V Shivaraman (शिवरामन्)

Hello! I am a research scholar at the Artificial Intelligence and Robotics Lab (AIRL) in the Department of Aerospace Engineering at the Indian Institute of Science (IISc) Bangalore, under the supervision of Prof. Suresh Sundaram. Previously, I was a Master's thesis student at AIRL, IISc Bangalore, supervised by Prof. Suresh Sundaram and Prof. Sujit P B.

I received my MS and BS in Electrical Engineering and Computer Science (EECS) with minors in Physics in 2023 from the Indian Institute of Science Education and Research Bhopal (IISER Bhopal).

I am keen on developing safe, robust learning-based algorithms for decision-making tasks in intelligent autonomous systems, with applications in autonomous navigation and biomedical systems.

 |   |  CV  | 

profile photo

Publications

Graph-based Prediction and Planning Policy Network (GP3Net) for scalable self-driving in dynamic environments using Deep Reinforcement Learning

Jayabrata Chowdhury*, V Shivaraman*, Suresh Sundaram, Sujit P B

Proceedings of the 38th AAAI Conference on Artificial Intelligence (AAAI). February 2024.

* equal contribution




Projects

Classical Reinforcement Learning Algorithms for Sokoban

I applied Reinforcement learning algorithms like Q-Learning, SARSA and REINFORCE to train an agent that can strategically push boxes onto target locations in complex environments. The agent learns to navigate through the game’s challenges, improving its performance over time through a process of trial and error.

Code



Risk-based Task Allocation: Resilient Coverage

Implemented a centralized framework to recalibrate and reposition operational robots to maintain optimal surveillance and minimize costs. In an event of robot failure, the framework repositions robots in a local neighbourhood of the failure to minimize the coverage loss. The implementation is based on paper that formulates the problem this as a submodular function optimization problem, solves it using a Greedy algorithm, and uses Mixed Integer Linear Programming for optimal fleet selection.

Code



Approximation Algorithms to Minimize Interference of Wireless Sensor Networks

Minimizing total interference (redundant connections) is a crucial problem in wireless networks. The efficiency and power consumption is linked directly to the interference. This problem is two dimensions is a NP-Hard problem. In this project, I study the complexity and optimality of a few algorithms proposed in literature. I have also implemented the algorithms to test their effectiveness empirically.