MUMBAI, India, Oct. 12 -- Intellectual Property India has published a patent application (202641118299 A) filed by Dr. Debasis Mohanta; Mr. Susant Roul; Dr. Jitendra Patra; Miss. Ankita Patra; Dr. Animesh Talapatra; Dr. Paramita Biswas; Dr. Paramita Bhowmik; Dr. Bankim Chandra Rudra; Dr. Bhanumati Sarkar; Dr. Nithar Ranjan Madhu; and Mr. Utpal Madhu on October 03, 2026, for Adaptive Autonomous Weed Intervention System With Uncertainty-Constrained Multi-Modal Actuation And Verified Outcome Feedback.
Inventors include Dr. Debasis Mohanta; Mr. Susant Roul; Dr. Jitendra Patra; Miss. Ankita Patra; Dr. Animesh Talapatra; Dr. Paramita Biswas; Dr. Paramita Bhowmik; Dr. Bankim Chandra Rudra; Dr. Bhanumati Sarkar; Dr. Nithar Ranjan Madhu; and Mr. Utpal Madhu.
The application for the patent was published on October 09, 2026, under issue no. 41/2026.
Abstract: The present invention discloses an adaptive autonomous weed-intervention system in which perception uncertainty is converted into machine-enforced spatial and actuator constraints before a physical treatment is selected. A mobile robotic platform acquires crop-and-weed imagery and determines crop geometry, weed geometry, target position, and one or more uncertainty values. An electronic controller generates a Dynamic Crop-Weed Intervention Envelope (DCWIE) defining a crop- protected region and, for each candidate intervention, determines a predicted intervention occupancy comprising a swept mechanical-tool region, localized dispensing footprint, or combined hybrid footprint. Candidate actions whose predicted occupancy violates the DCWIE, actuator workspace, environmental limits, or protected mission-resource reserve are rejected or parameter-constrained before optimization. The remaining feasible action set is evaluated to select a mechanical intervention, localized treatment-material dispensing intervention, coordinated hybrid intervention, or controlled deferred intervention together with target-specific actuator commands. After execution, the treated region is reacquired and evaluated using intervention-specific verification criteria. A verified partial, failed, or indeterminate outcome causes modified retry, cross-modal escalation, additional sensing, or deferred reacquisition rather than being treated as success merely because an actuator was activated. Verified outcomes, actuator settings, local field conditions, and regrowth observations are stored in spatially indexed Adaptive Field Memory and are used to modify subsequent safety margins, feasibility thresholds, actuator parameters, treatment priorities, or revisit states. The architecture thereby establishes a closed physical control loop from uncertain perception to admissible actuator motion and treatment, verified physical outcome, and later control adaptation while protecting crop safety and finite mission resources.
Disclaimer: Curated by HT Syndication.
