LAUSR

Abstract This paper investigates a Cyber-Physical & Human System (CPHS) comprised of a deterministic dynamical system plant model and a human actuator model. Namely, human decisions are stochastic inputs to the plant model. We examine a framework where human decisions cannot be directly controlled, but can be influenced via incentive control signals. Specifically, we use the framework of discrete choice models (DCMs) to capture human decision making, and then design optimal controllers for these human actuated dynamical systems. Existing literature on CPHS often treats human inputs as stochastic and exogenous inputs, and then formulates a disturbance rejection problem. Instead of treating human decision-making as an uncontrollable exogenous input, we directly incorporate human decision making into the modeling framework with DCMs. This paper thus adds two original contributions. (i) We develop a generalized human-actuated system framework based on DCM that predicts the probability of human decisions, conditioned on controllable incentives. (ii) We show that existing optimization schemes, such as Sequential Quadratic Programming (SQP) and Dynamic Programming (DP), can be applied to control the proposed human-actuated system. We conclude this paper by demonstrating the framework on a reference tracking problem and an inventory control problem.