K. Sreenath

This work proposes residual reinforcement learning as an implicit safety-filtering mechanism for safe humanoid control and shows that the residual policy can act as an implicit safety filter.

Safe-Stop enables humanoid robots to make safer emergency stop decisions by evaluating stoppability in real-time, rather than relying on fixed maneuvers.

ODEWorld achieves high-quality long-horizon predictions and planning-oriented dynamics by seamlessly integrating continuous-time modeling with ODE-based representations.

Action QFormer boosts navigation success rates from 18.8% to 56.3% by intelligently reorganizing multimodal information under action supervision.

Synthesizing 48,000 interaction trajectories without human input enables a humanoid robot to learn complex loco-manipulation tasks effectively.

Steering worst-case trajectories with an adversarial network and Boltzmann reweighting dynamics ensembles yields a surprisingly stable and efficient approach to robust RL under dynamics uncertainty.

Achieve natural and dynamically feasible humanoid robot motion by retargeting human motion data with a skeleton-aligned approach, significantly reducing inverse kinematics error compared to task-space methods.