This thesis was carried out in collaboration and co-funding of LSI of CEA/LIST and LBMC of IFSTTAR. The aim of the thesis was to study and develop a method for simulating the movement of a virtual manikin (VM) in a cluttered environment based on a priori knowledge. This thesis presents firstly motion capture (MoCap) experiments. The recorded data were analyzed to define some principles on human motion in cluttered environments. We then propose a general balance criterion and stability margin, based on a simplified model of VM. Then, we present a hierarchical framework that can generate and simulate dynamic movements of VM in a cluttered environment in three stages: a global trajectory of the center of mass (CoM) is generated at the global level to ensure balance in the VM's motion; then the trajectories of end-effectors (EE, ie feet, hands) and postures are generated locally under constraints of kinematics and collision avoidance; finally at the execution level, trajectories (CoM and EEs) and postures are used as references in a dynamic controller associated with VM so that the VM realizes the motion in a simulation. This framework is implemented in a car-ingress scenario in order to evaluate its performance and to suggest future improvements
