Distributed reinforcement learning based framework for energy-efficient UAV relay against jamming

Unmanned aerial vehicle (UAV) network is vulnerable to jamming attacks, which may cause severe damage like communication outages. Due to the energy constraint, the source UAV cannot blindly enlarge the transmit power, along with the complex network topology with high mobility, which makes the destination UAV unable to evade the jammer by flying at will. To maintain communication with a limited battery capacity in the UAV networks in the presence of a greedy jammer, in this paper, we propose a distributed reinforcement learning (RL) based energy-efficient framework for the UAV networks with constrained energy under jamming attacks to improve the communication quality while minimizing the total energy consumption of the network. This framework enables each relay UAV to independently select its transmit power based on historical state-related information without knowing the moving trajectory of other UAVs as well as the jammer. The location and battery level of each UAV need not be shared with other UAVs. We also propose a deep RL based anti-jamming relay approach for UAVs with portable computation equipment like Raspberry Pi to achieve higher and faster performance. We study the Nash equilibrium (NE) and the performance bounds based on the formulated power control game. Simulation results show that the proposed schemes can reduce the bit error rate (BER) and reduce energy consumption of the UAV network compared with the benchmark method.