MUMBAI, India, Oct. 26 -- Intellectual Property India has published a patent application (202611112642 A) filed by Dr. Nisha Aggarwal; Dr. Yogesh Sharma; Prof. Vikram Bali; Mr. Ashish Sharma; Dr. Sandeep Tayal; Ms. Savita Sharma; and Dr. Farzil Kidwai on September 19, 2026, for Self-Healing Sensor-Adaptive Neural Network System For Edge And Iot Devices.
Inventors include Dr. Nisha Aggarwal; Dr. Yogesh Sharma; Prof. Vikram Bali; Mr. Ashish Sharma; Dr. Sandeep Tayal; Ms. Savita Sharma; and Dr. Farzil Kidwai.
The application for the patent was published on October 23, 2026, under issue no. 43/2026.
Abstract: A self-healing sensor-adaptive neural network system (100) and method for operating a neural network on an edge or IoT device (110) with multiple input sensors (120). The system monitors a runtime health signal per sensor channel, detects channel-specific degradation, and identifies the dependent network substructure (134) via an attribution engine (150). A redundancy determination module (160) classifies the degraded channel as redundant or essential. If redundant, a structural adaptation module (170) prunes or reroutes the affected substructure during live inference without retraining. If essential, a confidence adjustment module (180) reduces output confidence and optionally issues a maintenance alert. Upon channel recovery confirmed over a predetermined period by a recovery and regrowth module (190), previously pruned structure is reintroduced. The combination of channel-specific attribution, redundancy-aware branching, and reversible regrowth enables autonomous topology adaptation on resource-constrained devices.
Disclaimer: Curated by HT Syndication.
