The field of humanoid robotics is witnessing significant advancements, with a focus on developing more robust, versatile, and autonomous systems. Recent research has explored innovative control methods, such as Latent Space Backward Planning and Coupled Hierarchical Diffusion, to improve the efficiency and accuracy of robotic planning. Moreover, the integration of reinforcement learning and imitation learning has led to more effective policy fine-tuning, as seen in approaches like IN-RIL. Noteworthy papers in this area include the presentation of Zippy, the smallest self-contained bipedal walking robot, and the introduction of FALCON, a dual-agent reinforcement-learning-based framework for robust force-adaptive humanoid loco-manipulation. These developments are paving the way for more sophisticated and capable humanoid robots that can perform complex tasks in various environments.
Advances in Humanoid Robotics and Control
Sources
Zippy: The smallest power-autonomous bipedal robot
Demystifying Diffusion Policies: Action Memorization and Simple Lookup Table Alternatives
Let Humanoids Hike! Integrative Skill Development on Complex Trails
JAEGER: Dual-Level Humanoid Whole-Body Controller
FALCON: Learning Force-Adaptive Humanoid Loco-Manipulation
Efficient Robotic Policy Learning via Latent Space Backward Planning
FACET: Force-Adaptive Control via Impedance Reference Tracking for Legged Robots
CHD: Coupled Hierarchical Diffusion for Long-Horizon Tasks
HuB: Learning Extreme Humanoid Balance
Motion Control of High-Dimensional Musculoskeletal Systems with Hierarchical Model-Based Planning
RT-cache: Efficient Robot Trajectory Retrieval System
Latent Theory of Mind: A Decentralized Diffusion Architecture for Cooperative Manipulation
Train a Multi-Task Diffusion Policy on RLBench-18 in One Day with One GPU
Learning Long-Context Diffusion Policies via Past-Token Prediction
IN-RIL: Interleaved Reinforcement and Imitation Learning for Policy Fine-Tuning