Nonlinear Model Predictive Control using a Modified Sequential Convex Programming

This paper presents a nonlinear model predictive control (NMPC) scheme based on sequential convex programming (SCP). We propose a modified SCP algorithm for nonlinear MPC that adaptively incorporates the effects of linearization errors into the state constraints, ensuring that the actual state trajectory—generated by the nonlinear dynamics from the returned solution—faithfully satisfies all constraints, even if the SCP process is terminated early. Subsequently, an NMPC framework for the nominal system is developed, eliminating the need for additional optimization. The approach is further extended to systems with additive disturbances by introducing a shrinking horizon and a self-triggering mechanism, which together reduce the problem dimensionality and computational frequency, thereby enhancing real-time performance. Theoretical analysis establishes recursive feasibility and closed-loop stability, even when the SCP procedure is terminated early. Finally, simulation results are presented to validate the effectiveness of the proposed algorithm.