Extended Kalman Filter (EKF) And some other useful Kalman stuff!

Slides:



Advertisements
Similar presentations
EKF, UKF TexPoint fonts used in EMF.
Advertisements

Properties of State Variables
Observers and Kalman Filters
Presenter: Yufan Liu November 17th,
Kalman Filter CMPUT 615 Nilanjan Ray. What is Kalman Filter A sequential state estimator for some special cases Invented in 1960’s Still very much used.
Introduction to Mobile Robotics Bayes Filter Implementations Gaussian filters.
Lecture 11: Recursive Parameter Estimation
280 SYSTEM IDENTIFICATION The System Identification Problem is to estimate a model of a system based on input-output data. Basic Configuration continuous.
Maximum likelihood (ML) and likelihood ratio (LR) test
Mobile Intelligent Systems 2004 Course Responsibility: Ola Bengtsson.
Laplace Transforms Important analytical method for solving linear ordinary differential equations. - Application to nonlinear ODEs? Must linearize first.
Probabilistic Robotics
Comparative survey on non linear filtering methods : the quantization and the particle filtering approaches Afef SELLAMI Chang Young Kim.
Kalman Filtering Jur van den Berg. Kalman Filtering (Optimal) estimation of the (hidden) state of a linear dynamic process of which we obtain noisy (partial)
Particle Filtering for Non- Linear/Non-Gaussian System Bohyung Han
Estimation and the Kalman Filter David Johnson. The Mean of a Discrete Distribution “I have more legs than average”
Course AE4-T40 Lecture 5: Control Apllication
Probabilistic Robotics Bayes Filter Implementations Gaussian filters.
Linear and generalised linear models
Linear and generalised linear models
Kalman Filtering Pieter Abbeel UC Berkeley EECS Many slides adapted from Thrun, Burgard and Fox, Probabilistic Robotics TexPoint fonts used in EMF. Read.
Maximum likelihood (ML)
Ordinary Differential Equations (ODEs) 1Daniel Baur / Numerical Methods for Chemical Engineers / Implicit ODE Solvers Daniel Baur ETH Zurich, Institut.
Lecture II-2: Probability Review
Overview and Mathematics Bjoern Griesbach
Adaptive Signal Processing
ROBOT MAPPING AND EKF SLAM
Principles of the Global Positioning System Lecture 13 Prof. Thomas Herring Room A;
1 Formation et Analyse d’Images Session 7 Daniela Hall 7 November 2005.
Kalman filter and SLAM problem
Lecture 11: Kalman Filters CS 344R: Robotics Benjamin Kuipers.
Biointelligence Laboratory, Seoul National University
Computer vision: models, learning and inference Chapter 19 Temporal models.
Computer Vision Group Prof. Daniel Cremers Autonomous Navigation for Flying Robots Lecture 6.2: Kalman Filter Jürgen Sturm Technische Universität München.
Computer vision: models, learning and inference Chapter 19 Temporal models.
Kalman Filter (Thu) Joon Shik Kim Computational Models of Intelligence.
Probabilistic Robotics Bayes Filter Implementations Gaussian filters.
Probabilistic Robotics Bayes Filter Implementations.
1 Chapter 2 1. Parametric Models. 2 Parametric Models The first step in the design of online parameter identification (PI) algorithms is to lump the unknown.
Human-Computer Interaction Kalman Filter Hanyang University Jong-Il Park.
NCAF Manchester July 2000 Graham Hesketh Information Engineering Group Rolls-Royce Strategic Research Centre.
Unscented Kalman Filter 1. 2 Linearization via Unscented Transform EKF UKF.
An Introduction To The Kalman Filter By, Santhosh Kumar.
By: Aaron Dyreson Supervising Professor: Dr. Ioannis Schizas
Extended Kalman Filter
Nonlinear State Estimation
Cameron Rowe.  Introduction  Purpose  Implementation  Simple Example Problem  Extended Kalman Filters  Conclusion  Real World Examples.
The Unscented Kalman Filter for Nonlinear Estimation Young Ki Baik.
École Doctorale des Sciences de l'Environnement d’ Î le-de-France Année Modélisation Numérique de l’Écoulement Atmosphérique et Assimilation.
DSP-CIS Part-III : Optimal & Adaptive Filters Chapter-9 : Kalman Filters Marc Moonen Dept. E.E./ESAT-STADIUS, KU Leuven
Probabilistic Robotics Bayes Filter Implementations Gaussian filters.
CSC321: Neural Networks Lecture 9: Speeding up the Learning
STATISTICAL ORBIT DETERMINATION Kalman (sequential) filter
STATISTICAL ORBIT DETERMINATION Coordinate Systems and Time Kalman Filtering ASEN 5070 LECTURE 21 10/16/09.
H.L. Tanaka and K. Kondo University of Tsukuba
PSG College of Technology
Unscented Kalman Filter
Unscented Kalman Filter
Kalman Filtering: Control with Limited/Noisy Measurements
The regression model in matrix form
Optimal prediction of x
Filtering and State Estimation: Basic Concepts
8. Stability, controllability and observability
Unscented Kalman Filter
Extended Kalman Filter
Kalman Filtering COS 323.
Principles of the Global Positioning System Lecture 13
Extended Kalman Filter
Laplace Transforms Important analytical method for solving linear ordinary differential equations. - Application to nonlinear ODEs? Must linearize first.
Laplace Transforms Important analytical method for solving linear ordinary differential equations. - Application to nonlinear ODEs? Must linearize first.
Presentation transcript:

Extended Kalman Filter (EKF) And some other useful Kalman stuff!

The Linear (normal) KF

What if the system are not described in a linear manner? I.e. if we divide by a state variable, multiplies two state variables or e.g. applies a function (e.g. a sinus) on one or more of our state variables. A lot of real world models are non-linear. How can we extend the Kalman filter – which is an linear optimal state estimator – to real world problems, which not always, in the first point of view, may not be described in a linear manner?

We goes from this system description, which is linear To this system description, which may be non-linear

How should we handle the Kalman gain?

First approach The obvious solution is to use a Kalman filter on a non-linear system, by linearize the model around a point of operation. But, this could require a significantly higher order of model, to be able to describe the non- linear behavior. Should we use a non-linear function in place of K?

The solution is much smarter! K will be replaced by another K, just obtained “slightly” different. Extension of the linear Kalman filter to the non- linear cases requires a few more steps in the implementation. But the most of the calculations may be done in compile time, which make the online calculations only a little more intense, in case than the non- linear functions themselves are not computationally heavier. We still only have to compute a single matrix inverse.

Partial Derivative Matrix A.k.a. Jacobian matrix will be the tool to handle the EKF. Remember that the Jacobian matrix is defined as Remember this!

Quiz Consider the system described by the following state update function: 1,1 2,1 1,2 2,2

The EKF algorithm (1/4) The system and measurement equations are given as follows: Initialize the system:

The EKF algorithm (2/4) For each time step k=1,2,…, compute the following: a)Compute the following partial derivative matrices (Jacobian):

The EKF algorithm (3/4) b)Perform the time update of the state estimate and estimation-error co-variance: c)Compute the partial derivative matrices:

The EKF algorithm (4/4) d)Perform the measurement update of the state estimate and the estimation error co-variance matrix: …and go back to a). Note that numerical more precise equations are available for the co-variance update function.

Discretization (1/5) There are different approaches to discretization. E.g. we could have a linear continuous time state space model described by the system equations This could be transformed to a discrete model by

Discretization (2/5) Another, and maybe the simplest approach, will be to transform the individual sums of products in the continuous time update equation from the definition of the derivative. E.g. like where Δt is the samplingstime.

Discretization (3/5) An obviously approach will also be to use Taylor expansion. This could make us skip the step of going through the continuous time state space model to reach the discrete time model. E.g. if M is a constant and F and (moment and angular position) are state variables we could discretize the expression

Discretization (4/5)

Discretization (5/5) Using up to second order the discrete time model will be

Other Need to Know about Kalman System is Reachable ((F,G) pair is Reachable) if, starting from a null initial state, we can always find an input signal which leads the system to any a-priori selected final state in finite time.

Condition to check Reachability

Other Need to Know about Kalman System S is called Observable (the (F,H) pair is observable) if two different initial states do not exist, such that their corresponding outputs are exactly the same, for each t≥0.

Condition to check Observability

Let’s check Observability

Nice to Know about Kalman If the following conditions hold: Uncorrelated process and measurement noise! And one of either: a)System is reachable from the noise’s point of view, and the system is observable. b)Or all F eigenvalues are strictly inside the unitary circle. It will make the system asymptotically converge.

Gain of Asymptotic Convergence

Exercise on the class! See the uploaded document: A Mechanical System.pdf

References [1]S. M. Savaresi, ”Model Identification and Adaptive Systems - KALMAN FILTER”, Milan: Politecnico di Milano, [2]D. Simon, “Optimal State Estimation, Kalman, H ∞ and Nonlinear Approaches”, Hoboken, New Jersey: Wiley, [3]Welch, G and Bishop, G “An introduction to the Kalman Filter”,