Next generation radio access networks (NG-RAN) are required to provide seamless connectivity for a myriad range of use cases such as Ultra Reliable Low Latency Communication (URLLC), enhanced Mobile Broadband (eMBB) and massive Machine Type Communications (mMTC) using gigahertz or terahertz wave spectrum. To address these stringent requirements, each layer of protocol suite has to play a significant role in providing the network coverage, speed, reliability and low latency for real time applications like voice over new radio (VoNR). Reinforcement Learning (RL) algorithms provide a highly agile way to fine-tune various network control, management settings and help in decision making for various telecommunication protocols of NG-RAN. This work demonstrates applicability of RL algorithms in reducing call mute instances by optimizing Robust header compression (ROHC) protocol. ROHC is deployed at packet data convergence protocol (PDCP) layer of NG-RAN to save radio resources by compressing recurring header fields in a given VoNR call. In case of networks experiencing high bit error rates and wherein ROHC is enabled, VoNR users often complain about degradation of service quality in the form of call muting. In this work, we develop a Reinforcement Learning based network decision mechanism to control and configure ROHC during suboptimal radio conditions to reduce packet discards and prevent call muting. The simulation results of our work show around 12 percent reduction in call mute instances.

https://ieeexplore.ieee.org/document/10008776