Advances in Robotic Manipulation and Control

The field of robotic manipulation and control is rapidly advancing, with a focus on developing systems that can learn efficiently from minimal human input and adapt to real-world uncertainties and diverse embodiments. Recent research has explored the use of vision-language grounding, task-aware decomposition, and hybrid diffusion models to enable generalizable bimanual manipulation and long-horizon task planning. Notable papers in this area include VLBiMan, which introduces a framework for deriving reusable skills from a single human example, and Hybrid Diffusion, which proposes a novel mix of discrete variable diffusion and continuous diffusion for simultaneous symbolic and continuous planning. Other noteworthy papers include DemoGrasp, which presents a simple yet effective method for learning universal dexterous grasping, and Super-Mimic, which enables zero-shot robotic imitation by directly inferring procedural intent from unscripted human demonstration videos. These advances have the potential to significantly improve the capabilities of robotic systems and enable more efficient and effective manipulation and control.

Sources

VLBiMan: Vision-Language Anchored One-Shot Demonstration Enables Generalizable Robotic Bimanual Manipulation

Hybrid Diffusion for Simultaneous Symbolic and Continuous Planning

DemoGrasp: Universal Dexterous Grasping from a Single Demonstration

From Watch to Imagine: Steering Long-horizon Manipulation via Human Demonstration and Future Envisionment

Learning to Ball: Composing Policies for Long-Horizon Basketball Moves

JointDiff: Bridging Continuous and Discrete in Multi-Agent Trajectory Generation

Control Your Robot: A Unified System for Robot Control and Policy Deployment

Annotation-Free One-Shot Imitation Learning for Multi-Step Manipulation Tasks

RoboPilot: Generalizable Dynamic Robotic Manipulation with Dual-thinking Modes

TubeDAgger: Reducing the Number of Expert Interventions with Stochastic Reach-Tubes

Symskill: Symbol and Skill Co-Invention for Data-Efficient and Real-Time Long-Horizon Manipulation

ARMADA: Autonomous Online Failure Detection and Human Shared Control Empower Scalable Real-world Deployment and Adaptation

Built with on top of