← ArchivePaper2007

SIMBICON: Simple Biped Locomotion Control

Kangkang Yin, Kevin Loken, Michiel van de Panne

SIGGRAPHAcademic590 cites37 descendantsMotion Synthesis

Simple feedback-based bipedal locomotion controller achieving robust walking and running on varied terrain with minimal parameter tuning.

Open the paper ↗Trace in the lineage graph

Abstract

SIMBICON is a simple control strategy for physics-based biped locomotion that can generate a wide variety of gaits and styles in real-time, including walking in all directions, running, skipping, and hopping. The framework combines a finite state machine of target poses driven by proportional-derivative controllers with a balance feedback law that adjusts the swing hip target angle based on the center of mass position and velocity. Controllers can be authored manually with a small set of parameters or reconstructed from motion capture data, and they remain robust to pushes, unexpected terrain variations, and changes in kinematic and dynamic parameters. The authors also apply feedback error learning to learn predictive feedforward torques, enabling low-gain control that produces smoother and more natural simulated motion.

How to read this

Category: Method: physics-based biped locomotion control
Contributions: A simple control strategy generating many real-time gaits and styles (walking in all directions, running, skipping, hopping).
A finite-state machine of target poses driven by PD controllers, combined with a balance feedback law adjusting the swing-hip target from center-of-mass position and velocity.
Controllers that are either hand-authored from few parameters or reconstructed from motion capture, with feedback error learning of feedforward torques for smoother low-gain control.
Context: Continues the physics-based human locomotion tradition (Hodgins et al.'s Animating Human Athletics), emphasizing a minimal, robust balance-feedback controller.Builds on: Animating Human Athletics
Correctness: Demonstrated to remain robust to pushes, terrain variation, and parameter changes; robustness is shown empirically across gaits rather than proven, so behavior under conditions far outside the demonstrated range should not be assumed.
Clarity: Notably accessible for a control paper; a first pass conveys the FSM-plus-balance-feedback core, a second pass for the parameterization and feedback-error-learning details.
How to read it: First pass for the FSM and the swing-hip balance law (the memorable core idea); second pass on parameter authoring and feedback error learning if you intend to build or extend the controller.

Builds on

Animating Human Athletics 1995

Built upon by

Terrain-Adaptive Locomotion Skills Using Deep Reinforcement Learning 2016

Related work

Animating Human Athletics 1995 / SIGGRAPH
Motion Grammars for Character Animation 2016 / CGF
Physics-based Motion Capture Imitation with Deep Reinforcement Learning 2018 / MIG
PFPN: Continuous Control of Physically Simulated Characters using Particle Filtering Policy Network 2021 / MIG

Keywords

simbicon biped locomotion control finite state machine proportional-derivative control balance feedback swing foot placement center of mass feedback feedback error learning feedforward torque pose control graph forward dynamics simulation motion capture imitation push recovery walking gait running skipping zero moment point limit cycle

This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →