English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

When limited by their own morphologies, humans and some species of animalshave the remarkable ability to use objects from the environment towardaccomplishing otherwise impossible tasks. Robots might similarly unlock a rangeof additional capabilities through tool use. Recent techniques for jointlyoptimizing morphology and control via deep learning are effective at designinglocomotion agents. But while outputting a single morphology makes sense forlocomotion, manipulation involves a variety of strategies depending on the taskgoals at hand. A manipulation agent must be capable of rapidly prototypingspecialized tools for different goals. Therefore, we propose learning adesigner policy, rather than a single design. A designer policy is conditionedon task information and outputs a tool design that helps solve the task. Adesign-conditioned controller policy can then perform manipulation using thesetools. In this work, we take a step towards this goal by introducing areinforcement learning framework for jointly learning these policies. Throughsimulated manipulation tasks, we show that this framework is more sampleefficient than prior methods in multi-goal or multi-variant settings, canperform zero-shot interpolation or fine-tuning to tackle previously unseengoals, and allows tradeoffs between the complexity of design and controlpolicies under practical constraints. Finally, we deploy our learned policiesonto a real robot. Please see our supplementary video and website athttps://robotic-tool-design.github.io/ for visualizations.

Learning to Design and Use Tools for Robotic Manipulation