Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 21 views

    AME 60617 Bayesian Data Assimilation Homework 3: 1 Problem 1: Standard Kalman Filtering For A Linear Problem

    Uploaded by

    christoumpingfotso
    This document provides instructions for three homework problems in Dr. Jian-Xun Wang's AME 60617 Bayesian Data Assimilation course. Problem 1 involves standard Kalman filtering for a linear object tracking problem. Problem 2 requires deriving the equations for the Extended Kalman Filter. Problem 3 applies the Extended Kalman Filter to track a nonlinear dynamical system involving the evolution of angle, angular velocity, and amplitude over time. Students are asked to generate synthetic data, perform the filtering, and provide plots comparing the results to ground truth.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    AME 60617 Bayesian Data Assimilation Homework 3: 1 Problem 1: Standard Kalman Filtering For A Linear Problem

    Uploaded by

    christoumpingfotso
    21 views3 pages
    This document provides instructions for three homework problems in Dr. Jian-Xun Wang's AME 60617 Bayesian Data Assimilation course. Problem 1 involves standard Kalman filtering for a linear object tracking problem. Problem 2 requires deriving the equations for the Extended Kalman Filter. Problem 3 applies the Extended Kalman Filter to track a nonlinear dynamical system involving the evolution of angle, angular velocity, and amplitude over time. Students are asked to generate synthetic data, perform the filtering, and provide plots comparing the results to ground truth.

    Original Title

    HW3

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document provides instructions for three homework problems in Dr. Jian-Xun Wang's AME 60617 Bayesian Data Assimilation course. Problem 1 involves standard Kalman filtering for a linear object tracking problem. Problem 2 requires deriving the equations for the Extended Kalman Filter. Problem 3 applies the Extended Kalman Filter to track a nonlinear dynamical system involving the evolution of angle, angular velocity, and amplitude over time. Students are asked to generate synthetic data, perform the filtering, and provide plots comparing the results to ground truth.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    21 views3 pages

    AME 60617 Bayesian Data Assimilation Homework 3: 1 Problem 1: Standard Kalman Filtering For A Linear Problem

    Uploaded by

    christoumpingfotso
    This document provides instructions for three homework problems in Dr. Jian-Xun Wang's AME 60617 Bayesian Data Assimilation course. Problem 1 involves standard Kalman filtering for a linear object tracking problem. Problem 2 requires deriving the equations for the Extended Kalman Filter. Problem 3 applies the Extended Kalman Filter to track a nonlinear dynamical system involving the evolution of angle, angular velocity, and amplitude over time. Students are asked to generate synthetic data, perform the filtering, and provide plots comparing the results to ground truth.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 3

    AME 60617 Bayesian Data Assimilation

    Homework 3
    Instructor: Dr. Jian-Xun Wang
    Assistant Professor, Department of Aerospace and Mechanical Engineering

    Note: Please upload your solution to each problem in your homework assignment as a single
    PDF file to the appropriate assignment section. Use the following file naming convention: DA-
    Lastname-FirstName-HW3.pdf. Source codes should also be attached.

    1 Problem 1: Standard Kalman filtering for a linear problem


    Let’s do the example that was showed in the class: object tracking problem. We track an object
    moving in a 2D space with a sensor, which only monitor the object’s position in a Cartesian
    coordinate x and y. For this system, the state will include position (x, y), velocity (ẋ, ẏ), and
    acceleration (ẍ, ÿ). The forward model is a linear one with Gaussian noise,

    xk = Axk−1 + q,

    q ∼ N (0, Q)
    where A and Q is given as,

     1 5 1 4 1 3 
    20 ∆t 0 8 ∆t 0 6 ∆t 0
     0 1 5 1 4 1 3
    1 4 20 ∆t 0 8 ∆t 0 6 ∆t 
     ∆t 1 3 1 2
    8 0 3 ∆t 0 2 ∆t 0 
    Q=
     0 1 4 1 3 1

    2
     8 ∆t 0 3 ∆t 0 2 ∆t 
     1 ∆t3 0 1 2
    6 2 ∆t 0 ∆t 0 
    1 3 1 2
    0 6 ∆t 0 2 ∆t 0 ∆t

    1
    where ∆t = 0.5 and q = 1.

    1. Please generate synthetic data by using the forward model given above with the “ground
    (T ruth)
    truth” of x0 = [0, 0, 1, 0.6, 0.4, 0.8]T (corrupted with some Gaussian measurement noise.)

    2. Use the synthetic data to perform Kalman fitering

    3. Play with the measurement noise as we did in class.

    Note: (1) you can start from x0 = [0, 0, 0, 0, 0, 0]T with state covariance P0 = diag([0.1, 0.1, 0.1, 0.1, 0.5, 0.5]).
    (2) Please show plots of filtered results compared with synthetic observation and ground truth.

    2 Problem 2: Derive non-additive EKF


    Please derive the Extended Kalman filter (non-additive form) algorithm (page 30 in Chapter5-1),
    as what we did in class.

    3 Problem 3: Extended Kalman filtering for a nonlinear problem


    Let’s do another example I showed in the class: considering to track a simple nonlinear dynamic
    system, a random sine signal (θ), where the angular velocity (ω) and the amplitude (a) can very
    over time. So the state is x = (θk , ωk , ak )T . The evolution of the angle θ is modeled with a
    discretized Wiener velocity model, which can be discretized as,

    where q1 = 0.2, q2 = 0.1, and ∆t = 0.01


    The nonlinearity is from the measurement model,

    yk = h(xk , k) + rk = ak sin(θk ) + rk

    and rk is a scalar Gaussian RV.

    1. Please generate synthetic data by using the forward model given above with the “ground
    (T ruth)
    truth” of x0 = [0, 10, 1]T . (corrupted with some Gaussian measurement noise.)

    2. Derive EKF for this problem (explicitly write out Jacobi, forecase, and update ...)

    3. Use the synthetic data to perform Extended Kalman fitering

    2
    4. Play with the measurement noise.

    5. Play with q1 and q2 and see what happens.

    Note: (1) you can start from x0 = [0, 10, 1]T with state covariance P0 = diag([3, 3, 3]). (2) Please
    show plots of filtered results compared with synthetic observation and ground truth.

    You might also like