Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
and JavaScript.
A tendon-driven robotic limb learns movements autonomously from sparse experience, by a short period of ‘motor babbling’ (that is, repeated exploratory movements), followed by a phase of reinforcement learning. In the photo, the limb is learning to make cyclic movements to propel the treadmill. The approach is a step towards designing robots with the versatility and robustness of vertebrates, which can adapt quickly to everyday environments.
Artificial intelligence (AI) has recently re-emerged from the intersection of many fields, directing its collective energy at the building and studying of intelligent machines.
Affordances are ways in which an animal or a robot can interact with the environment. The concept, borrowed from psychology, inspires a fresh take on the design of robots that will be able to hold their own in everyday tasks and unpredictable situations.
David Oh was lead flight director for the Curiosity Mars rover and is now part of NASA’s mission to Psyche, a 200-km-wide metal asteroid. Our editor Yann Sweeney met with David at SIGGRAPH Asia to discuss whether advances in AI could improve autonomous robots for space exploration.
Research on reinforcement learning in artificial agents focuses on a single complex problem within a static environment. In biological agents, research focuses on simple learning problems embedded in flexible, dynamic environments. The authors review the literature on these topics and suggest areas of synergy between them.
To perform complex tasks, robots need to learn the relationship between their bodies and dynamic environments. A biologically plausible approach to hardware and software design shows that a robotic tendon-driven limb can make effective movements based on a short period of learning.
Present day quantum technologies enable computations with tens and soon hundreds of qubits. A major outstanding challenge is to measure and benchmark the complete quantum state, a task that grows exponentially with the system size. Generative models based on restricted Boltzmann machines and recurrent neural networks can be employed to solve this quantum tomography problem in a scalable manner.
Juxi Leitner recounts how he and his team took part in — and won — the 2017 Amazon Robotics Challenge and reflects on the importance of solving big picture problems in robotics.