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

Humans are able to perform a myriad of sophisticated tasks by drawing uponskills acquired through prior experience. For autonomous agents to have thiscapability, they must be able to extract reusable skills from past experiencethat can be recombined in new ways for subsequent tasks. Furthermore, whencontrolling complex high-dimensional morphologies, such as humanoid bodies,tasks often require coordination of multiple skills simultaneously. Learningdiscrete primitives for every combination of skills quickly becomesprohibitive. Composable primitives that can be recombined to create a largevariety of behaviors can be more suitable for modeling this combinatorialexplosion. In this work, we propose multiplicative compositional policies(MCP), a method for learning reusable motor skills that can be composed toproduce a range of complex behaviors. Our method factorizes an agent's skillsinto a collection of primitives, where multiple primitives can be activatedsimultaneously via multiplicative composition. This flexibility allows theprimitives to be transferred and recombined to elicit new behaviors asnecessary for novel tasks. We demonstrate that MCP is able to extractcomposable skills for highly complex simulated characters from pre-trainingtasks, such as motion imitation, and then reuse these skills to solvechallenging continuous control tasks, such as dribbling a soccer ball to agoal, and picking up an object and transporting it to a target location.

MCP: Learning Composable Hierarchical Control with Multiplicative Compositional Policies