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

Though robot learning is often formulated in terms of discrete-time Markovdecision processes (MDPs), physical robots require near-continuous multiscalefeedback control. Machines operate on multiple asynchronous sensing modalities,each with different frequencies, e.g., video frames at 30Hz, proprioceptivestate at 100Hz, force-torque data at 500Hz, etc. While the classic approach isto batch observations into fixed-time windows then pass them throughfeed-forward encoders (e.g., with deep networks), we show that there exists amore elegant approach -- one that treats policy learning as modeling latentstate dynamics in continuous-time. Specifically, we present 'InFuser', aunified architecture that trains continuous time-policies with NeuralControlled Differential Equations (CDEs). InFuser evolves a single latent staterepresentation over time by (In)tegrating and (Fus)ing multi-sensoryobservations (arriving at different frequencies), and inferring actions incontinuous-time. This enables policies that can react to multi-frequency multisensory feedback for truly end-to-end visuomotor control, without discrete-timeassumptions. Behavior cloning experiments demonstrate that InFuser learnsrobust policies for dynamic tasks (e.g., swinging a ball into a cup) notablyoutperforming several baselines in settings where observations from one sensingmodality can arrive at much sparser intervals than others.

Multiscale Sensor Fusion and Continuous Control with Neural CDEs