On Functional Observability of Nonlinear Systems and the Design of Functional Observers with Assignable Error Dynamics

Basic Information

• Paper ID: 2501.00167
• Title: On Functional Observability of Nonlinear Systems and the Design of Functional Observers with Assignable Error Dynamics
• Author: Costas Kravaris (Texas A&M University)
• Classification: eess.SY cs.SY
• Publication Date: January 2025
• Paper Link: https://arxiv.org/abs/2501.00167

Abstract

This paper proposes a novel methodology for designing functional observers for nonlinear systems with linear error dynamics and assignable poles. First, sufficient conditions for functional observability are derived, establishing functional relationships between the Lie derivatives of the output to be estimated and the measured output. These relationships are directly utilized in the proposed functional observer design. The functional observer is defined in differential input-output form, satisfying appropriate invariance conditions evolved from state-space invariance conditions in the literature. The concept of functional observer index is introduced to characterize the minimum feasible order of functional observers with pole assignment. The proposed methodology is illustrated through two chemical reactor applications.

Research Background and Motivation

Problem Definition

The core problem addressed in this research is the design of functional observers for nonlinear systems, particularly in applications where only a function of the state rather than the entire state vector needs to be estimated. Traditional state observers require reconstruction of the complete state vector, which is unnecessary and computationally complex in many practical applications.

Problem Significance

Functional observers have important value in numerous engineering applications, particularly in:

1. Fault detection and diagnosis systems
2. Control systems requiring feedback of specific state functions
3. Reducing observer order to improve computational efficiency
4. Chemical engineering process monitoring

Limitations of Existing Methods

1. Linear System Approaches: Luenberger's pioneering work primarily addresses linear systems and cannot be directly applied to nonlinear systems
2. Exact Linearization Methods: Existing nonlinear functional observer designs based on exact linearization require stringent conditions
3. State-Space Formulation: Most existing methods start from linear error dynamics requirements in state-space form, limiting design flexibility

Research Motivation

The innovation of this paper lies in designing functional observers directly in input-output form starting from functional observability, avoiding limitations of traditional methods and providing a design methodology with linear error dynamics and assignable poles for any locally functionally observable nonlinear system.

Core Contributions

1. Proposed new definitions and sufficient conditions for functional observability of nonlinear systems, establishing functional relationships based on Lie derivatives
2. Developed functional observer design methodology in input-output form, avoiding limitations of state-space formulation
3. Introduced the concept of functional observer index to characterize the minimum feasible order of functional observers
4. Provided linear error dynamics design with assignable poles, ensuring observer stability and convergence
5. Established correspondence between state-space invariance conditions and input-output invariance conditions
6. Verified practical applicability through chemical reactor examples

Methodology Details

Problem Formulation

Consider a nonlinear system of the following form:

dx/dt = F(x)
y = H(x)  
z = q(x)

where x ∈ ℝⁿ is the system state, y ∈ ℝᵖ is the measured output vector, and z is the output function to be estimated. The objective is to construct a functional observer that generates an estimate of z based solely on output measurements y.

Core Theoretical Framework

1. Functional Observability Definition

Definition 4: A system is functionally observable on Χ if for any two initial states x₁, x₂ ∈ Χ:

H_F(t;x₁) = H_F(t;x₂) ⟹ q(x₁) = q(x₂)

This is equivalent to the Lie derivative condition:

L_F^i H_j(x₁) = L_F^i H_j(x₂), ∀i=0,1,2,..., j=1,...,p ⟹ q(x₁) = q(x₂)

2. Sufficient Conditions for Functional Observability

Proposition 2: If there exists a positive integer m such that q(x) can be locally expressed as a function of elements in the observability set:

q(x) = ψ(L_F^i H_j(x), i=0,1,...,m-1, j=1,...,p)

then the system is locally functionally observable.

3. Input-Output Form Functional Observer

Definition 6: A dynamic system defined by a ν-th order differential equation:

d^ν ẑ/dt^ν = Φ(ẑ, dẑ/dt, ..., d^(ν-1)ẑ/dt^(ν-1), y, dy/dt, ..., d^ν y/dt^ν)

is called a functional observer for the system if it satisfies the invariance condition:

L_F^ν q(x) = Φ(q(x), L_F q(x), ..., L_F^(ν-1) q(x), H(x), L_F H(x), ..., L_F^ν H(x))

Linear Functional Observer Design

Design Template

A linear functional observer form is proposed:

d^ν ẑ/dt^ν + α_(ν-1) d^(ν-1)ẑ/dt^(ν-1) + ... + α₁ dẑ/dt + α₀ ẑ = 
Γ(y, dy/dt, ..., d^(ν-1)y/dt^(ν-1))

Invariance Condition

The invariance condition for the linear observer is:

L_F^ν q(x) + α_(ν-1) L_F^(ν-1) q(x) + ... + α₁ L_F q(x) + α₀ q(x) = 
Γ(H(x), L_F H(x), ..., L_F^(ν-1) H(x))

Error Dynamics Analysis

Defining the error e = ẑ - q(x), the error dynamics equation is obtained:

d^ν e/dt^ν + α_(ν-1) d^(ν-1)e/dt^(ν-1) + ... + α₁ de/dt + α₀ e = 0

By selecting the roots of the polynomial λⁿ + α_(ν-1)λ^(ν-1) + ... + α₁λ + α₀ to have negative real parts, asymptotic convergence of the error to zero is guaranteed.

Functional Observer Index

Definition 7: The minimum positive integer ν satisfying the condition:

L_F^k q(x) = ψₖ(H_j(x), L_F H_j(x), ..., L_F^ν H_j(x)), k=0,1,...,ν, j=1,...,p

is called the functional observer index. This index characterizes the minimum feasible order of functional observers achieving pole assignment.

Experimental Setup

Application Case 1: Isothermal Batch Chemical Reactor

Consider the continuous irreversible chemical reaction A → B → C → D:

• First and third reactions follow first-order kinetics
• Second reaction follows second-order kinetics
• Measurement: c_B (concentration of B)
• Estimation: c_A (concentration of A)

Application Case 2: Non-isothermal Continuous Chemical Reactor

Consider the irreversible exothermic chemical reaction A → B:

• First-order kinetics with Arrhenius reaction rate
• Cooled via cooling jacket
• Measurements: reaction mixture temperature θ and jacket temperature θ_J
• Estimation: concentration of A, c_A

Design Process

For each case:

1. Establish system dynamics equations
2. Compute Lie derivative relationships
3. Determine functional observer index
4. Design linear functional observer
5. Verify invariance conditions

Experimental Results

Case 1 Results

• Functional Observer Index: ν = 1
• Functional Relationships:

  ψ₀(y, dy/dt) = (1/k₁)y + (1/k₂)(dy/dt)
  ψ₁(y, dy/dt) = -k₂(dy/dt) + (dy/dt)
  

• Observer Equation:

  dẑ/dt - λẑ = -λ(1/k₁ + 1/k₂)(dy/dt) - λ(1/k₁)y
  

Case 2 Results

• Functional Observer Index: ν = 1
• Successfully established both input-output and state-space forms of the observer
• Verified the effectiveness of the theoretical design

Key Findings

1. Low-Order Observers: Both cases achieved first-order functional observers, significantly lower than the order of complete state observers
2. Pole Assignment: Successfully implemented pole assignment for error dynamics, allowing arbitrary specification of convergence rate
3. Practicality: Observers can be directly implemented in input-output form, facilitating engineering applications

Related Work

Functional Observers for Linear Systems

• Luenberger (1966, 1971): Pioneering definition of functional observer concept for linear systems
• Darouach (2000): Existence and design of linear functional observers
• Fernando et al. (2010): Minimum-order linear functional observer design

Observers for Nonlinear Systems

• Hermann & Krener (1977): Foundational observability theory for nonlinear systems
• Krener & Isidori (1983): Linearization-based observers using output injection
• Kazantzis & Kravaris (1998): Nonlinear observer design using Lyapunov auxiliary theorem

Advantages of This Work

1. Direct Input-Output Design: Avoids restrictive conditions of exact linearization
2. Functional Observability Theory: Provides a more general theoretical framework
3. Pole Assignment Capability: Achieves pole assignment for nonlinear system functional observers

Conclusions and Discussion

Main Conclusions

1. Theoretical Contribution: Established a complete theoretical framework for functional observability of nonlinear systems
2. Design Methodology: Provided direct input-output form functional observer design methodology
3. Practical Value: Applicable to any locally functionally observable nonlinear system
4. Performance Guarantee: Achieves linear error dynamics with assignable poles

Limitations

1. Local Nature: Theoretical results are local, requiring validity within specific regions
2. Smoothness Requirements: System functions must possess sufficient smoothness
3. Computational Complexity: Calculation of high-order Lie derivatives may be complex
4. Real-Time Implementation: Requires computation of output derivatives, potentially introducing noise sensitivity

Future Directions

1. Global Results: Extension to global functional observability and observer design
2. Robustness: Consideration of model uncertainty and measurement noise effects
3. Numerical Methods: Development of efficient Lie derivative computation methods
4. Application Extension: Expansion to broader engineering application domains

In-Depth Evaluation

Strengths

1. Theoretical Innovation:
   • Proposes a new paradigm for input-output form functional observer design
   • Establishes rigorous theoretical foundation for functional observability and Lie derivative relationships
   • Functional observer index concept has significant theoretical value
2. Methodological Advantages:
   • Avoids restrictive conditions of exact linearization
   • Achieves pole assignment, providing design freedom
   • Input-output form facilitates engineering implementation
3. Experimental Verification:
   • Chemical reactor cases have practical engineering background
   • Detailed demonstration of each design step
   • Validates feasibility of theoretical methodology

Weaknesses

1. Theoretical Limitations:
   • Results limited to local regions, lacking global analysis
   • High smoothness requirements limit applicability
   • Does not consider model uncertainty and external disturbances
2. Experimental Deficiencies:
   • Lacks numerical simulation results validating observer performance
   • No quantitative comparison with other methods
   • Cases are relatively simple, lacking complex system verification
3. Practical Issues:
   • Requires precise computation of output derivatives, potentially introducing noise in practice
   • High-order Lie derivative computation is complex, potentially affecting real-time performance
   • Lacks robustness analysis

Impact Assessment

1. Academic Value: Provides new research direction for nonlinear system functional observer theory
2. Engineering Value: Provides practical design methodology for real engineering applications
3. Reproducibility: Clear theoretical derivations and explicit design steps ensure good reproducibility

Applicable Scenarios

1. Process Control: Chemical engineering, bioengineering, and other process monitoring
2. Fault Diagnosis: Fault detection systems requiring specific state function estimation
3. Robot Control: Control applications requiring only partial state information
4. Aerospace: Aircraft attitude and trajectory parameter estimation

References

1. Luenberger, D.G. (1966). Observers for multivariable systems. IEEE Trans. Automat. Contr.
2. Hermann, R. and Krener, A.J. (1977). Nonlinear Controllability and Observability. IEEE Trans. Automat. Contr.
3. Kravaris, C. and Venkateswaran, S. (2021). Systems & Control Letters.
4. Montanari, A.N. et al. (2022). Functional observability and subspace reconstruction in nonlinear systems. Physical Review Research.

Overall Assessment: This is a paper with significant contributions to nonlinear system observer theory. The author's proposed input-output form functional observer design methodology is innovative, with rigorous theoretical derivations and appropriate application examples. Despite certain limitations, it provides valuable theoretical tools and design methods for the advancement of this field.