Evolving robots that learn to manipulate objects without preset parts
Evolving Dexterous Robots from Scratch
RoboticsMachine Learning
Summary
Designing robot hands to handle objects is tricky, and copying animal body parts doesn’t always work well for machines. The authors developed a method where robot shapes and movements evolve freely from scratch, without starting with any set parts like fingers or claws. They used artificial intelligence techniques to explore many robot designs and trained the best ones to pick up, hold, and move different objects. Some evolved robots even made it from computer design to physical machines that worked right away. This approach can discover entirely new robot shapes suited for handling tasks.
What this means in practice
- •For roboticists: Design robots that evolve novel hands and manipulators without predefining body parts, improving ability to handle varied objects.
- •For robot manufacturers: Automatically convert evolved robot designs into printable blueprints and build them without redesigning for physical testing.$Commercial implications: Enables robotic hardware companies to create custom dexterous robots efficiently from AI-generated designs, shortening development time.
Authors
Zihan Guo, Shuzhe Zhang, Muhan Li, Peiyang Li, Sam Kriegman
Abstract
Little is known about how to manually design agents capable of dexterous manipulation. Some design principles have been inferred from close examination of how animals manipulate objects, but these structures and behaviors have so far resisted biomimicry and may not be optimal for artificial machines. Here we evolve freeform robots to pick up, hold, rotate, and use diverse objects. Unlike other approaches to optimizing robot hands, we do not presuppose the presence, articulation, or geometry of any part of the body. Although familiar prehensile forms such as tails, beaks, paws and claws may emerge spontaneously under certain conditions--and while such conditions could be of interest to evolutionary biologists--de novo manipulator design can also reveal whole new solutions, overlooked or unknown structures which may be better suited for the task at hand. We use contrastive learning to create a highly searchable genetic embedding of design space, an autoregressive developmental model to decode designs, evolutionary strategies to find good designs, and reinforcement learning to train each evolved design. Winning designs were automatically converted into a manufacturable blueprint, printed, assembled and tested in the real world in a zero-shot manner. The results represent the state-of-the-art in evolutionary robotics in terms of performance, diversity and complexity.