MUMBAI, India, Jan. 2 -- Intellectual Property India has published a patent application (202541123830 A) filed by Malla Reddy (MR) Deemed to be University; Malla Reddy University; Malla Reddy Engineering College For Women; Malla Reddy College Of Engineering And Technology; and Malla Reddy Vishwavidyapeeth, Medchal-Malkajgiri, Telangana, on Dec. 9, 2025, for 'dynamic bandwidth allocation controller using deep reinforcement learning.'
Inventor(s) include Mr. Pothuganti Srikanth; Dr. M. Naresh; Mr. Badana Upendra; I Uma Mahesh; and Adlakadi Anand.
The application for the patent was published on Jan. 2, under issue no. 01/2026.
According to the abstract released by the Intellectual Property India: "The current invention reveals a smart Dynamic Bandwidth Allocation (DBA) Controller which uses Deep Reinforcement Learning (DRL) to automatically control network resources in extremely dynamical telecommunication conditions. Conventional methods of bandwidth management, including traditional provisioning or basic heuristic scaling, are not responding fast enough to burst traffic patterns typical of modern applications such as 4K streaming, online gaming or the Internet of Things (IoT). Such traditional approaches normally lead to either wastage of the expensive network spectrum or gross overload that undermines the user experience. The invention addresses these shortcomings through the deployment of a self-learning agent that is able to predict traffic demand and automatically adjust bandwidth allocation during real-time even without the intervention of a human being. The essential key of the invention is its closed-loop control design, in which an agent based on neural networks monitors the state of the network, aiding it in measures like packet loss rate, latency, jitter and buffer occupancy. As opposed to the supervised learning models which demand immense volumes of labeled historical data, the given DRL agent learns by working with the network environment. It makes decisions, including widening channel, or giving priority to certain flows, and gets a reward signal that is computed by the resulting Quality of Experience (QoE). The system will learn to predict congestion before it happens and in the long run, the system will maximize its policy to maximize this reward. Moreover, the system is developed to work under Software Defined Networking (SDN) and Network Function Virtualization (NFV) systems. The DRA agent is the brain of SDN controller, which provides updates to flow tables to network switches to attractively allocate resources. This enables network slicing control on a granular basis and the critical traffic has guaranteed bandwidth as the best-effort traffic is efficiently handled. The outcome is a strong, responsive network that is resilient enough to keep the throughput and low latency levels high even in the times of unforeseen surges in traffic. Lastly, the invention will solve the problem of fairness and resource contention in multi-tenant networks. The controller can make sure that although the aggregate network throughput is maximized, no individual user or service is starved to connectivity by including a fairness index in the reward function. This efficiency of the entire system and the fairness of individual users is what makes the solution especially appropriate within 5G networks and enterprise settings where different applications needs have to coexist on common infrastructure."
Disclaimer: Curated by HT Syndication.
