Intelligent anti-jamming decision algorithm based on proximal policy optimization

The existing intelligent anti-jamming methods based on deep reinforcement learning are applied to space-ground TT&C and communication links, in which the deep neural network used for decision-making has a complex structure, and the resources of satellites and other vehicles are limited, making it difficult to independently complete the timely training of complex neural network under the constraints of limited complexity, and the decision-making of anti-jamming cannot converge. Aiming at the above problems, an intelligent anti-jamming decision algorithm based on proximal policy optimization was proposed, which deployed the decision-making neural network and the training neural network in the vehicles and the ground station, respectively. The ground station conducted the optimal offline training based on the empirical information feedback from the vehicles, and assisted the decision-making neural network in parameter updating, thereby achieving the effective selection of anti-jamming strategies while satisfying the resource constraints of the vehicles. The simulation results demonstrate that the convergence speed of the proposed algorithm is increased by 37%, and the system capacity after convergence is increased by 25%, compared with the decision algorithms of policy gradient and deep Q-learning. © 2024 Editorial Board of Journal on Communications. All rights reserved.