Intellectual Property India Publishes Patent Application for 'Proactive Traffic Signal Optimization With Multi-Intersection Coordination Using Vision-Based Demand Forecasting' Filed by Jadhav Dhiraj Shashikant; Nama Omkar Vilas; Nikhare Shreyash Vinod; Khan Moin Usman; and Mohol Vedant Bharat

MUMBAI, India, Feb. 6 -- Intellectual Property India has published a patent application (202521123404 A) filed by Jadhav Dhiraj Shashikant; Nama Omkar Vilas; Nikhare Shreyash Vinod; Khan Moin Usman; and Mohol Vedant Bharat, Pune, Maharashtra, on Dec. 8, 2025, for 'proactive traffic signal optimization with multi-intersection coordination using vision-based demand forecasting.'

Inventor(s) include Jadhav Dhiraj Shashikant; Nama Omkar Vilas; Nikhare Shreyash Vinod; Khan Moin Usman; and Mohol Vedant Bharat.

The application for the patent was published on Feb. 6, under issue no. 06/2026.

According to the abstract released by the Intellectual Property India: "Proactive Traffic Signal Optimization with multi-intersection coordination Using Vision-Based Demand Forecasting is an integrated, AI-driven system for real-time optimization of urban traffic signals using live video, deep learning, and multi-agent control. It combines high-resolution camera feeds with a YOLOv8 detector to estimate lane-wise vehicle counts, queue lengths, and densities with over 95% accuracy, and augments these measurements with LSTM-based forecasts of short-term traffic inflow so that controllers can anticipate demand surges rather than react only to current congestion. Each intersection runs a Deep Q-Network agent that selects signal phases at fixed decision intervals while exchanging summarized inflow/outflow data with neighboring intersections, creating coordinated green waves and reducing queue spillovers across the network. Deployed in a hybrid edge-central architecture, Proactive Traffic Signal Optimization with multi-intersection coordination Using Vision-Based Demand Forecasting executes perception and control on intersection-level edge devices with sub-200 ms decision latency, while a central server aggregates data, retrains models, and hosts a supervisory dashboard. Simulations on a sixteen-intersection (4x4) grid with 500-2,000 vehicles per hour per lane demonstrate 40-55% lower average waiting time, 35-45% higher throughput, and roughly 30-45% lower estimated fuel consumption than fixed-time and actuated controllers, making our system a practical, scalable candidate for next-generation smart city traffic management."

Disclaimer: Curated by HT Syndication.