Tata Consultancy Services Limited

Hardik Meisheri

Harshad Khadilkar

Vinita Baniwal

Vinay Kulkarni

Souvik Barat

Prashant Kumar

Monika Gajrani

Omkar Dilip Shelke

Dheeraj Anilkumar Shah

Durgesh Kalwar

Pranavi Pathakota

Padmakumar Manikandasadanam Anandan Kartha

Harshad Dilip Khadilkar

The present disclosure provides a Reinforcement Learning (RL) based architecture to efficiently learn action embeddings in low dimensional space. In conventional methods, the embeddings are learnt with the sole objective of improving policy learning, and there are no specific requirements on the quality of the embeddings. Initially, the system receives a goal to be reached by a mobile robot and a current location of the mobile robot is obtained. Simultaneously current transition dynamics associated with the plurality of directional actuators are obtained using a Reinforcement Learning (RL) technique. Further, a plurality of embeddings is computed based on the current location of the mobile robot and the current transition dynamics using a trained Dual Channel Training (DCT) based autoencoder decoder model. Finally, a displacement vector for current navigation of the mobile robot is computed based on the computed plurality of embeddings using the RL technique.

Method and system for reinforcement learning and dual channel action embedding based robotic navigation

Generally, in complex systems such as business organizations and supply chain networks, decisions making and control are challenging due to dynamic environment thereof. The embodiments herein provide method and system that uses reinforcement learning (RL) for exploring policies and deciding control actions, and actor-based modelling and simulation for performing accurate long-term rollouts of the policies, in order to optimize operation of the complex systems. In an embodiment, a method for actor based simulation of complex system includes modeling, actors of the complex system using an actor abstraction, where actors includes RL agents embodied in each subsystem of the complex system, and executes micro-level interactions amongst the subsystems. An emergent macro-behavior of the complex system is simulated by the using the RL agents based on the micro-level interactions. Optimal decisions pertaining to actions of the RL agents are learnt based on the emergent macro-behavior of the complex subsystem.

System and method for actor based simulation of complex system using reinforcement learning

This disclosure provides a system and method for concurrent dynamic optimization of a replenishment decision in a networked node environment by receiving at least one current state attribute associated with the networked node environment at a predetermined time interval; obtaining an optimized entity actionable matrix by mapping the at least one of current state attribute and an associated actionable entity; computing a predictive replenishment matrix for an impending cycle of the networked node environment based on the optimized entity actionable matrix; identifying a reward function for the predictive replenishment matrix; and determining the replenishment suggestive actionable element based on the reward function. The reward function includes a computation of overall cost with regard to a realized revenue associated with aggregation of the plurality of nodes.

System and method for concurrent dynamic optimization of replenishment decision in networked node environment

Growing community of inventors

Thane, India

Hardik Meisheri