Development of RL -Based Adaptive Congestion Control Algorithm for Wireless Networks

Existing queue-centric algorithms for the congestion management faces problems of reliability, flexibility, and adaptability. Recently, several algorithms focuses on managing queue length and windows for congestion control. Congestion is a major bottleneck in wireless networks. This paper introduces an algorithm for congestion control in wireless networks, leveraging machine-learning techniques. The machine-learning algorithm presented here is an expansion of reinforced learning (RL). The RL algorithm segments the traffic scenario into three conditions: underload, overload, and ideal load. The proposed algorithm predicts future traffic behaviors and controls congestion in wireless communication. Ad-hoc On-Demand Distance Vector (AODV) is the primary routing algorithm, and queue management is a protocol modification. The results of the proposed algorithm were highly efficient, showing significant improvements in terms of packet delivery, throughput, and network loss. The empirical difference in results with the RED and CLM algorithms, such as packet drop ratio increasing by 18% and delay decreasing by 28%.