Research Article
Collaborative Interference Source Search and Localization Based on Reinforcement Learning and Two-Stage Clustering
@INPROCEEDINGS{10.1007/978-3-030-69069-4_36, author={Guangyu Wu and Yang Huang and Simeng Feng}, title={Collaborative Interference Source Search and Localization Based on Reinforcement Learning and Two-Stage Clustering}, proceedings={Wireless and Satellite Systems. 11th EAI International Conference, WiSATS 2020, Nanjing, China, September 17-18, 2020, Proceedings, Part I}, proceedings_a={WISATS}, year={2021}, month={2}, keywords={Reinforcement learning Clustering Unmanned aerial vehicle Localization Wireless communications}, doi={10.1007/978-3-030-69069-4_36} }- Guangyu Wu
Yang Huang
Simeng Feng
Year: 2021
Collaborative Interference Source Search and Localization Based on Reinforcement Learning and Two-Stage Clustering
WISATS
Springer
DOI: 10.1007/978-3-030-69069-4_36
Abstract
Exploiting unmanned aerial vehicles (UAVs) to locate the position of interferences has attracted intensive research interests, due to UAVs’ flexibility and the feature of suffering less multi-path interference. However, in order to find the position of an interference source, off-the-shelf Q-learning-based schemes require the UAV to keep searching until it arrives at the target. This obviously degrades time efficiency of localization. To balance the accuracy and the efficiency of searching and localization, this paper proposes a collaborative search and localization approach, where search and remote localization are iteratively performed with a swarm of UAVs. For searching, a low-complexity reinforcement learning algorithm is proposed to decide the direction of flight (in every time interval) for each UAV. In the following remote localization phase, a two-stage clustering algorithm is proposed to estimate the position of the interference source, by processing intersections of the extensions of UAVs’ trajectories. Numerical results reveal that in the proposed collaborative search and localization scheme, the proposed reinforcement-learning-based searching can benefit the collaborative localization, in terms of the accuracy of localization. Moreover, compared to the Q-learning-based approach, the proposed approach enables remote localization and can well balance accuracy, the robustness and time efficiency of localization.
