Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Efficient robust model predictive control using ellipsoidal sets
Robust model predictive control (MPC) extends the guarantees provided by nominal MPC to systems subject to uncertainties, thereby ensuring safety.
Recent advances in robust MPC propose ways to parameterize uncertain state and input trajectories using ellipsoidal sets. In this project, novel strategies will be investigated to improve the computational complexity of the MPC optimization problems.
Keywords: Robust model predictive control, Convex optimization
Robust model predictive control (MPC) is a control technique which enables to incorporate model mismatches and exogenous unknown disturbances into the control design problem. With the increasing demand for safe-autonomy and safe learning-based methods, robust MPC algorithms have gathered a lot of attention. One way to implement robust MPC is to generate a tube of predicted trajectories, which contains all feasible trajectories of the system. In order to obtain computationally tractable optimization problems, parameterized sets (ellipsoids or polytopes) are used to propagate uncertain trajectories of the system.
Recently, a novel ellipsoidal tube MPC approach has been proposed for linear systems affected by dynamic model uncertainties [1]. The method is scalable and can be applied to large-scale systems, but the MPC optimization is much slower compared to existing MPC methods using polytopes. In this project, novel strategies will be investigated to reduce the computational complexity of the optimization problem, for example by imposing additional properties on the ellipsoidal sets.
The project is conceived as a SemesterArbeit, but can be extended into a MasterArbeit depending upon the interest of the student. Having attended the courses "Model Predictive Control" and "Advanced Model Predictive Control" is a prerequisite for the project.
**Highlights of the project:**
- Understand the different ways to ensure safety using model predictive control
- Design scalable MPC controllers which can be applied to large systems
- Demonstrate the versatility of the ideas by implementing controllers on multiple real-world examples of your choice
Interested students should apply with a CV and a transcript of their courses, with grades.
**Publications**: If the final results are promising they can be turned into a publication.
[1] Parsi, A., Iannelli, A. and Smith, R.S., 2022. Scalable tube model predictive control of uncertain linear systems using ellipsoidal sets. arXiv preprint arXiv:2204.02134.
Robust model predictive control (MPC) is a control technique which enables to incorporate model mismatches and exogenous unknown disturbances into the control design problem. With the increasing demand for safe-autonomy and safe learning-based methods, robust MPC algorithms have gathered a lot of attention. One way to implement robust MPC is to generate a tube of predicted trajectories, which contains all feasible trajectories of the system. In order to obtain computationally tractable optimization problems, parameterized sets (ellipsoids or polytopes) are used to propagate uncertain trajectories of the system.
Recently, a novel ellipsoidal tube MPC approach has been proposed for linear systems affected by dynamic model uncertainties [1]. The method is scalable and can be applied to large-scale systems, but the MPC optimization is much slower compared to existing MPC methods using polytopes. In this project, novel strategies will be investigated to reduce the computational complexity of the optimization problem, for example by imposing additional properties on the ellipsoidal sets.
The project is conceived as a SemesterArbeit, but can be extended into a MasterArbeit depending upon the interest of the student. Having attended the courses "Model Predictive Control" and "Advanced Model Predictive Control" is a prerequisite for the project.
**Highlights of the project:**
- Understand the different ways to ensure safety using model predictive control
- Design scalable MPC controllers which can be applied to large systems
- Demonstrate the versatility of the ideas by implementing controllers on multiple real-world examples of your choice
Interested students should apply with a CV and a transcript of their courses, with grades.
**Publications**: If the final results are promising they can be turned into a publication.
[1] Parsi, A., Iannelli, A. and Smith, R.S., 2022. Scalable tube model predictive control of uncertain linear systems using ellipsoidal sets. arXiv preprint arXiv:2204.02134.
1. Study the literature on robust MPC and familiarize with models described by dynamic uncertainties.
2. Study the design of ellipsoids, linear matrix inequalities and conic constraints
3. Critically understand the technical aspects of designing robust MPC controllers, and the ways in which guarantees of stability and feasibility can be preserved.
4. Formulate an efficient robust MPC scheme by choosing parameterized ellipsoidal sets
5. Perform simulation studies on real-world examples (quadrotor control, building climate control, energy hub management) comparing the proposed appraoch with state-of-the-art techniques.
1. Study the literature on robust MPC and familiarize with models described by dynamic uncertainties. 2. Study the design of ellipsoids, linear matrix inequalities and conic constraints 3. Critically understand the technical aspects of designing robust MPC controllers, and the ways in which guarantees of stability and feasibility can be preserved. 4. Formulate an efficient robust MPC scheme by choosing parameterized ellipsoidal sets 5. Perform simulation studies on real-world examples (quadrotor control, building climate control, energy hub management) comparing the proposed appraoch with state-of-the-art techniques.
Anil Parsi - aparsi@control.ee.ethz.ch
Raffaele Soloperto – soloperr@control.ee.ethz.ch
Prof. Roy S. Smith - rsmith@control.ee.ethz.ch
Anil Parsi - aparsi@control.ee.ethz.ch Raffaele Soloperto – soloperr@control.ee.ethz.ch Prof. Roy S. Smith - rsmith@control.ee.ethz.ch