Scenario-Based Model Predictive Control for Energy Harvesting Actuators

Nicholas J. Watkins, Konstantinos Gatsis, Manfred Morari, George J. Pappas

Abstract: In this paper, we develop a method of control for energy harvesting devices-i.e. systems which can restore their energy reserves online-fit for the purpose of using actuators for tracking a stochastic trajectory. This is a problem of significant novelty, insofar that most prior work concerning energy harvesting devices focuses on energy harvesting communication nodes, and not on devices capable of actuation. Moreover, it is a problem with significant potential applications, as endowing devices with a means for determining how and when energy should be used to best accomplish an assigned task is a central engineering problem faced in developing automation on a large scale. We take a scenario-based model predictive control approach to solving this problem, in which stochastic models of the energy arrival and target evolution processes are used in order to generate a random optimization problem, the solution of which generates a sequence of controls to be applied by the device's actuators. We show that the optimization required by this approach admits a small convex formulation, which may be solved efficiently. We examine the efficacy of the designed controller through numerical simulations.