Dr. Alexander V. Schperberg

We present a whole-body loco-manipulation controller combining model-based admittance control for the manipulator with a reinforcement-learning locomotion policy for the legs. Experiments on a Unitree Go2 quadruped with a 6-DoF arm demonstrate compliant interaction, accurate force tracking, and safe operation during dynamic human-robot interaction.

We present an energy-efficient motion planner for legged robots that improves trajectory generation while accounting for robustness and hardware-relevant constraints. Results demonstrate effective and practical planning performance for legged locomotion.

We propose GRAM, a deep reinforcement learning framework that unifies adaptive and robust generalization across in-distribution and out-of-distribution dynamics via a robust adaptation module. Experiments on a quadruped robot demonstrate strong performance and reliable deployment across diverse environments.

State estimation for legged robots is challenging due to their highly dynamic motion and limitations imposed by sensor accuracy. By integrating Kalman filtering, optimization, and learning-based modalities, we propose a hybrid solution that combines proprioception and exteroceptive information for estimating the state of the robot’s trunk. Leveraging joint encoder and IMU measurements, our Kalman filter is enhanced through a single-rigid body model that incorporates ground reaction force control outputs from convex Model Predictive Control optimization. The estimation is further refined through Gated Recurrent Units, which also considers semantic insights and robot height from a Vision Transformer autoencoder applied on depth images. This framework not only furnishes accurate robot state estimates, including uncertainty evaluations, but can minimize the nonlinear errors that arise from sensor measurements and model simplifications through learning. The proposed methodology is evaluated in hardware using a quadruped robot on various terrains, yielding a 65% improvement on the Root Mean Squared Error compared to our VIO SLAM baseline.

SCALER is a versatile multilimbed robot capable of free-climbing in extreme terrains. The platform integrates legged locomotion with grasping capabilities, enabling robust traversal and manipulation in complex environments.