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

Deformable Object Manipulation (DOM) is of significant importance to bothdaily and industrial applications. Recent successes in differentiable physicssimulators allow learning algorithms to train a policy with analytic gradientsthrough environment dynamics, which significantly facilitates the developmentof DOM algorithms. However, existing DOM benchmarks are eithersingle-object-based or non-differentiable. This leaves the questions of 1) howa task-specific algorithm performs on other tasks and 2) how adifferentiable-physics-based algorithm compares with the non-differentiableones in general. In this work, we present DaXBench, a differentiable DOMbenchmark with a wide object and task coverage. DaXBench includes 9 challenginghigh-fidelity simulated tasks, covering rope, cloth, and liquid manipulationwith various difficulty levels. To better understand the performance of generalalgorithms on different DOM tasks, we conduct comprehensive experiments overrepresentative DOM methods, ranging from planning to imitation learning andreinforcement learning. In addition, we provide careful empirical studies ofexisting decision-making algorithms based on differentiable physics, anddiscuss their limitations, as well as potential future directions.

DaXBench: Benchmarking Deformable Object Manipulation with Differentiable Physics