Using rewards to change a person´s behavior: A double-integrator output-feedback dynamic control approach

Using a nonlinear discrete-time model that captures how rewards affect a person\´s attitudes and behavior, we consider in this paper how to find a reward sequence that drives behavior to (or beyond) some desired level. We develop an output-feedback controller by re-arranging the plant equations so that the plant and controller form a linear feedback loop with a "disturbance" that captures all nonlinear effects. We then argue that the "disturbance" signal is rate limited, and propose a controller that contains two discrete-time integrators, a zero, and a saturator. Some simulation and analysis results are included to show that the control scheme is effective.