MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641109716 A) filed by Syed Shouib Ahmed; Shifa Anjum; Rama Devi Ks; and Sajida Nazneen on September 12, 2026, for System And Method For Hierarchical Reinforcement Learning With Context-Sensitive Policy Switching.
Inventors include Syed Shouib Ahmed; Shifa Anjum; Rama Devi Ks; and Sajida Nazneen.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: The present invention relates to a system and method for hierarchical reinforcement learning with context-sensitive policy switching. The system comprises an input interface, a sensing unit, a memory unit, a processing unit, an interaction interface, and a feedback processing unit. The sensing unit acquires environmental observation data and context data, while the memory unit stores hierarchical control policies including higher-level policies for operational objectives and lower-level policies for execution of corresponding subtasks. The processing unit generates a current state representation, selects an objective and policy level, retrieves a corresponding policy, and generates a control action. The processing unit selectively switches between policies based on context conditions, temporal validity, confidence information, and switching history while maintaining an intermediate control state. The interaction interface provides control actions and receives resultant state and feedback information, and the feedback processing unit updates policy states based on the received information.
Disclaimer: Curated by HT Syndication.
