On Linear Time Invariant Systems Analysis via A Single Trajectory: A Linear Programming Approach

In this note, a novel methodology that can extract a number of analysis results for linear time-invariant systems (LTI) given only a single trajectory of the considered system is proposed. The superiority of the proposed technique relies on the fact that it provides an automatic and formal way to obtain a valuable information about the controlled system by only having access to a single trajectory over a finite period of time (i.e., the system dynamics is assumed to be unknown). At first, we characterize the stability region of LTI systems given only a single trajectory dataset by constructing the associated Lyapunov function of the system. The Lyapunov function is found by formulating and solving a linear programming (LP) problem. Then, we extend the same methodology to a variety of essential analysis results for LTI systems such as deriving bounds on the output energy, deriving bounds on output peak, deriving L2 and RMS gains. To illustrate the efficacy of the proposed data-driven paradigm, a comparison analysis between the learned LTI system metrics and the true ones is provided.

[1]  Sandra Hirche,et al.  How Training Data Impacts Performance in Learning-Based Control , 2020, IEEE Control Systems Letters.

[2]  Sriram Sankaranarayanan,et al.  Learning Lyapunov (Potential) Functions from Counterexamples and Demonstrations , 2017, Robotics: Science and Systems.

[3]  Sicun Gao,et al.  Neural Lyapunov Control , 2020, NeurIPS.

[4]  A. Papachristodoulou,et al.  On the construction of Lyapunov functions using the sum of squares decomposition , 2002, Proceedings of the 41st IEEE Conference on Decision and Control, 2002..

[5]  K. Narendra,et al.  A common Lyapunov function for stable LTI systems with commuting A-matrices , 1994, IEEE Trans. Autom. Control..

[6]  Manfred Morari,et al.  Learning Lyapunov Functions for Hybrid Systems , 2020, 2021 55th Annual Conference on Information Sciences and Systems (CISS).

[7]  S. Shankar Sastry,et al.  Provably safe and robust learning-based model predictive control , 2011, Autom..

[8]  A. Papachristodoulou,et al.  A tutorial on sum of squares techniques for systems analysis , 2005, Proceedings of the 2005, American Control Conference, 2005..

[9]  Robert Shorten,et al.  On Linear Copositive Lyapunov Functions and the Stability of Switched Positive Linear Systems , 2007, IEEE Transactions on Automatic Control.

[10]  Sandra Hirche,et al.  Parameter Optimization for Learning-based Control of Control-Affine Systems , 2020, L4DC.

[11]  E. Yaz.  Linear Matrix Inequalities In System And Control Theory , 1998, Proceedings of the IEEE.

[12]  Sergey Levine,et al.  Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models , 2018, NeurIPS.

[13]  Andreas Krause,et al.  Safe controller optimization for quadrotors with Gaussian processes , 2015, 2016 IEEE International Conference on Robotics and Automation (ICRA).

[14]  Alessandro Abate,et al.  Formal Synthesis of Lyapunov Neural Networks , 2020, IEEE Control Systems Letters.

[15]  R. Stephenson.  A and V , 1962, The British journal of ophthalmology.

[16]  Carl E. Rasmussen,et al.  Gaussian Processes for Data-Efficient Learning in Robotics and Control , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[17]  Sandra Hirche,et al.  The Impact of Data on the Stability of Learning-Based Control , 2020, L4DC.

[18]  Dana Kulic,et al.  Stable Gaussian Process based Tracking Control of Euler-Lagrange Systems , 2018, Autom..

[19]  Jaime F. Fisac,et al.  A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems , 2017, IEEE Transactions on Automatic Control.

[20]  Michael Jackson,et al.  Optimal Design of Experiments , 1994 .