Massive Random Access Control for VLC: A Reinforcement Learning Driven Approach

Visible light communication (VLC) has been widely applied to provide dense network access. Usually random access requests are sparse, and the users can be efficiently detected using compressed sensing (CS) based methods. However, in case of bursting network traffic, massive access requests can significantly degrade the performance of user detection. To this end, we propose an intelligent massive random access control scheme, i.e., Sparse Adaptive Random Access (SARA), based on reinforcement learning (RL). Through iterative interactions with the complex and time-varying environment, the proposed scheme of SARA can smartly provide appropriate flow control levels for users with different priorities. Thus, it can not only respond to high-priority users in a timely manner, but also avoid the low detection accuracy caused by massive access requests. The simulation results demonstrate that the proposed scheme outperforms the benchmark schemes in case of high concurrent traffic.