Statement of Purpose

I have a strong desire to pursue multidisciplinary research for optimal decision making in uncertain environments. In particular, I am interested in devising efficient optimal solutions to sequential decision making tasks modelled as an MDP or Markov game. My research has been in the area of Hybrid Intelligent Systems based on Markov games, their applications to control and comparison with standard MDP based RL techniques. My doctoral work broadly covered Reinforcement learning algorithms and their control applications.

In general, the research question I intend to tackle is “Designing reliable, adaptive, robust, cost effective and intelligent control schemes for countering disturbances, noise and parameter variations in real time high dimensional systems”. The basic framework could be Reinforcement Learning which may include Markov games in conjunction with cutting-edge soft computing techniques like fuzzy inference systems, decision trees, support vector machines, and neural networks. Currently I am working on Decentralized POMDPs and Multi agent POMDPs with Fuzzy Reinforcement Learning.

I would like to extend my research to optimization of multi agent systems modeled as either a POMDP or a POSG, depending on the case. As a matter of fact, some of the algorithms that I worked on are more suited to a multi agent scenario. So far my research domain centered on finding adaptive optimal policies for uncertain environments wherein a single agent learns correct/optimal behavior through repeated interactions with the environment or the Reinforcement Learning paradigm. However, significant research has been done in multi agent RL wherein agents share the information gained from the environment to arrive at an optimal policy.