Nonlinear Attractor Dynamics for Guiding an Anthropomorphic Robot by Means of Speech Commands

We present a unified approach to design robot control systems by means of nonlinear dynamical systems. The robot’s behavior is generated by evaluating differential equations for variables which parametrize the behavior. These differential equations are designed such that potential behavioral goals specify stable fixed points (attractors) of the behavioral variables. The strength of these attractors is controlled by another dynamical system depending on the sensor input and predefined logical rules which express the interrelation between the different behavioral goals. The quantitative variation of the attractors’ strengths induces bifurcations of the behavioral dynamics leading to qualitative changes of the robot’s behavior. We show that our approach is particularly suited to guide an anthropomorphic robot by means of natural and intuitive speech commands. A number of elementary behaviors are implemented as contributions to a control dynamics which acts on the mobile platform and on the manipulator arm of the robot. The contributions are activated and deactivated depending on the commands given by the human operator.