Robust Adaptive MPC Under Nonlinear Time-Varying Uncertainties: An Uncertainty Compensation Approach
New MPC method handles nonlinear uncertainties for safer autonomous flight and spacecraft landing.
A team of researchers including Ran Tao, Pan Zhao, Ilya Kolmanovsky, and Naira Hovakimyan has introduced a novel robust adaptive control framework designed to handle complex, real-world uncertainties. Their paper, "Robust Adaptive MPC Under Nonlinear Time-Varying Uncertainties: An Uncertainty Compensation Approach," presents the Uncertainty Compensation-based MPC (UC-MPC) method. This framework is specifically engineered for linear systems plagued by nonlinear, time-varying uncertainties—a common and challenging scenario in advanced robotics, aerospace, and autonomous systems. The core innovation is a dual-controller architecture that separates and tackles different types of uncertainty to maintain both performance and rigorous safety guarantees.
The UC-MPC framework's technical approach is twofold. First, it employs an L1 adaptive controller to actively compensate for 'matched' uncertainties that align with the system's control channels. Second, it uses a robust feedback controller, designed via linear matrix inequalities (LMIs), to mitigate the effect of 'unmatched' uncertainties on critical output channels. By mathematically deriving uniform bounds on the errors between the real, uncertain system and a nominal, uncertainty-free model, the researchers can preemptively 'tighten' the state and input constraints for the controller. This allows a Model Predictive Controller (MPC) to be designed for the simpler nominal system while still guaranteeing that the actual, complex system will satisfy all safety constraints. Simulation results demonstrating the framework's effectiveness for an aircraft flight control problem and a high-stakes spacecraft landing on an asteroid underscore its potential to enable more reliable and high-performing autonomy in safety-critical applications.
- Integrates L1 adaptive control with robust MPC to handle both matched and unmatched nonlinear uncertainties.
- Uses linear matrix inequalities to design a feedback controller that guarantees constraint satisfaction under tightened bounds.
- Validated in high-fidelity simulations for autonomous flight control and precision asteroid landing scenarios.
Why It Matters
Enables safer, more reliable autonomous systems for aerospace, robotics, and other fields where uncertainty is unavoidable.