CIM WELCOMES NEW ASSOCIATE MEMBER PROFESSOR ADITYA MAHAJAN
CIM is extremely pleased to announce the addition of Prof. Aditya Mahajan as Associate Member of our Centre.
Prof. Mahajan is an assistant professor in the Department of Electrical and Computer Engineering. He is a graduate of the University of Ann Arbour, Michigan, where he completed his PhD in Electrical and Computer Engineering in 2008, and then went on to complete a post-doctoral fellowship at Yale University.
The main focus of Professor Mahajan’s research is to develop an axiomatic framework for sequential decision making in decentralized multi-agent systems. Such systems arise in almost all technological systems including control systems, communication systems, sensor networks, and robotics.
The fundamental conceptual difficulty in the design of these systems is the following: To choose an optimal decision making policy, an agent must pick the best action; to pick the best action, it must evaluate the utility of each action; to evaluate the utility of an action, it must know how other agents will react to that action--thus know the decision making policy of other agents. Hence, to choose their policies, all agents must second guess the policies of others, leading to infinite regress. Even if a set of equilibrium policies are found, they are only person-by-person (or locally) optimal and may not be jointly optimal.
Professor Mahajan is developing a solution methodology that circumvents this difficulty by looking at the system from the point of view of a fictitious agent, called the coordinator, whose observations are common knowledge to all agents. From the point of view of this coordinator, the system as a single agent partially observable Markov decision process and we can then use existing solution techniques for partially observable Markov decision processes to find optimal decision making policies of all agents. This methodology has been effectively applied in applications in real-time communication, networked control systems, and sensor scheduling. This methodology is also useful for path planning and task scheduling in multi-robot systems.
Welcome aboard Aditya!