DONGHO KANG

Logo

Robotics student living in Zurich.

News

Projects

Writings

24 February 2018

RaiSim: Efficient Pipeline for Rigid Body Simulation with Contacts

by Dongho Kang

title-image

RaiSim is a rigid body simulator developed by researchers at RSL, ETH Zurich (Jemin Hwangbo and Dongho Kang).

The goal of RaiSim project is to provide simulation for data-driven robotics and animation research.

raisim-zerogravity-gif
raisim-anymals-gif

Abstract

Solving contact dynamics is a NP-hard problem due to its non-convexity and discontinuity. In order to make the problem tractable, rigid-body simulators often over-relax the contact problem that might leads to inaccurate result. In this project, the multi-contact simulation pipeline raiSim was implemented based on Hwangbo’s method that results in accurate solution in an efficient manner. By using Hwangbo’s method, the penetration problem is significantly reduced. Moreover, raiSim obtains contact solution using sparse formulation of contact problem thus fast and memory-efficient. For the integration, raiSim computes gyroscopic force by Implicit Euler scheme and integrate the object velocity by Semi-implicit Euler that improves numerical stability of the rigid-body simulation. In the benchmark tests, our simulation pipeline performed similarly or better than the state-of-the-art simulators in terms of speed and better in terms of accuracy. Remarkably, it can simulate 18-DOF quadruped robot at 29k time step per second in a 12 contact points case. This performance could be achievable by the virtue of accurate and fast solver, sparse representation of the contact problem, efficient memory management and numerically stable integration scheme.

Dongho Kang, "RaiSim: Efficient Pipeline for Rigid Body Simulation with Contacts," MSc Semester Thesis, Departement Maschinenbau und Verfahrenstechnik, ETH Zürich, Jan. 2018


Demo videos



tags: RaiSim - ETH