Nonlinear Robust Control Using an Inside-Out Strategy With User-Friendly Tuning

Inspired by ideas about the brain's control systems as inside-out models, this paper develops an inside-out strategy for a robot arm i.e. generally placing as much importance on controlling the internal model as it does on controlling the real system. In this work, adjusting the control involves varying how much attention it pays to external sensor information versus internal states, and independently changing how much effort it uses to control the external system versus the amount of virtual control directed to the internal model. Using a popular model of human experience, internal and external attention vary as complementary opposites, as do internal and external control effort. The method provides a way to tune a robot in real time in order to adjust for different situations, in a way that even untrained operators may find intuitive i.e. a friendly human-machine interface for adjusting a control system. Results for a single-link robot arm demonstrate the effects of tuning on performance, stability margins, disturbance rejection, and trade-offs for various scenarios including sinusoidal disturbance, elastic-element dynamics, and time delay.