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

We present a simple and intuitive approach for interactive control ofphysically simulated characters. Our work builds upon generative adversarialnetworks (GAN) and reinforcement learning, and introduces an imitation learningframework where an ensemble of classifiers and an imitation policy are trainedin tandem given pre-processed reference clips. The classifiers are trained todiscriminate the reference motion from the motion generated by the imitationpolicy, while the policy is rewarded for fooling the discriminators. Using ourGAN-based approach, multiple motor control policies can be trained separatelyto imitate different behaviors. In runtime, our system can respond to externalcontrol signal provided by the user and interactively switch between differentpolicies. Compared to existing methods, our proposed approach has the followingattractive properties: 1) achieves state-of-the-art imitation performancewithout manually designing and fine tuning a reward function; 2) directlycontrols the character without having to track any target reference poseexplicitly or implicitly through a phase state; and 3) supports interactivepolicy switching without requiring any motion generation or motion matchingmechanism. We highlight the applicability of our approach in a range ofimitation and interactive control tasks, while also demonstrating its abilityto withstand external perturbations as well as to recover balance. Overall, ourapproach generates high-fidelity motion, has low runtime cost, and can beeasily integrated into interactive applications and games.

A GAN-Like Approach for Physics-Based Imitation Learning and Interactive Character Control