postPerspective

Motion Control

Tuesday, 26 July, 3:45 pm - 5:15 pm, Anaheim Convention Center, Ballroom E
Session Chair: Jernej Barbič, University of Southern California

Unified Motion Planner for Fishes With Various Swimming Styles

A unified motion planner that generates variations in fish swimming styles. By modeling the common decision-making mechanism in fish that allows them to instantly decide “where and how to swim”, the paper demonstrates 12 types of fishes and massive fish school with completely different sizes and skeletal structures.

Daiki Satoi, University of Tsukuba

Mikihiro Hagiwara
University of Tsukuba

Akira Uemoto
University of Tsukuba

Hisanao Nakadai
University of Tsukuba

Junichi Hoshino
University of Tsukuba

Guided Learning of Control Graphs for Physics-Based Characters

A method for learning robust control graphs that supports real-time physics-based simulation of multiple characters, each capable of a diverse range of movement skills.

Libin Liu
Disney Research Pittsburgh

Michiel van de Panne
The University of British Columbia

KangKang Yin
National University of Singapore

Terrain-Adaptive Locomotion Skills Using Deep Reinforcement Learning

Deep reinforcement learning is used to develop controllers giving simulated characters the capability to dynamically navigate across challenging terrains, directly using a height map of the upcoming terrain as input. The method uses a mixture of actor-critic experts, which results in specialized subcontrollers.

Xue Bin Peng
The University of British Columbia

Glen Berseth
The University of British Columbia

Michiel van de Panne
The University of British Columbia

Task-Based Locomotion

This locomotion model that supports a rich variety of task-specific stepping behaviors, including side-steps, toe pivots, heel pivots, intentional foot slides, and effort-adapted positioning of the body.

Shailen Agrawal
The University of British Columbia

Michiel van de Panne
The University of British Columbia