MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202641044888 A) filed by Pwin Technologies Private Limited on April 08, 2026, for System And Method For Adaptive Training Of Autonomous Agents Using Multi-Source Feedback.
Inventors include Pratap Kumar Behera; and Chinmoy Kumar Nanda.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: ABSTRACT A system (106) and method for adaptive training of autonomous agents (110) using multi-source reinforcement is disclosed. The method comprises performing actions by the autonomous agents (110) and receiving a plurality of feedback signals corresponding to the actions, the plurality of feedback signals comprising an environmental reward signal, a human feedback signal, and an artificial intelligence-based critic evaluation signal. Intermediate evaluation parameters are generated, and a first composite reinforcement metric is computed based on the intermediate evaluation parameters. A similarity measure between a current action and stored prior actions is determined, and the first composite reinforcement metric is modified to generate a second composite reinforcement metric. Alignment of the actions with predefined behavioral constraints is evaluated to generate an alignment score, and a policy update is determined based on the second composite reinforcement metric and the alignment score to modify parameters of the autonomous agents (110).
Disclaimer: Curated by HT Syndication.
