Tuning Controls for Interaction with Unstructured Environments based on Human Personality Types

Traditional robust control system design methodologies typically provide methods for dealing with well-understood, easily-modelled disturbances of similar magnitude. However, many proposed applications for robotics require interaction with an unstructured environment, which may even include people and other robots. Providing a more advanced toolkit than simple feedback laws, fuzzy logic offers advantages in such situations; however designers often use trial-and-error (and eventually experience) in constructing their fuzzy membership sets. Moreover, many interactive tasks easily accomplished by human beings remain beyond the ability of current robotics technology. We propose a more formal methodology at the high-level design phase, based on understandings of personality types. Specifically, similarities between human personalities and robot controls allow one to choose a personality for the control as the first stage of the design process. This might provide a powerful method of tuning when one only has intuitive understanding (and order-of-magnitude estimates) of the robot, its environment, people, and other robots. Our method allows combining this type of knowledge with intuitive understanding of human personalities when tackling a control design problem. To illustrate the methodology, a design that combines four distinct personalities in a fuzzy-PID control allows a simulated system to react more reasonably, and quite differently, to unmodeled disturbances that differ by an order of magnitude.