Yesh Godse
Papers
2
Total Citations
8
H-Index
2
About
Yesh Godse is a roboticist whose research lies at the intersection of reinforcement learning and dynamic legged locomotion, with a focus on bridging the gap between simulation and the real world. In his highly cited work, "Sim-to-Real Learning of All Common Bipedal Gaits via Periodic Reward Composition" (2021), Godse tackled the challenge of teaching a physical bipedal robot the full spectrum of natural gaits—from walking to running—using sim-to-real RL. His key innovation was designing intuitive, periodic reward functions that allowed a single policy to produce multiple distinct gaits, a breakthrough that has garnered 5 citations. In his complementary work, "Learning Spring Mass Locomotion: Guiding Policies With a Reduced-Order Model" (2021, 3 citations), Godse introduced a hierarchical control framework that combines the theoretical elegance of reduced-order models (like the spring-mass model) with the adaptability of RL. This approach enables high-level behavior planning through simplified dynamics while low-level policies handle real-world complexities. Godse’s contributions are notable for making complex bipedal control more accessible and robust, directly advancing the field of legged robotics toward practical, energy-efficient locomotion.
Research Focus
Key Achievements
Top Papers
- 1
- 2