RENFORCE DYNAMICS

Building humanoid intelligence from the ground up.

We are an open robotics team dedicated to advancing humanoid motion intelligence: the ability of bipedal machines to move with agility, react to unexpected contact, and operate in unstructured real-world environments. We believe general intelligence in robots will emerge bottom-up — starting from robust physical controllers and expanding through interaction, imitation, and experience.

Our current focus spans dynamic locomotion, parkour-like agility, whole-body balance under perturbation, and the sim-to-real transfer of contact-rich skills. We build open tools, policies, and datasets that lower the barrier to capable, real-world humanoids.

GitHub Our Approach

Our Approach to Humanoid Intelligence

We pursue intelligence from the bottom up: first mastering physical interaction, then scaling behavior through learning and real-world experience.

Dynamic Whole-Body Locomotion

We develop controllers that let humanoid robots walk, run, jump, vault, and recover balance over uneven terrain and obstacles. Our focus is on whole-body coordination under dynamic contact changes, rather than constrained flat-ground gaits.

Bottom-Up Skill Acquisition

We combine reinforcement learning, motion imitation, and skill composition to grow a repertoire of behaviors from simple primitives to complex, chained maneuvers. Skills emerge from physics-based interaction rather than top-down scripting.

Sim-to-Real Transfer

We close the gap between simulation and reality through domain randomization, actuator modeling, and perception-driven closed-loop control. Our goal is policies that deploy reliably on physical humanoid hardware without exhaustive on-robot tuning.

Research Focus

We tackle the problems that matter most for capable, real-world humanoids.

Dynamic Locomotion Parkour & Agile Maneuvers Whole-Body Balance & Recovery Motion Imitation Reinforcement Learning Skill Composition Sim-to-Real Transfer Perception for Locomotion Contact-Rich Control Embodied Intelligence

Partner With Us

We are actively seeking collaborators to advance humanoid robotics — from dynamic locomotion and sim-to-real transfer to the foundations of embodied intelligence. Whether you are an academic lab, an industry team, or an open-source contributor, we welcome your expertise.

Initiate Discussion (Email) Open a GitHub Issue