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
Robots that can work together safely with humans in unstructured environments need to be able to effectively manipulate objects for a variety of tasks. Much attention has been focused on the challenge of grasping different objects with robotic hands. However, Ortenzi et al. argue that grasping depends on the task. They develop a new metric that measures the success of robot manipulation focused on the goal of the task itself, for instance handing over an object, rather that the action of grasping.
As machine learning methods are adopted across the scientific community, strong code sharing and reviewing practices are required. Our policy mandates that code essential to the main results is made available to reviewers, and to readers on publication. Our partnership with Code Ocean helps authors and reviewers navigate this process.
DeepMind’s AlphaFold recently demonstrated the potential of deep learning for protein structure prediction. DeepFragLib, a new protein-specific fragment library built using deep neural networks, may have advanced the field to the next stage.
To prepare robots for working autonomously under real-world conditions, their resilience and capability to recover from damage needs to improve radically. A fresh take on robot design suggests that instead of adapting the robotic control strategy, we could enable robots to change their physical bodies to recover more effectively from damage.
Traditional robotic grasping focuses on manipulating an object, often without considering the goal or task involved in the movement. The authors propose a new metric for success in manipulation that is based on the task itself.
An approach to protein structure prediction is to assemble candidate structures from template fragments, which are extracted from known protein structures. Wang et al. demonstrate that combining deep neural network architectures with a relatively small but high-resolution fragment dataset can improve the quality of the sample fragment libraries used for protein structure prediction.
For some combinatorial puzzles, solutions can be verified to be optimal, for others, the state space is too large to be certain that a solution is optimal. A new deep learning based search heuristic performs well on the iconic Rubik’s cube and can also generalize to puzzles in which optimal solvers are intractable.
When neural networks are retrained to solve more than one problem, they tend to forget what they have learned earlier. Here, the authors propose orthogonal weights modification, a method to avoid this so-called catastrophic forgetting problem. Capitalizing on such an ability, a new module is introduced to enable the network to continually learn context-dependent processing.
Deep neural networks can contain arbitrary mathematical operators, as long as they are derivable. The authors investigate how knowledge about a problem can be incorporated into machine learning through the use of operators that are related to the problem.
As nations come together in Tokyo next summer to celebrate the spirit of human potential in the 2020 Olympic Games, they will have a chance to take part in another international competition hosted by Japan soon after, this time with challenges designed for robot contenders.