A human-like learning control for digital human models in a physics-based virtual environment

This paper presents a new learning control framework for virtual humans in a physics-based virtual environment. This framework combines multi-objective control with learning technique. Multi-objective control is based on humanlike properties (combined feedforward and feedback controller) and learning properties (human-beings ability to learn novel task dynamics through the minimization of instability, error and effort). This controller performs multiple tasks simultaneously (balance, contacts, manipulation) in real time and adapts feedforward force as well as impedance to counter environment disturbance. It is very useful to deal with unstable manipulations as tool-use tasks and to compensate perturbations. The novelty of our controller is that it is implemented in cartesian space with joint stiffness, damping and torque learning in a multi-objective control framework. The interest of the proposed control method to model human motor adaptation has been demonstrated by various simulations.