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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Legged robots are physically capable of traversing a wide range ofchallenging environments, but designing controllers that are sufficientlyrobust to handle this diversity has been a long-standing challenge in robotics.Reinforcement learning presents an appealing approach for automating thecontroller design process and has been able to produce remarkably robustcontrollers when trained in a suitable range of environments. However, it isdifficult to predict all likely conditions the robot will encounter duringdeployment and enumerate them at training-time. What if instead of trainingcontrollers that are robust enough to handle any eventuality, we enable therobot to continually learn in any setting it finds itself in? This kind ofreal-world reinforcement learning poses a number of challenges, includingefficiency, safety, and autonomy. To address these challenges, we propose apractical robot reinforcement learning system for fine-tuning locomotionpolicies in the real world. We demonstrate that a modest amount of real-worldtraining can substantially improve performance during deployment, and thisenables a real A1 quadrupedal robot to autonomously fine-tune multiplelocomotion skills in a range of environments, including an outdoor lawn and avariety of indoor terrains.

Legged Robots that Keep on Learning: Fine-Tuning Locomotion Policies in the Real World