MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202641100801 A) filed by Gaurav Kumar; Girisha Gs; and Shamanth N on August 20, 2026, for Goal-Based Td3 With Hindsight Experience Replay For Scalable And Efficient Routing In Sdn.
Inventors include Gaurav Kumar; Girisha Gs; and Shamanth N.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: The rapid development of traffic in Software-Defined Networking (SDN) environments demands adaptive and effective routing strategies to ensure high Quality of Service (QoS). Conventional algorithms suffer from poor adaptability, local optimal and inefficiency in dynamic and sparse reward scenarios and Deep Reinforcement Learning (DRL) approaches suffer from slow convergence and unstable learning behaviour under highly dynamic conditions. To address these drawbacks, this invention proposed a Goal-Based Twin Delayed Deep Deterministic Policy Gradient with Hindsight Experience Replay (GB-TB3 + HER) model to enable scalable and robust traffic engineering in SDN. The Goal-based TD3 component stabilises policy learning through overcoming Q- value overestimation using twin critics and delayed policy updates, when HER converts sparse reward into dense feedback through relabeling failed trajectories with alternative goals. The reward function ensures latency minimization, congestion avoidance and increasing throughput. The proposed GB-TD3 + HER model obtains less packet loss of 1.12, latency of 117.3 ms, throughput of 98.47% and execution time of 15.62 ms when compared to existing models, thereby ensuring scalable, energy-efficient and reliable routing decisions for dynamic SDN environments.
Disclaimer: Curated by HT Syndication.
