MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202611024014 A) filed by Yushu Excellence Technologies Private Limited, Hisar, Haryana, on Feb. 28, for 'a hierarchical multi-agent ai system for adaptive traffic signal control with temporal prediction.'
Inventor(s) include Amit Dua.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "The present invention discloses a hierarchical multi-agent artificial intelligence system for adaptive traffic signal control employs Large Language Models at three decision-making levels: Claude Opus for strategic city-wide planning (hours timescale), Claude Sonnet for tactical corridor coordination (minutes timescale), and Claude Haiku for operational junction control (seconds timescale). The system integrates a Temporal Graph Neural Network that predicts traffic states with 6.01 km/h mean absolute error over a 30-minute horizon. Multi-source data fusion combines real-time information from Google Maps and TomTom APIs using confidence-weighted Kalman filtering. Digital twin validation in SUMO simulation demonstrates 352% throughput improvement, 65% delay reduction, and 17% speed increase compared to fixed-time signals (p 0.001). The 30-second operational cycle enables real-time adaptation while maintaining interpretable natural language decision-making. Validated deployment on six-junction Bangalore Outer Ring Road network confirms substantial performance improvements."
Disclaimer: Curated by HT Syndication.
