Intellectual Property India Publishes Patent Application for "decentralized Multimodal Edge Ai Sytem For Real-Time Traffic Prediction, Adaptive Route Optimization, And Energy-Aware Control In Intelligent Transportation Networks''

MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202521039450 A) filed by Ms. Sweta Kahurke on April 24, 2025, for "decentralized Multimodal Edge Ai Sytem For Real-Time Traffic Prediction, Adaptive Route Optimization, And Energy-Aware Control In Intelligent Transportation Networks''.

Inventors include Ms. Sweta Kahurke; and Prof. Dr. Ktv Reddy.

The application for the patent was published on October 30, 2026, under issue no. 44/2026.

Abstract: ABSTRACT OF INVENTION: The invention discloses a comprehensive and modular multimodal hardware and AI-based system designed for traffic prediction and adaptive route optimization within Intelligent Transportation Systems (ITS). The system comprises a decentralized network of edge compute nodes, each equipped with a heterogeneous sensor fusion suite and integrated AI algorithms, enabling real-time, privacy-preserving, and context-aware traffic management at the edge. Each node is powered by a high-performance, energy-efficient edge compute unit (e.g., NVIDIA Jetson Orin NX), capable of processing data from multiple sensors including thermal cameras, LIDAR, radar, GNSS modules, and V2X communication devices. These multimodal data streams are fused using a custom transformer-based deep learning model, termed MMTFormer, which performs real-time scene understanding by leveraging cross-modal attention and contrastive loss mechanisms. This results in accurate detection of vehicles, pedestrians, road types, and traffic density under diverse environmental conditions. For dynamic traffic flow management, the system incorporates a decentralized, multi-agent reinforcement learning module (MARTIN) that allows nodes to cooperatively negotiate traffic behaviors at intersections, significantly reducing congestion and wait times. In parallel, an energy-aware sensing controller (EcoSenseNet) dynamically manages node power consumption using an LSTM-based energy forecast and FSM-controlled duty cycling, ensuring long-term operation in energy-constrained scenarios. To enhance route personalization while maintaining user data privacy, a federated learning module (FedRoute++) enables vehicles to collaboratively learn optimal routing strategies without transmitting raw GPS data, utilizing differential privacy mechanisms. Furthermore, the system includes a context-driven policy switch engine (CoPoSwitch), which dynamically adjusts traffic signals and routing decisions in response to critical events such as the detection of emergency vehicles or temporary hazards in the process. This integrated framework, referred to as MITRO-Fx, provides a scalable and intelligent infrastructure for modern ITS, supporting real-time multimodal data fusion, cooperative traffic control, and adaptive route optimization in a fully decentralized and privacy-respecting manner.

Disclaimer: Curated by HT Syndication.