My interests lie in dealing with the problem of sequential decision making in real world uncertain domains.  

In my thesis, I have used Partially Observable Markov Decision Problems (POMDPs) and Distributed POMDPs for modeling these kinds of uncertain domains. My contributions are in providing efficient techniques for solving these models, while providing error bounds on the quality of the solutions. Given below are some of my key contributions: 

(i) The first part of my work provides techniques that exploit structure in the dynamics of the domain to efficiently solve Partially Observable Markov Decision Problems (or POMDPs) exactly. This work was done in the context of the addressing the problem of Adjustable Autonomy in CALO (Cognitive Agent that Learns and Organizes) project [publication  in AAMAS-05]. 

(ii) My second contribution focuses on providing an technique for solving POMDPs by approximating in the value space. This technique provides a novel approach to approximation in solving POMDPs that facilitates providing error bounds on the quality of solutions [publication in IJCAI-07]. 

(iii) Thirdly, I have provided efficient techniques for solving distributed POMDPs. The key idea was to exploit structure in the interactions of the agents. As part of this work, we provided techniques for solving distributed POMDPs, that were inspired from DCOP algorithms (DPOP, DBA and DSA). [publication  in AAAI-05].

(iv) The fourth contribution is in providing efficient locally optimal techniques for solving distributed POMDPs with continuous belief spaces [publication in AAMAS-06]. 

Other things that I have worked on: 

(i) Applying asimovian principles to teams of humans and agents [publication  in  PROMAS-06].

(ii) Measuring privacy loss in the context of cooperative Multi Agent settings [ ex: meeting scheduling in office environments] [publication  in  AAMAS-05].

(iii) In my first semester at TEAMCORE, I worked on taking DCOP (Distributed Constraint Optimization) to real world domains (Meeting Scheduling and Distributed Sensor Networks) [publication  in AAMAS-04].