Nico Bohlinger

Nico Bohlinger joined the Intelligent Autonomous System lab on July 15, 2023, as a PhD student.

Nico is working on learning embodied intelligence by applying Reinforcement Learning (RL) at scale to robotics.
Besides many side projects, his core work focuses on:

• The Unified Robot Morphology Architecture (URMA), a fully morphology-aware neural network architecture, to learn a single policy to control any legged robot embodiment. This approach shows embodiment scaling laws and can be scaled to millions of embodiments.
• Learning natural and omnidirectional locomotion in just 8 minutes with efficient off-policy RL directly on a real quadruped robot.
• Currently: Training VLAs for Humanoid Loco-Manipulation by combining VLMs with online RL in large scale randomized simulation environments.
• Currently: Enabling robot co-design through embodiment-aware value functions trained with multi-embodiment RL.

His main robot platforms are currently the Unitree H1, Unitree G1 and Booster T1 humanoids and the Unitree Go2 and MAB Silver Badger quadrupeds.

Before he started his PhD, Nico obtained a bachelor's degree in Business Informatics from the Offenburg University of Applied Sciences and a master's degree in Bioinformatics from the Goethe University Frankfurt.
He was previously part of the RoboCup teams Magma and Sweaty in Offenburg and developed RL environments to learn complex soccer skills. Furthermore, he is developing the research-focused RL framework RL-X.