Receding Horizon Control

Abstract¶

In this article we have shown that receding horizon control offers a straightforward method for designing feedback controllers that deliver good performance while respecting complex constraints. A designer specifies the RHC controller by specifying the objective, constraints, prediction method, and horizon, each of which has a natural choice suggested directly by the application. In more traditional approaches, such as PID control, a designer tunes the controller coefficients, often using trial and error, to handle the objectives and constraints indirectly. In contrast, RHC controllers can often obtain good performance with little tuning. In addition to the straightforward design process, we have seen that RHC controllers can be implemented in real time at kilohertz sampling rates. These speeds are useful for both real-time implementation of the controller as well as rapid Monte Carlo simulation for design and testing purposes. Thus, receding horizon control can no longer be considered a slow, computationally intensive policy. Indeed, RHC can be applied to a wide range of control problems, including applications involving fast dynamics.

Summary¶

Provides a tutorial-style case for receding horizon control as an intuitive way to design constrained feedback controllers. The article emphasizes that, with custom online optimization and code generation, RHC can be implemented at kilohertz rates, challenging the older view that MPC is only for slow processes.

Links¶

Tags¶

Similar Papers¶