Radio Interference Source Localization with UAV-Based Q-Learning

Expanding aerial monitoring methods is of great significance to radio management, that contributes to regulatory spectrum monitoring and interference hunting especially aviation interference. In this paper, a Q-Learning algorithm based on unmanned aerial vehicle (UAV) is proposed to locate radio interference sources under the Rayleigh channel. This method locates radio interference sources through times of interaction with the environment and self-learning, which can operate without any prior knowledge about the environment or the model for received signal strength (RSS) observations. Firstly, a signal model considering Rayleigh fading is established. Secondly, the problems of state division, action division, immediate reward and exploration strategy are discussed. Finally, Matlab is used to verify the feasibility of using Q-Learning algorithm to autonomously locate radio interference sources with a UAV. The impacts of the number of actions, two kinds of action selection method (ε-greedy and Softmax) on the path were compared by simulation. Comparative experimental results prove that the Q-Learning algorithm can enable the UAV to locate the radio interference source autonomously with a short path. And ε-greedy outperform softmax. When the number of actions is 8, the exploration factor ε is 0.1, and the learning rate is 0.9, the UAV locate the radio interference source with the shortest path.