Reduced precision solution criteria for solving optimal control problems in NMPC

Nonlinear model predictive control (NMPC) needs to solve optimal control problems (OCPs) instantaneously so that new input can be applied in time. Computational delay, which may destroy the stability of the closed-loop, usually arises in the solution procedure of non-convex and nonlinear OCP with differential equations as its constraints. The traditional termination criteria for optimization algorithms usually cause the solution procedure to converge slowly and waste computing resources. In this paper, new termination rules called Reduced Precision Solution (RPS) criteria are proposed. The RPS criteria take the practical solution precision into account and lead to good approximate solutions of OCPs. Thus, the computation time is saved, and the negative effect of computational delay is reduced. A classical nonlinear model formulated by ordinary differential equations is simulated to demonstrate the efficiency of the RPS criteria.