Research On QoE-Oriented Reinforcement Learning Based UAV Anti-Jamming Video Transmission Technique

Unmanned aerial vehicle(UAV)video transmission faces dynamic network topology,limited energy resources and vulnerable to jamming attacks,which degrades the received video quality and even interrupts the communications.Traditional UAV video transmission schemes usually use constant compression and coding parameters and transmit power,which are difficult to meet the various requirements of the video qualityof-experience(QoE)against smart jammers that can change their jamming strategies.Thus,this paper uses game theory and reinforcement learning(RL)to optimize the UAV anti-jamming video transmission policy to improve the video quality and reduce the transmission latency and energy consumption against jamming,which is helpful to improve the effectiveness and reliability of UAV video transmission.First,this paper constructs a UAV anti-jamming video transmission game,derives the Nash equilibrium strategies under different network environment,reveals the impact of the factors such as the task priority and channel gain on the video transmission quality,latency and energy consumption of UAV.For important video capturing tasks such as traffic accident monitoring,when the channel gain between the UAV and the ground control station is larger than a lower bound influenced by the video QoE requirements,the maximum code rate of channel coding and the number of available modulation types,the UAV uses the maximum quantization parameter in video compression coding,the maximum code rate in channel coding,the highest order modulation type for spectrum shifting and the minimum transmit power to transmit the video signals.Meanwhile,the jammer keeps silent.This paper proposes an RL-based UAV anti-jamming video transmission scheme named RL-AJ for dynamic UAV networks,which chooses the video compression coding,channel coding and modulation,and power control strategies without knowing the video service model,channel model and jamming model.This scheme enables the UAV to guarantee the video QoE and reduce the energy consumption when the jammer exists.A safe RL-based approach named SRL-AJ is further proposed for UAVs with strong computing ability,which uses convolutional neural networks to compress the state space to accelerate the UAV learning process,combines with safe RL to reduce the video transmission outage probability.Simulation results based on H.264 compression coding standard and low-density parity-check channel coding show that the proposed RL-AJ scheme can improve the peak signal-to-noise ratio(PSNR)by 31.7%,reduce the latency and energy consumption by 35.6%and 47.3%,respectively,compared with the time division multiple access-based UAV resource allocation scheme.The proposed SRL-AJ scheme can further improve 6.1%PSNR,reduce 22.8%latency and 27.9%energy consumption.