Motion Synergies for Real-Time Postural Control in Virtual Environments

Virtual prototyping has become an important part during the product development process. The common way to let a user interact with the virtual prototype is the use of a three dimensional virtual mannequin that is controlled by the user. These human figures are difficult to animate such that they move in a coordinated and human-like fashion. Motion capture is a very popular technology to bring natural motions into human figure animation. Many techniques have been proposed to reuse and to edit the motion captured data to produce human-like motions. However, it can be difficult to reuse the captured motions in new situations (e.g. retargeting to a new environment). For more realistic animations of virtual mannequins, we introduce several techniques. We propose a simple model for the coupling behavior of the human spine that is capable of exhibiting anatomically correct motions of the vertebrae in virtual mannequins. The adjustments of the joints due to the coupling is made with several simple (in)equality constraints that can be transparently integrated into a Prioritized Inverse Kinematics framework. A key benefit is to prevent the Prioritized Inverse Kinematics algorithm from providing infeasible postures. To better model the kinematic anisotropy of joint limits for virtual mannequin, we introduce the progressive clamping method that damps only the joints' variation component heading towards the limit. We present how our approach is exploited for the major classes of rotational joints. The Prioritized Inverse Kinematics algorithm is vulnerable to the full extension singularity of the limbs. In such contexts the convergence is reduced and/or less believable intermediate solutions are produced. We address this issue by developing a new type of analytic constraint that smoothly integrates within the Prioritized Inverse Kinematics framework. In this thesis, we also propose a hybrid control approach taking advantage of data-driven and goal-driven methods while overcoming their limitations. In particular, we take advantage of the latent space characterizing a given motion database. We are focusing on highly coordinated movements (e.g. reach) that remain very similar independently of the goal location that are the expression of a high level motion synergy. Thus, the captured motion can be expressed in a reduced latent space. We introduce a motion constraint, 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 Kinematics framework to channel the postural control through a spatio-temporal pattern representative of the motion database while achieving a broader range of goals. Comparative results illustrate the interest of joint coupling, progressive clamping and the flexion-extension constraint. We illustrate our hybrid control approach with a database of large range forward-stepping reach motions. Finally, we present full-body interactions in a CAVE-like environment in which a reduced set of optical markers is used to track the full body.