Versatile & Efficient
Created to be an atomic skill library based on real-world implementation scenarios, AnySkill enhances the general applicability of basic skills, empowering robots to combine multiple skills for complex tasks and adapt seamlessly to dynamic environments. This streamlined capability accelerates the rapid development and deployment of robotics applications in diverse settings.
Typical Atomic Skill Library
General Object Grasping Skill
Without prior modeling, robots can rapidly grasp unknown objects, including rigid, deformable, and textureless objects.
General Deformable Object Folding Skill
Without prior modeling, robots can autonomously manipulate flexible items, such as towels and clothing, from recognition through unfolding to folding.
General Surface Shaving Skill
Without prior modeling, robots can perform tasks such as shaving and peeling, while adapting to surface variations for safe and precise manipulation.
General Object Opening Skill
Without prior modeling, robots can open various articulated objects, such as drawers, doors, and refrigerators.
General Object Scooping Skil
Without prior modeling, robots can rapidly scoop up unknown items, including diverse solids and liquids.
General Object Rearranging Skill
Without prior modeling, robots can rearrange unknown objects, such as shoes and tableware, based on user instructions.
Core Values
Convenient and User-friendly
A single skill offers strong versatility and general applicability, enabling rapid reuse and combination across diverse scenarios. This allows a streamlined skill set to handle complex tasks and environments.
Perception-driven
Noematrix uses a combination of visual perception and interactive feedback to adjust operational behaviour dynamically in response to real-time environmental information.
Universal Handling
Skill execution is model-free, allowing it to directly manipulate both rigid and deformable objects. This approach ensures a high level of task generalization, vastly extending a solutions utility.
