Vis Comput (2009) 25: 509–518 DOI 10.1007/s00371-009-0336-2 ORIGINAL ARTICLE Motion constraint Daniel Raunhardt · Ronan Boulic Published online: 3 March 2009 © Springer-Verlag 2009 Abstract In this paper, we propose a hybrid postural con- trol approach taking advantage of data-driven and goal- oriented methods while overcoming their limitations. In par- ticular, we take advantage of the latent space characteriz- ing a given motion database. We introduce a motion con- straint operating in the latent space to benefit from its much smaller dimension compared to the joint space. This allows its transparent integration into a Prioritized Inverse Kine- matics framework. If its priority is high the constraint may restrict the solution to lie within the motion database space. We are more interested in the alternate case of an intermedi- ate priority level that channels the postural control through a spatiotemporal pattern representative of the motion data- base while achieving a broader range of goals. We illustrate this concept with a sparse database of large range full-body reach motions. Keywords Inverse kinematics · Motion editing · Posture control 1 Introduction The applications of human-like character animation at in- teractive frame rates are manifold, ranging from computer This work has been supported by the Swiss National Foundation under the grant N° 200020-117706. D. Raunhardt ( ) · R. Boulic Ecole Polytechnique Fédérale de Lousanne, VRLAB Station 14, 1015 Lausanne, Switzerland e-mail: daniel.raunhardt@epfl.ch R. Boulic e-mail: ronan.boulic@epfl.ch entertainment to ergonomic studies or virtual prototyping. Human figures are difficult to animate such that they move in a coordinated and human-like fashion [7, 17]. We focus in the present paper on the problem of controlling a full- body goal directed motion (i.e. reach) where locomotion is not necessary. Previous work has led to two different approaches for realistic character animation: data-driven and goal-oriented. Data-driven approaches create animations from motion cap- tured data that contain the movement details from a human actor. The embedded natural flow of motion is their most at- tractive feature; on the other hand it can become tedious to generate movements that are outside the space of the cap- tured movements. Goal-oriented techniques minimize the norm of postural variations and/or other physically-based cost functions while achieving a wide range of user-defined tasks. Their key advantage is their versatility at the cost of a frequent lack of naturalness when it comes to reproduce human activities. In this paper we propose a hybrid approach aiming to combine the positive features of goal-oriented and data- driven approaches. Our goal is to benefit from the nat- ural flow of movement provided by a motion database to channel the convergence of a Prioritized Inverse Kinemat- ics (PIK) solver. This makes sense as the dimension of the user-defined goals are generally much smaller than the di- mension of the joint space. Hence, even when exploited at a low priority, a motion database can influence each succes- sive postural variation of the optimization convergence. As already mentioned we use the action of full-body reach to illustrate the methodology throughout the paper; its generalization to other movement types is discussed at the end of the paper. We use a small number of reach mo- tions where both feet remain fixed on the ground while the brought to you by CORE View metadata, citation and similar papers at core.ac.uk provided by RERO DOC Digital Library