Ishan Durugkar

Papers

Total Citations

H-Index

About

Ishan Durugkar is a researcher advancing the frontiers of reinforcement learning and robotics, with a focus on intrinsic motivation and real-world deployment. His key contributions lie in developing algorithms that enable agents to learn more efficiently by leveraging adversarial objectives. In his highly cited work, "Adversarial Intrinsic Motivation for Reinforcement Learning" (2021, 4 citations), Durugkar explores how minimizing the Wasserstein-1 distance between a policy's state visitation distribution and a reference distribution can drive exploration and imitation, bridging ideas from generative modeling to RL. This approach offers a principled way to shape agent behavior without explicit rewards. More recently, he has tackled practical challenges in robotics with "Towards a Real-Time, Low-Resource, End-to-End Object Detection Pipeline for Robot Soccer" (2023, 3 citations), demonstrating his commitment to deploying efficient, low-latency vision systems in competitive, resource-constrained environments. Durugkar's work stands out for its blend of theoretical insight and applied impact, making him a notable figure in the intersection of reinforcement learning, intrinsic motivation, and robotic perception.