Kinematic modeling of the human operator

The simulation of positions and motions of human operators interacting with machines and other operators in both industrial and non-industrial environments is essential for an ergonomic design of work places in production lines or car interiors. Therefore, modeling of the operator and his highly redundant link mechanisms is essential for such a simulation. With a predefined analytical biomechanical model of the operator it is feasible to compute the position of any end-effector for a given link configuration. Much more difficult is to find the solution of the inverse task: Given a point in world coordinates, determine the corresponding joint angles. A corresponding learning procedure is based on local linear mapping between link coordinates and effector coordinates where the global mapping remains highly nonlinear. The training data of the human motions were recorded from different subjects equipped with markers on proper anatomical landmarks at the body using a system of four IR cameras for motion analysis. Learning of the inverse kinematics is done by fuzzy clustering and local linear (affine) fuzzy models being smoothly connected. Moreover a simple operator dynamics and well defined restrictions on joint angles and contributing joint torques are included in the model.