Research Analyzer
← Back ICRA 2026

Guaranteed Robust Nonlinear MPC Via Disturbance Feedback

Antoine Leeman, Johannes Köhler, Melanie N. Zeilinger

PDF

AI summary

Key figure (auto-extracted from paper)
A scalable robust MPC framework guarantees safety, stability, and real-time feasibility for nonlinear systems by jointly optimizing trajectories and disturbance-feedback controllers.
Robust MPC Nonlinear Control Disturbance Feedback Sequential Convex Programming Real-Time Control Safety Guarantees

Problem

Existing model-based control methods struggle to guarantee safety and stability under disturbances without excessive conservatism or computational cost, while data-driven approaches lack formal guarantees. This paper addresses how to achieve robust constraint satisfaction and stability for nonlinear systems in real time.

Approach

The method decomposes the uncertain nonlinear system into a nominal model, disturbance-feedback controllers, and error bounds, then jointly optimizes them using sequential convex programming to solve efficient convex subproblems.

Key results

  • Formal guarantees of robust constraint satisfaction, recursive feasibility, and input-to-state stability
  • Efficient sequential convex programming algorithm enabling real-time deployment
  • Demonstrated robust performance and real-time feasibility on rocket landing and quadcopter dynamics
  • Open-source implementation for scalable application to high-dimensional systems

Why it matters

Enables safe, real-time deployment of autonomous robots and spacecraft in uncertain environments by bridging theoretical robustness guarantees with computational tractability.

Abstract

Robots must satisfy safety-critical state and input constraints despite disturbances and model mismatch. We in- troduce a robust model predictive control (RMPC) formulation that is scalable and compatible with real-time implementa- tion. Our formulation guarantees robust constraint satisfac- tion, input-to-state stability (ISS) and recursive feasibility. The key idea is to decompose the uncertain nonlinear system into (i) a nominal nonlinear dynamic model, (ii) disturbance- feedback controllers, and (iii) bounds on the model error. These components are optimized jointly using sequential convex programming. The resulting convex subproblems are solved efficiently using a recent disturbance-feedback MPC solver. The approach is validated across multiple dynamics, including a rocket-landing problem with steerable thrust. An open- source implementation is available at https://github. com/antoineleeman/robust-nonlinear-mpc.

Index terms

Optimization and Optimal Control Robust/Adaptive Control

Related papers