View source for SURF 2017: Data-driven models for temporal logic control

'''[[SURF 2017|2017 SURF]]: project description'''
* Mentor: Richard M. Murray
* Co-mentor: Sofie Haesaert

The quest for reliable controller design satisfying complex functionality requirements has recently led to the development of automatic design and verification methods based on temporal logic specifications [1].

The dependence of these methods on exact model knowledge inhibits industrial adoption in the control engineering community, where knowledge on system dynamics is often uncertain and models built from data are common practice. Using system identification [3] a math- ematical model can be selected based on experiment data. For these data-driven models, the certification of the resulting controllers with respect to performance criteria and the cost-effective gathering of data for control applications is a mature field [2]. Performance objectives for which these identification methods yield certified controllers include refer- ence tracking, disturbance attenuation, stabilization, etcetera. As of yet, there are little results on the design and certification of data-driven controllers with respect to temporal logic specifications.

Within the field of data-driven methods for control synthesis contributions can be made by the SURF towards :
# Certifying temporal logic specifications for identified models: As a first topic the SURF can consider the use of linear time invariant models ob- tained from data for the design of controllers. For this the development of theoretical guarantees on the satisfaction of temporal logic specifications is of interest.
# Temporal logic control with integrated experiment design: The quality of a model that is estimated from measurements depends on the informa- tion content of the data. To avoid costly down-time, it is often of interest to perform experiments online during the normal operation of the system [4]. The optimal design of these experiments, while still satisfying temporal logic specifications with respect the normal operation, is a secondary potential topic of interest.

Familiarity with the following topics is useful but not necessary:
* Bayesian inference or system identification
* Control based on temporal logic specifications
* Receding horizon control and dynamic programming

References
# Wongpiromsarn, T., Topcu, U., Ozay, N., Xu, H., & Murray, R. M. “TuLiP: a software toolbox for receding horizon temporal logic planning.” In Proceedings of the 14th inter- national conference on Hybrid systems: computation and control, pp. 313-314. ACM, 2011.
# Gevers, M., “Identification for Control: From the Early Achievements to the Revival of Experiment Design.”, European Journal of Control, 2005, pp. 335-352.
# Ljung, L. “System identification.” Signal Analysis and Prediction. Birkh ̈auser Boston, 1998. 163-173.
# Larsson, C. A., Annergren, M., Hjalmarsson, H., Rojas, C. R., Bombois, X., Mesbah, A., & Mod ́en, P. E. ,“Model predictive control with integrated experiment design for output error systems,” 2013 European Control Conference (ECC), Zurich, 2013, pp. 3790-3795.