Shivam Sood

Indian Institute of Technology Kharagpur

Papers

Total Citations

H-Index

About

Shivam Sood is a rising roboticist whose work is reshaping how legged robots move—shifting the paradigm from rigid, position-based control to compliant, torque-driven locomotion. His research sits at the intersection of optimal control, deep reinforcement learning (DRL), and force-based actuation for quadrupedal robots. In his highly cited work, "DecAP: Decaying Action Priors for Accelerated Imitation Learning of Torque-Based Legged Locomotion Policies" (2024), Sood introduces a novel framework that accelerates imitation learning by leveraging decaying action priors, enabling robots to learn torque-based policies that are both robust and agile. This work, alongside "Force Control for Robust Quadruped Locomotion: A Linear Policy Approach" (2023), demonstrates his ability to distill complex control problems into elegant, scalable solutions—here, a simple linear policy that directly commands end-foot trajectories and centroidal wrenches for dynamic, force-driven locomotion. Though early in his career, Sood’s contributions are already gaining traction, with each paper cited twice, signaling growing interest from the robotics community. His focus on torque control and force-based policies addresses a critical bottleneck in legged robotics: achieving the compliance and adaptability needed for real-world terrain. For students and researchers, Sood’s work offers a clear path from theory to practice, proving that sometimes the most powerful controllers are the simplest.