Abstract
We propose an algorithm to automate fault management in an outdoor cellularnetwork using deep reinforcement learning (RL) against wireless impairments.This algorithm enables the cellular network cluster to self-heal by allowing RLto learn how to improve the downlink signal to interference plus noise ratioand spectral efficiency through exploration and exploitation of various alarmcorrective actions. The main contributions of this paper are to 1) introduce adeep RL-based fault handling algorithm which self-organizing networks canimplement in a polynomial runtime and 2) show that this fault management methodcan improve the radio link performance in a realistic network setup. Simulationresults show that our proposed learns an action sequence to clear alarms andimprove the performance in the cellular cluster better than existingalgorithms, even against the randomness of the network fault occurrences anduser movements.