Victor G. Lopez,Matthias A. Müller,Paolo Rapisarda
Abstract:We illustrate a novel version of Willems' lemma for data-based representation of continuous-time systems. The main novelties compared to previous works are two. First, the proposed framework relies only on measured input-output trajectories from the system and no internal (state) information is required. Second, our system representation makes use of exact system trajectories, without resorting to orthogonal bases representations and consequent approximations. We first establish sufficient and necessary conditions for data-based generation of system trajectories in terms of suitable latent variables. Subsequently, we reformulate these conditions using measured input-output data and show how to span the full behavior of the system. Furthermore, we show how to use the developed framework to solve the data-based continuous-time simulation problem.
What problem does this paper attempt to address?
This paper aims to solve the problem of data - based trajectory representation for continuous - time systems. Specifically, it attempts to overcome two main limitations in existing methods:
1. **No Dependence on Internal State Information**: Existing methods, such as the framework proposed in reference [5], require the measurement of the internal state of the system (i.e., state variables), which may be difficult to achieve in practical applications. The method proposed in this paper only depends on the measured input - output trajectories and does not require internal state information.
2. **Accurate Representation of System Trajectories**: Existing methods, such as the orthogonal basis representation method used in reference [12], approximate continuous - time trajectories, which results in the generated trajectories being essentially approximations of the real trajectories. The method proposed in this paper can accurately calculate data - driven continuous - time system trajectories, and these trajectories only depend on input - output measurements.
### Main Contributions
1. **New Perspective**: By introducing the concept of "jets", the differential equations required for generating system trajectories are related to the input - output trajectories and their derivatives.
2. **Extended Method**: The method in reference [5] is extended from internal state measurement to input - output measurement, achieving data - based accurate trajectory generation.
3. **Connecting Different Methods**: The relationships between the methods in references [5], [11], and [12] are explored, providing a more comprehensive understanding.
4. **Data - Driven Simulation Problem**: It is shown how to use the proposed method to solve the data - driven continuous - time simulation problem.
### Problem Statement
In the behavioral framework, a dynamic system is described by its behavior \( \mathcal{B} \), which contains all input - output trajectories allowed by the system dynamics. This paper considers the linear differential behavior \( \mathcal{B} \) with \( m \) inputs and \( p \) outputs. The specific problem can be formalized as:
**Problem 1**: Determine the necessary and sufficient conditions for data - based system representation of continuous - time systems using only the measurements of external variables.
### Main Results
1. **Data - Driven Controllable System Representation**: For a controllable linear differential behavior \( \mathcal{B} \), it can be described by a \( d \)-dimensional latent variable \( \ell \). By defining time - dependent matrices \( H(u) \), \( H(y) \), and \( H(\ell) \) and using these matrices, the necessary and sufficient conditions for input - output trajectories can be derived.
2. **Externally Traversed with Sufficient Information**: The concept of "jets" is introduced, and the \( L \)-order jet \( J_L(u, y) \) of the system trajectory is defined. By defining the data matrix, the concept of an externally traversed with sufficient information is proposed, ensuring a one - to - one correspondence between the left annihilator of the data matrix and the differential equations describing the system \( \mathcal{B} \).
3. **From Input - Output Data to Acceptable \( L \)-order Jets**: A data - based linear differential system characterization method is proposed, which only depends on external variables \( u \) and \( y \), and the necessary and sufficient conditions for generating acceptable \( L \)-order jets are given.
### Data - Driven Simulation Problem
The problem of how to find the corresponding output trajectory when given the initial conditions and input trajectory is solved. By solving an implicit differential equation system, the output trajectory satisfying the initial conditions can be obtained.
### Conclusion
This paper proposes a novel continuous - time version of Willems' fundamental lemma, providing the necessary and sufficient conditions for data - based system trajectory representation. This method overcomes the limitations of existing methods, can accurately represent system trajectories, and only depends on the measurements of external variables. Future research directions include calculating non - minimal state variables and their applications in data - driven system representation.