Stefan Kersting
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Stefan Kersting has a total of 2 co-authors. Here are the top co-authors by number of shared publications:
- Martin Buss
Chair of Automatic Control Engineering (LSR), Technical University of Munich, Munich, Germany
2 shared publications Martin Buss
Chair of Automatic Control Engineering, Department of Electrical and Computer Engineering, Technical University of Munich , Munich, Germany
1 shared publication
Latest Publications
- Open Access
- 20 October 2017
- Published by Taylor & Francis in International Journal of Control
- Vol. 92 (6) , 1264-1281
Recursive estimation in piecewise affine systems using parameter identifiers and concurrent learning
Piecewise affine systems constitute a popular framework for the approximation of non-linear systems and the modelling of hybrid systems. This paper addresses the recursive subsystem estimation in continuous-time piecewise affine systems. Parameter identifiers are extended from continuous-time state-space models to piecewise linear and piecewise affine systems. The convergence rate of the presented identifiers is improved further using concurrent learning, which makes concurrent use of current and recorded measurements. In concurrent learning, assumptions on persistence of excitation are replaced by the less restrictive linear independence of the recorded data. The introduction of memory, however, reduces the tracking ability of concurrent learning because errors in the recorded measurements prevent convergence to the true parameters. In order to overcome this limitation, an algorithm is proposed to detect and remove erroneous measurements at run-time and thereby restore the tracking ability. Detailed examples are included to validate the proposed methods numerically.
- 20 October 2017
- Published by Taylor & Francis in International Journal of Control
- Vol. 92 (6) , 1264-1281
Open AccessRecursive estimation in piecewise affine systems using parameter identifiers and concurrent learning
Piecewise affine systems constitute a popular framework for the approximation of non-linear systems and the modelling of hybrid systems. This paper addresses the recursive subsystem estimation in continuous-time piecewise affine systems. Parameter identifiers are extended from continuous-time state-space models to piecewise linear and piecewise affine systems. The convergence rate of the presented identifiers is improved further using concurrent learning, which makes concurrent use of current and recorded measurements. In concurrent learning, assumptions on persistence of excitation are replaced by the less restrictive linear independence of the recorded data. The introduction of memory, however, reduces the tracking ability of concurrent learning because errors in the recorded measurements prevent convergence to the true parameters. In order to overcome this limitation, an algorithm is proposed to detect and remove erroneous measurements at run-time and thereby restore the tracking ability. Detailed examples are included to validate the proposed methods numerically.
- Open Access
- 9 August 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automatic Control
- Vol. 63 (4) , 1075-1089
How to Systematically Distribute Candidate Models and Robust Controllers in Multiple-Model Adaptive Control: A Coverage Control Approach
Distributing nominal models in multiple-models applications constitutes a long standing problem. The set of models needs to be distributed in such a way that their corresponding controllers can stabilize all possible system configurations in a large uncertainty set. This article presents a systematic solution by phrasing the distribution as coverage control problem, in which each model covers a subset of the uncertainty. The subsets are derived based on a combination of the nu-gap metric, which serves as a distance measure, and the generalized stability margin. Characterizing coverage in terms of the nu-gap also motivates the use of H-inf controller synthesis to design a set of controllers. The proposed algorithms are initialized with suboptimal model configurations. Two update laws optimize the model parameters and minimize the coverage function. The first algorithm performs a gradient descent on the coverage function and the second algorithm performs pairwise optimizations. Due to computational complexity, a discretized implementation is derived, which reduces the optimization to an efficient graph search. The proposed algorithms are evaluated in numeric benchmark examples.
- 9 August 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automatic Control
- Vol. 63 (4) , 1075-1089
Open AccessHow to Systematically Distribute Candidate Models and Robust Controllers in Multiple-Model Adaptive Control: A Coverage Control Approach
Distributing nominal models in multiple-models applications constitutes a long standing problem. The set of models needs to be distributed in such a way that their corresponding controllers can stabilize all possible system configurations in a large uncertainty set. This article presents a systematic solution by phrasing the distribution as coverage control problem, in which each model covers a subset of the uncertainty. The subsets are derived based on a combination of the nu-gap metric, which serves as a distance measure, and the generalized stability margin. Characterizing coverage in terms of the nu-gap also motivates the use of H-inf controller synthesis to design a set of controllers. The proposed algorithms are initialized with suboptimal model configurations. Two update laws optimize the model parameters and minimize the coverage function. The first algorithm performs a gradient descent on the coverage function and the second algorithm performs pairwise optimizations. Due to computational complexity, a discretized implementation is derived, which reduces the optimization to an efficient graph search. The proposed algorithms are evaluated in numeric benchmark examples.
- 1 June 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 738-743
Metrics to compare and control dynamical systems
The focus of this paper is on similarity based controller selection, which is thought of as an essential component of cognitive control. The strength of cognitive systems lies in the ability to compare the current situation with previously experienced situations. This allows for an efficient reuse of previously successful solutions. Transferring this motivation to control theory requires a distance measure between dynamical systems. Given a new dynamical system, it is then possible to find the closest previously controlled system and reuse the corresponding controller. This paper highlights four distance measures and analyzes their suitability for similarity based controller selection. The compared distance measures are Multiple Models Adaptive Control, the Binet-Cauchy kernels, the Martin distance and the ν-gap metric.
- 1 June 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 738-743
Metrics to compare and control dynamical systems
The focus of this paper is on similarity based controller selection, which is thought of as an essential component of cognitive control. The strength of cognitive systems lies in the ability to compare the current situation with previously experienced situations. This allows for an efficient reuse of previously successful solutions. Transferring this motivation to control theory requires a distance measure between dynamical systems. Given a new dynamical system, it is then possible to find the closest previously controlled system and reuse the corresponding controller. This paper highlights four distance measures and analyzes their suitability for similarity based controller selection. The compared distance measures are Multiple Models Adaptive Control, the Binet-Cauchy kernels, the Martin distance and the ν-gap metric.