that N-DDPG outperforms other methods in reducing both delay and energy consumption in UAV-assisted Ad-hoc Computing-aware Networks.