Deep Reinforcement Learning-Based Computation Offloading in UAV Swarm-Enabled Edge Computing for Surveillance Applications

The rapid development of the Internet Things and wireless communication has resulted in emergence many latency-constrained computation-intensive applications such as surveillance, virtual reality, disaster monitoring. To satisfy computational demand reduce prolonged transmission delay to cloud, mobile edge computing (MEC) evolved a potential candidate that can improve task completion efficiency reliable fashion. Owing its high nature ease use, promising candidates, unmanned aerial vehicles (UAVs) be incorporated with MEC support latency-critical applications. However, determining ideal offloading decision for UAV on basis characteristics still remains crucial challenge. In this paper, we investigate surveillance application scenario hierarchical swarm includes an UAV-enabled team UAVs surveilling area monitored. determine optimal policy, propose deep reinforcement learning based computation (DRLCO) scheme using double Q-learning, which minimizes weighted sum cost by jointly considering execution energy consumption. A performance study shows proposed DRLCO technique significantly outperforms conventional schemes terms cost, consumption, delay. better convergence effectiveness method over are also demonstrated.