Researchers at WMG, University of Warwick, have produced groundbreaking artificial intelligence algorithms that enable a robot to perform dexterous manipulation of objects just like humans. While dexterous manipulation of objects is a basic and perfunctory task for humans, it is still very difficult for autonomous robotic hands to master before this seeming breakthrough.
The Shadow Robot Dexterous Hand compares well with a human hand and is capable of reproducing all of its functionalities. Simulations showed that the robotic hands learn on their own how to coordinate movements and execute tasks of varying difficulty.
Robot hands can be used in many important applications needing precision and versatility in manufacturing, surgery, and activities involving radioactive materials. The utilization of robotic hands in assembly lines has yielded higher productivity and reduced risks exposure from work risk situations to human workers.
In a paper titled ‘Solving Challenging Dexterous Manipulation Tasks With Trajectory Optimisation and Reinforcement Learning’, Professor Giovanni Montana and Dr. Henry Charlesworth, both from WMG, the University of Warwick proposed new AI algorithms (also called the “brain”) to enable robots to learn how to coordinate the fingers’ movements and manipulate objects.
The versatility of the algorithms is that the robots can learn any task so far it can be simulated. The 3D simulations were developed using MuJoCo (Multi-Joint Dynamics with Contact), a physics engine developed by the University of Washington.
The researchers’ approach uses two algorithms – an initial planning algorithm and then a reinforcement learning algorithm. By taking this approach, the researchers were able to produce significantly better movements compared to current methodologies.
The research team is expected to continue working with Shadow Robot and deploy the algorithms on real robotic hardware, which could see the hand advance one step closer to use in real day to day life.
In another paper, ‘PlanGAN: Model-based Planning With Sparse Rewards and Multiple Goals’, to be presented at the 2021 NeurIPS conference, the WMG researchers have also developed a general AI approach that enables robots to learn various tasks in order to improve their hand manipulation capabilities.
Rich Walker, Managing Director of the Shadow Robot Company, based in London, stated that the company has come a long way in, stressing that the research group is now delivering the promise of the hardware by creating algorithms clever enough to control the robot hand – soon perhaps we will see a super-human performance. (via)
UP NEXT: Oppo X 2021 unveiled as the world’s first smartphone concept with an extending display
