CARL: Controllable agent with reinforcement learning for quadruped locomotion

Abstract: Motion synthesis in a dynamic environment has been a long-standing problem for character animation. Methods using motion capture data tend to scale poorly in complex environments because of their larger capturing and labeling requirement. Physics-based controllers are effective in this regard, albeit less controllable. In this paper, we present CARL, a quadruped agent that can be controlled with high-level directives and react naturally to dynamic environments. Starting with an agent that can imitate individual animation clips, we use Generative Adversarial Networks to adapt high-level controls, such as speed and heading, to action distributions that correspond to the original animations. Further fine-tuning through the deep reinforcement learning enables the agent to recover from unseen external perturbations while producing smooth transitions. It then becomes straightforward to create autonomous agents in dynamic environments by adding navigation modules over the entire process. We evaluate our approach by measuring the agent’s ability to follow user control and provide a visual analysis of the generated motion to show its effectiveness.

14/06/2020

action and behavior recognition, force prediction, human-object interaction, physical understanding, contact point prediction, motion and tracking, video understanding

4:59

02/02/2021

Tim Seyde, Igor Gilitschenski, Wilko Schwarting and
Bartolomeo Stellato, Martin Riedmiller, Markus Wulfmeier, Daniela Rus

action synthesis, video synthesis, joint generative model, human action generation, end-to-end learning, conditional video generation

3:02

16/11/2020

Tianlu Wang, Xuezhi Wang, Yao Qin and
Ben Packer, Kang Li, Jilin Chen, Alex Beutel, Ed Chi

3d object detection, point cloud, starcture-aware, fast detection, single-stage detector, auxiliary task learning, autonomous driving, convolutional neueral network, deep learning, computer vision

1:01

17/08/2020

Baohe Zhang, Raghu Rajan, Luis Pineda and
Nathan Lambert, André Biedenkapp, Kurtland Chua, Frank Hutter, Roberto Calandra

mot, multi-task learning, motion, affinity, attention, online

1:03

02/02/2021

Train a One-Million-Way Instance Classifier for Unsupervised Visual Representation Learning

Yu Liu, Lianghua Huang, Pan Pan and
Bin Wang, Yinghui Xu, Rong Jin

unsupervised domain adaptation, self-training, 3D object detection, scene flow, LiDAR point cloud, source-free domain adaptation

3:06

14/06/2020