Lokesh Krishna

Indian Institute of Technology BHU, Banaras Hindu University, Viterbo University

Papers

Total Citations

H-Index

About

Lokesh Krishna is a robotics researcher specializing in legged locomotion, reinforcement learning, and robust control for bipedal and quadrupedal robots. His work has made significant strides in developing lightweight, deployable control frameworks that enable robots to navigate challenging real-world terrains without relying on computationally expensive methods. Krishna's most influential contribution demonstrates that surprisingly simple linear policies are sufficient to achieve robust bipedal walking on uneven and sloped terrains — a counterintuitive result that has garnered 29 citations and reshaped assumptions about the complexity required for effective locomotion control. Building on this insight, his 2021 work extended linear policy learning to varying slopes using gradient-free optimization techniques like Augmented Random Search, earning 14 citations. His earlier research on quadrupedal locomotion established the foundation for this philosophy, applying linear policies to low-cost hardware platforms such as Stoch 2. More recently, Krishna has ventured into sim-to-real transfer with DiffCoTune, addressing the persistent challenge of bridging simulation and real-world deployment through differentiable co-tuning. Collectively, his research advocates for elegant, efficient solutions in robot control — proving that simplicity, when thoughtfully designed, can be remarkably powerful.