Nowadays, digital human models are broadly used in many research domains and industries. A digital human can represent human body characteristics and simulate human behaviors and capacities, thus we can evaluate or predict human performances in a scenario via simulations with digital human models. Currently, ergonomic evaluation of human motions depends largely on kinematic human models and Motion Capture techniques. Dynamic human models and associated control techniques have advantages over kinematic ones in terms of physical feasibility and they may lead to a reduced number of experiments with real human subjects. However, the dynamics-based techniques are very time-consuming when they are applied in complex scenarii, especially when the environment is cluttered. In this context, this thesis aims to explore an efficient dynamics-based method for digital human motion simulation in cluttered environments. The method is aimed to be applied in ergonomic study of daily human activities, such as car-ingress or car-egress motions. At the beginning of this study, we carry out MoCap experiments in which we obtain a set of recorded human motions in some cluttered environments. We learn human motion characteristics and principles in cluttered environments by observing and analyzing the recorded motions. With the help of the obtained heuristic knowledge, we make reasonable hypotheses and choose the main problematics of this research. Then we work on the topic of balance. A simplified digital human model for balance formulation is proposed in which the human is taken as a mass point located at its center of mass (CoM) interacting with the environment via a few supports (contact or grasps). Support feasibility (stability) brings constrains on exterior wrenches applied at supports. The constraints then result in an admissible space for the pseudo-wrench (a 6D vector calculated from position and acceleration of CoM) of the model. The admissible pseudo-wrench space defines a criterion for verifying balance of a digital human model during its motions. The “stability margin” concept associated with this criterion is studied which evaluates robustness (or quality) of a balance state. Thereafter, a hierarchical framework is developed which can realize motion simulations via three levels. Firstly, a global CoM trajectory is generated at a global level. The CoM trajectory is generated by optimization under motion state imposition, balance, geometric and timing constraints. Time durations for transition phases are generated as well. Then at a local level, the trajectories of end-effectors (e.g. foot or hand) and whole-body postures are piecewisely generated under the constraints of inverse kinematics and collision-freeness. The results obtained at the previous two levels are finally used as control references in dynamic simulation. A dynamic controller actuates the dynamic digital human model to realize the generated motion. Approaches at each level of the framework are applied and tested in several scenarii. Eventually, the framework is applied in a car-ingress scenario relying on heuristics of car-ingress motions.
| Identifer | oai:union.ndltd.org:CCSD/oai:tel.archives-ouvertes.fr:tel-00991144 |
| Date | 05 December 2012 |
| Creators | Qiu, Zhaopeng |
| Source Sets | CCSD theses-EN-ligne, France |
| Language | English |
| Detected Language | English |
| Type | PhD thesis |
Page generated in 0.1366 seconds