Radio Environment Map (REM) is used to enhance the communication efficiency in spatial spectrum sharing. Using the information from terminals’ reports is one of the methods to generate the REM. The open environment always leads the security problems. High accuracy of terminals’ reports is an essential problem in REM construction. Reinforcement Learning (RL) can estimate and predict the channel condition in a short time, and Multi-armed Bandit (MAB) is a good way to maximize the expected gain by allocating the choices. In this paper, we proposed an algorithm named Terminal Selection Based on MAB (TS-MAB), aim to select a group of reliable terminals to predict the channel condition. For our knowledge, this is the first time using MAB to deal with threatening environment. Simulation results show our method has a good convergence speed with a high accuracy. As long as the percentage of malicious data is less than half of the total data, the algorithm can give a good solution to do the terminal selection and have a high precision prediction of channel conditions.
Terminal Selection Based on Multi-armed Bandit under Threatening Environment for Radio Environment Map Construction
2022-06-01
1707424 byte
Conference paper
Electronic Resource
English
Mobility-Aware Federated Learning: Multi-Armed Bandit Based Selection in Vehicular Network
ArXiv | 2024
|Transportation Research Record | 2019
|