4. We noted in the text that the EKF linearization is an approximation. To see how bad this approximation is, we ask you to work out an example. Suppose we have a mobile robot operating in a planar environment. Its state is its r-y-location and its global heading direction 8. Suppose we know x and y with high certainty, but the orientation is unknown. This is reflected by our initial estimate (000) and Σ = 0 0.01 0 0 0.01 0 0 0 10000 (a) Draw, graphically, your best model of the posterior over the robot pose after the robot moves d = 1 units forward. For this exercise, we assume the robot moves flawlessly without any noise. Thus, the expected lo- cation of the robot after motion will be (E) y + sin For your drawing, you can ignore and only draw the posterior in r-y-coordinates. (b) Now develop this motion into a prediction step for the EKF. For that, you have to define a state transition function and linearize it. You then have to generate a new Gaussian estimate of the robot pose using the linearized model. You should give the exact mathematical equations for each of these steps, and state the Gaussian that results. (c) Draw the uncertainty ellipse of the Gaussian and compare it with your intuitive solution. (d) Now incorporate a measurement. Our measurement shall be a noisy projection of the z-coordinate of the robot, with covariance Q = 0.01. Specify the measurement model. Now apply the measurement both to your intuitive posterior, and formally to the EKF estimate using the standard EKF machinery. Give the exact result of the EKF, and compare it with the result of your intuitive analysis. (e) Discuss the difference between your estimate of the posterior, and the Gaussian produced by the EKF. How significant are those differences? What can be changed to make the approximation more accurate? What would have happened if the initial orientation had been known, but not the robot's y-coordinate?
4. We noted in the text that the EKF linearization is an approximation. To see how bad this approximation is, we ask you to work out an example. Suppose we have a mobile robot operating in a planar environment. Its state is its r-y-location and its global heading direction 8. Suppose we know x and y with high certainty, but the orientation is unknown. This is reflected by our initial estimate (000) and Σ = 0 0.01 0 0 0.01 0 0 0 10000 (a) Draw, graphically, your best model of the posterior over the robot pose after the robot moves d = 1 units forward. For this exercise, we assume the robot moves flawlessly without any noise. Thus, the expected lo- cation of the robot after motion will be (E) y + sin For your drawing, you can ignore and only draw the posterior in r-y-coordinates. (b) Now develop this motion into a prediction step for the EKF. For that, you have to define a state transition function and linearize it. You then have to generate a new Gaussian estimate of the robot pose using the linearized model. You should give the exact mathematical equations for each of these steps, and state the Gaussian that results. (c) Draw the uncertainty ellipse of the Gaussian and compare it with your intuitive solution. (d) Now incorporate a measurement. Our measurement shall be a noisy projection of the z-coordinate of the robot, with covariance Q = 0.01. Specify the measurement model. Now apply the measurement both to your intuitive posterior, and formally to the EKF estimate using the standard EKF machinery. Give the exact result of the EKF, and compare it with the result of your intuitive analysis. (e) Discuss the difference between your estimate of the posterior, and the Gaussian produced by the EKF. How significant are those differences? What can be changed to make the approximation more accurate? What would have happened if the initial orientation had been known, but not the robot's y-coordinate?
Algebra & Trigonometry with Analytic Geometry
13th Edition
ISBN:9781133382119
Author:Swokowski
Publisher:Swokowski
Chapter7: Analytic Trigonometry
Section7.6: The Inverse Trigonometric Functions
Problem 94E
Related questions
Question
Transcribed Image Text:4. We noted in the text that the EKF linearization is an approximation. To
see how bad this approximation is, we ask you to work out an example.
Suppose we have a mobile robot operating in a planar environment. Its
state is its r-y-location and its global heading direction 8. Suppose we
know x and y with high certainty, but the orientation is unknown. This
is reflected by our initial estimate
(000) and Σ = 0
0.01 0
0
0.01
0
0
0 10000
(a) Draw, graphically, your best model of the posterior over the robot pose
after the robot moves d = 1 units forward. For this exercise, we assume
the robot moves flawlessly without any noise. Thus, the expected lo-
cation of the robot after motion will be
(E)
y + sin
For your drawing, you can ignore and only draw the posterior in
r-y-coordinates.
(b) Now develop this motion into a prediction step for the EKF. For that,
you have to define a state transition function and linearize it. You then
have to generate a new Gaussian estimate of the robot pose using the
linearized model. You should give the exact mathematical equations
for each of these steps, and state the Gaussian that results.
(c) Draw the uncertainty ellipse of the Gaussian and compare it with your
intuitive solution.
(d) Now incorporate a measurement. Our measurement shall be a noisy
projection of the z-coordinate of the robot, with covariance Q = 0.01.
Specify the measurement model. Now apply the measurement both
to your intuitive posterior, and formally to the EKF estimate using the
standard EKF machinery. Give the exact result of the EKF, and compare
it with the result of your intuitive analysis.
(e) Discuss the difference between your estimate of the posterior, and the
Gaussian produced by the EKF. How significant are those differences?
What can be changed to make the approximation more accurate? What
would have happened if the initial orientation had been known, but
not the robot's y-coordinate?
AI-Generated Solution
AI-generated content may present inaccurate or offensive content that does not represent bartleby’s views.
Unlock instant AI solutions
Tap the button
to generate a solution
Recommended textbooks for you
Algebra & Trigonometry with Analytic Geometry
Algebra
ISBN:
9781133382119
Author:
Swokowski
Publisher:
Cengage
Algebra & Trigonometry with Analytic Geometry
Algebra
ISBN:
9781133382119
Author:
Swokowski
Publisher:
Cengage