Nonlinear model predictive control with a robot in the loop

The interplay between heterogeneous elements in control systems, such as processes and robots, often leads to hybrid optimization problems. This work integrates the path planning of sensing agents into the process control problem while circumventing this complexity. To this end, the proposed Model Predictive Control (MPC) strategy utilizes continuous variables to model the robots’ movements and optimizes both system performance and path planning over a prediction horizon. As a result, robots move and sense in ways that maximize control performance. Moreover, the stochastic nonlinear formulation of the MPC controller allows it to dynamically adjust to constraint violations while maintaining probabilistic guarantees. To illustrate the proposed method, an academic example is employed in which a single robot monitors two separated tanks. Our simulations show that the proposed strategy enhances the flexibility of control systems with agents in the loop, providing a viable and efficient solution for applications ranging from industrial automation to resource management in uncertain environments.