Session 3: Computing and Multi-Agent Cooperation

In this short talk, I will share my invaluable experience that I gained from my interactions with Prof Debasish Ghose during my PhD days and how it significantly motivated and facilitated my decision to choose an academic career. I will reflect on my PhD work and the impact it has had on my professional path till date and how it placed me in a position where I am today. I will also highlight the lasting influence of Dr. Debasish Ghose's mentorship on my research and teaching philosophy. 

Multi-robotic systems find application in many fields such as search and rescue, agriculture, exploration and coverage of unknown environment, and many more. Cooperating between individual robots/agents is one of the important challenges in multi-robotic systems research. Voronoi partition is a very useful tool in achieving a natural cooperating between robots and optimality in task allocation among the robots in a multi-robotic system. This talk explores a few problems in multi-robotic systems solved used Voronoi partition as a tool. 

In contested environments where GPS and traditional navigation aids may be degraded or denied, collaborative navigation offers a resilient alternative. This work explores  algorithms that enable multiple autonomous agents to share sensor data and state estimates to improve positioning accuracy and robustness. By leveraging inter-agent communication, onboard sensing, and cooperative estimation techniques, the proposed framework enhances situational awareness and navigation reliability in contested environments.

Multi-agent cooperation is essential to distribute tasks among themselves for efficient execution. However, the decision-making becomes difficult under fuel, sensor, communication and localisation constraints. In this talk, I will show how we can use reinforcement learning framework for persistent monitoring and auction-based methods for task allocation while reducing the number of messages.