DreamControl: Human-Inspired Whole-Body Humanoid Control for Scene Interaction via Guided Diffusion
This work addresses the challenge of enabling humanoid robots to perform complex, natural-looking whole-body interactions in real-world scenarios, representing a novel integration of methods rather than an incremental improvement.
The paper tackles the problem of autonomous whole-body humanoid control for scene interaction by introducing DreamControl, which combines a diffusion prior trained on human motion data with reinforcement learning to guide policy learning in simulation, resulting in the robot successfully completing challenging tasks like opening drawers and picking up objects on a Unitree G1 robot.
We introduce DreamControl, a novel methodology for learning autonomous whole-body humanoid skills. DreamControl leverages the strengths of diffusion models and Reinforcement Learning (RL): our core innovation is the use of a diffusion prior trained on human motion data, which subsequently guides an RL policy in simulation to complete specific tasks of interest (e.g., opening a drawer or picking up an object). We demonstrate that this human motion-informed prior allows RL to discover solutions unattainable by direct RL, and that diffusion models inherently promote natural looking motions, aiding in sim-to-real transfer. We validate DreamControl's effectiveness on a Unitree G1 robot across a diverse set of challenging tasks involving simultaneous lower and upper body control and object interaction. Project website at https://genrobo.github.io/DreamControl/