Multi-fingered robotic hands hold an immense potential for flexible manipulation in human environments. From using tools like scissors to preparing food, handling cloth, and robustly tying shoelaces, many mundane activities are enabled by – or become more flexible with – additional degrees of freedom, allowing for in-hand manipulation and dexterous environment interaction. With the ongoing waves of advancement in artificial intelligence, learning techniques, robust teleoperation, the proliferation of affordable tactile sensors, and the expansion of diverse datasets, we are closer than ever to overcome major existing challenges and unlock unprecedented levels of dexterity in robotic manipulation.However, we also acknowledge the current state of research: In practical applications, multi-fingered hands usually demonstrate a lower grasping and manipulation success rate compared to two-fingered parallel grippers; more robotic datasets and benchmarks focus on grippers than on multi-fingered hands; and grippers already serve a wide range of industrial applications. In light of these challenges, this workshop is dedicated to exploring a critical question: how can we overcome the challenges of dexterous manipulation and unlock the potential of multi-fingered hands?
This workshop intends to answer this question in four key aspects:
1. Investigating Manipulation Limitations: Understanding the constraints linked to mechanical design and control
2. Examining Advanced Algorithms: Reviewing state-of-the-art algorithms for dexterous manipulation
3. Enhancing Dexterity and Adaptability: Exploring new paradigms to improve the dexterity and adaptability of robotic hands
4. Envisioning the Future: Considering how foundational models can revolutionize dexterous manipulation
The objective of this workshop is to foster collaboration among researchers at all career stages with a strong interest in robotics, artificial intelligence, control theory, mechanical design, and related fields.
