EECS 465/ROB 422 Homework Assignment  5 solved

Original price was: $35.00.Current price is: $28.00.



5/5 - (1 vote)


1. (10 points) Consider the binary random variable A. Starting from P(a ∨ ¬a) = 1, use the axioms of
probability to derive P(¬a) = 1 − P(a).
2. (10 points) AI Book Chapter 13 #13.8
3. (10 points) AI Book Chapter 13 #13.16
4. (10 points) AI Book Chapter 14 #14.5 part a
5. (15 points) AI Book Chapter 14 #14.6
6. (10 points) AI Book Chapter 15 #15.2
7. (15 points) AI Book Chapter 15 #15.13
8. (10 points) Describe the advantages and disadvantages of a Particle filter vs. an Extended Kalman
Filter vs. an Unscented Kalman Filter. Specifically, consider the computational cost, the ability to
handle different kinds of dynamics, and the type of state uncertainty distributions each can handle.


1. Download and unzip Run You should see some lines and points plotted. If
you do not see this, make sure matplotlib is installed.


The following implementation problem should be done in python starting from the provided templates.
Only edit what is inside the ### YOUR CODE/IMPORTS HERE ### block. Feel free to look around the
internet for example code for reference, but you should implement your own code from scratch unless
explicitly stated otherwise. Include all code you write in your zip file.

1. In this problem we will implement a Kalman filter given the motion model, sensor model, and data.
Consider a robot with the following motion model:
xt+1 = xt + 1.5u1 + 0.1u2 + ζx
yt+1 = yt + 0.2u1 − 0.5u2 + ζy
and sensor model:
z1 = 1.05x + 0.01y + δ1
z2 = 0.01x + 0.90y + δ2

a (10 points) To use a Kalman filter, we first need to specify the models in the standard form:
xt = Axt−1 + But + ζt
zt = Cxt + δt
where ζ is the motion noise and δ is the sensor noise. For this form, specify x (the state), u (the
controls), and the matrices A, B, and C from the equations above in your pdf. Input your A, B,
and C matrices in

b (15 points) We also need to specify the matrices R and Q, the motion and sensor noise covariances,
respectively. To do this, we will compute the covariances from data. We will assume that both
the motion and sensor noise are zero-mean and normally-distributed. Write code in
to compute the covariances.

This script loads in the data file kfdata.dat, which contains the
measurements, ground truth states, and actions. loads in the A, B, and C matrices
from, so make sure you have completed part (a) before you do this part. Include
what you compute for R and Q in your pdf.

c (30 points) Now we are ready to write the Kalman Filter. Open and carefully read through
the code to understand what it is doing. When you run python, you will see a plot with
the following:
• Blue line: ground truth state trajectory
• Blue dots: measurements
• Red line: estimate state trajectory (the output of your Kalman Filter)
• Black ellipses: covariance plotted as an ellipse at 2σ. The covariance is only drawn at every
3rd state so the plot doesn’t get too cluttered

It will also print out the total error, which is the sum of the magnitudes of the errors between your
estimated states and the ground truth states. loads in the A, B,C,R, and Q matrices from, so make sure you complete parts
(a) and (b) before doing this part.

You will fill in the KalmanFilter function in so that it outputs the estimated mean and
covariance after receiving a new measurement and action. Your Kalman filter should not use
the ground truth states in any way, those are only used to compute the total error.

If you have
completed all the parts correctly, your output should look similar to the figure below.
The total error should be 22 ± 1. Include a screenshot of your output and your total error in your
pdf. We will run your code to verify it. If the code crashes, you will not receive credit for this part
of the assignment.

3 EECS 465/ROB 422 Homework Assignment  5