Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
194 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

A Unified Filter for Simultaneous Input and State Estimation of Linear Discrete-time Stochastic Systems (1309.6627v4)

Published 25 Sep 2013 in math.OC, cs.SY, and math.DS

Abstract: In this paper, we present a unified optimal and exponentially stable filter for linear discrete-time stochastic systems that simultaneously estimates the states and unknown inputs in an unbiased minimum-variance sense, without making any assumptions on the direct feedthrough matrix. We also derive input and state observability/detectability conditions, and analyze their connection to the convergence and stability of the estimator. We discuss two variations of the filter and their optimality and stability properties, and show that filters in the literature, including the Kalman filter, are special cases of the filter derived in this paper. Finally, illustrative examples are given to demonstrate the performance of the unified unbiased minimum-variance filter.

Citations (166)

Summary

  • The paper’s main contribution is a unified filter that achieves optimal, unbiased estimation of states and unknown inputs.
  • It presents two variants, ULISE and PLISE, using system transformations to decouple components with and without direct feedthrough.
  • The approach guarantees global optimality and exponential stability under standard observability and detectability conditions for practical applications.

Analysis of a Unified Filter for Simultaneous Input and State Estimation in Linear Discrete-time Stochastic Systems

This paper introduces a comprehensive filter designed to achieve optimal and unbiased minimum-variance estimation of states and unknown inputs within linear discrete-time stochastic systems. The crux of the work is the simultaneous estimation without restrictions on the direct feedthrough matrix, marking an advancement over prior methodologies that imposed limitations on the rank and structure of feedthrough matrices. This novelty circumvents issues faced by classical filters like the Kalman filter, which traditionally handle Gaussian white noise disturbances well but are suboptimal with non-Gaussian inputs.

Core Contributions and Theoretical Implications

The authors leverage a specialized system transformation to decouple the output equations, distinguishing between components with and without direct feedthrough. This allows the adoption of established techniques separately on each component while utilizing a singular framework. Two filter variations are proposed: the Updated Linear Input and State Estimator (ULISE) and the Propagated Linear Input and State Estimator (PLISE). ULISE utilizes updated state estimates for input prediction, while PLISE uses the propagated estimates. Both variants hold promises of unbiasedness and optimality but differ in their approach to sequence step execution within the estimation process.

The primary contribution lies in demonstrating that ULISE is globally optimal among all linear filters due to its structural design. Furthermore, the paper establishes connections and expands upon existing filters, including those by Gillijns and De Moor, and filters that deal with systems with full feedthrough matrices. The paper also provides detailed algorithms for these filters, outlines their implementation, and discusses matrix transformations crucial for satisfying observability and detectability conditions.

Implications and Future Directions

From a theoretical perspective, the paper makes significant strides in showing how exponential stability of filter estimates can be deduced from system detectability and stabilizability conditions. For systems where the direct feedthrough matrix has full rank, as proved, the filter's structure naturally converges to existing globally optimal solutions. The paper also identifies the strong detectability condition as pivotal for attaining unbiased and stable estimates, tying convergence properties of complex stochastic systems directly to their observability conditions.

Practically, the ability to estimate unknown system inputs while maintaining unbiased state predictions is crucial in various fields like autonomous vehicle navigation, fault detection, and control systems design where inputs are often unmeasurable or corrupted. This enriches the operational reliability and security of systems engaging in real-world environments subject to unpredictable actions or errors.

Looking forward, application of this unified filter to more complex nonlinear, multi-agent, or continuous-time systems remains an open field of research. Extending this work towards hybrid settings with non-linearity and real-world parameter variations will further demonstrate its flexibility and robustness.

In conclusion, this paper contributes valuably by introducing a unified framework for state and input estimation in linear stochastic systems without constraints on feedthrough matrices. The authors have methodically provided evidence for global optimality of their proposed approach and established a strong theoretical foundation for future adaptations and applications. Through simulation examples, they substantiate the practical significance, opening avenues for extensive research in efficient estimation under uncertain conditions.