Sale!

MP4: IMU PDR

$30.00

Category:
Rate this product

ECE/CS 434 | MP4: IMU PDR

Objective
In this MP, you will:

Implement a step counting algorithm using accelerometer data.
Apply signal processing and linear algebra functions such as low/high pass filtering, peak detection, fft, etc. to the step counting algorithm.
Calculate phone orientation using a single static accelerometer reading.
Track phone orientation through a sequence of gyroscope data by performing integration.
Problem Overview
In pedestrian dead-reckoning applications, two pieces of information need to be tracked: how far a user walked, and the direction of the walk. In the first part of this MP, you will write a step counter using accelerometer data as input. In the second part, you will derive the initial orientation of the phone using a single accelerometer reading and calculate the final orientation using a sequence of gyroscope data.

Imports & Setup
Installing requirements correctly
First. we will make sure that the correct versions of required modules are installed. This ensures that your local Python environment is consistent with the one running on the Gradescope autograder. Just convert the following cell to code and run:

Note: It’s preferred that your local environment matches the autograder to prevent possible inconsistencies. However, if you’re running into annoying Python version issues but haven’t had any issues getting consistent results on the autograder, there is no need to stress over it. Just skip for now and come back when you do encounter inconsistencies:) Ditto below.
WARNING: ENSURE THE FOLLOWING CELL IS MARKDOWN OR DELETED BEFORE SUBMITTING. THE AUTOGRADER WILL FAIL
if name == ‘main’: import sys !{sys.executable} -m pip install -r requirements.txt

Your imports
Write your import statements below. If Gradescope reports an error and you believe it is due to an unsupported import, check with the TA to see if it could be added.

import numpy as np
import pandas as pd

# This function is used to format test results. You don’t need to touch it.
def display_table(data):
from IPython.display import HTML, display

html = “<table>”
for row in data:
html += “<tr>”
for field in row:
html += “<td><h4>{}</h4><td>”.format(field)
html += “</tr>”
html += “</table>”
display(HTML(html))
Sanity-check
Running the following code block verifies that the correct module versions are indeed being used.

Try restarting the Python kernel (or Jupyter) if there is a mismatch even after intalling the correct version. This might happen because Python’s import statement does not reload already-loaded modules even if they are updated.

if __name__ == ‘__main__’:
from IPython.display import display, HTML

def printc(text, color):
display(HTML(“<text style=’color:{};weight:700;’>{}</text>”.format(color, text)))

_requirements = [r.split(“==”) for r in open(
“requirements.txt”, “r”).read().split(“\n”)]

import sys
for (module, expected_version) in _requirements:
try:
if sys.modules[module].__version__ != expected_version:
printc(“[✕] {} version should to be {}, but {} is installed.”.format(
module, expected_version, sys.modules[module].__version__), “#f44336”)
else:
printc(“[✓] {} version {} is correct.”.format(
module, expected_version), “#4caf50”)
except:
printc(“[–] {} is not imported, skipping version check.”.format(
module), “#03a9f4”)
Part 1. Step Counter
We have provided you with smartphone accelerometer data collected under three circumstances

walking with phone in pant pocket
walking with phone held in the hand statically as if the user is looking at it while walking
walking with phone in hand and the hand swinging
For each file, there are three columns, representing the accelerometer readings in three local axes(unit:  m/s2 ). The accelerometer is sampled at 100Hz.
Implement your algorithm in the count_steps(walk_accl_file) function below. Do NOT change the function signature. You are, however, free to define and use helper functions. You are expected to use common signal processing and linear algebra functions (e.g., high/low pass filtering, convolution, cross correllation, peak detection, fft etc.)

# This function takes 1 argument:
#     walk_accl_file  (string) – name of data file for accelerometer data
# It returns an integer, the number of steps
def count_steps(walk_accl_file):
# Your implementation starts here:

return 0
Run & Test
Use the cell below to run and test count_steps(walk_accl_file).

def estimate_steps_score(calculated, expected):
delta = abs(calculated – expected)
return 1 if(delta <= 2) else max((1 – abs(delta – 2) / expected), 0)

if __name__ == ‘__main__’:
walk_accl_files = [‘data/holdstatic_20steps.csv’, ‘data/inpocket_26steps.csv’,
‘data/inpocket_36steps.csv’, ‘data/swing_32steps.csv’, ‘data/swing_38steps.csv’]
groundtruth = [20, 26, 36, 32, 38]
output = [[‘Dataset’, ‘Expected Output’, ‘Your Output’, ‘Grade’]]
for i in range(len(groundtruth)):
calculated = count_steps(walk_accl_files[i])
score = estimate_steps_score(calculated, groundtruth[i])
output.append([walk_accl_files[i], groundtruth[i],
calculated, “{:2.2f} / 5.00”.format(score * 5)])
output.append([‘<i>👻 Hidden test 1 👻</i>’,'<i>???</i>’, ‘<i>???</i>’, ‘<i>???</i> / 15.00’])
output.append([‘<i>…</i>’, ‘<i>…</i>’, ‘<i>…</i>’, ‘<i>…</i>’])
output.append([‘<i>👻 Hidden test 5 👻</i>’,'<i>???</i>’, ‘<i>???</i>’, ‘<i>???</i> / 15.00′])
display_table(output)
Part 2. Orientation Tracking
Part 2.1 Initial Orientation Calculation
Assume the phone is static at the initial moment. We will provide you with the accelerometer reading at that moment (unit:  m/s2 ). Your goal is to identify the initial phone orientation from this reading. We will not provide compass data here since all the data are collected indoor and compass won’t give an accurate north indoor. Instead, assume at the initial moment, the projection of the phone’s local Y axis onto the horizontal plane is pointing towards the global Y axis. This will also give a fixed phone initial orientation.

We expect you to output the global direction in which the phone’s local X axis is pointing at.

Hint: Find the global Y axis’s direction in the local frame and let this direction be a 3 × 1 vector  v1 . Let the gravity in the local frame be another 3 × 1 vector  v2 . Then essentially you need to solve the following equation:
R[v1v2]=⎡⎣⎢010001⎤⎦⎥
Part 2.2 3D Orientation Tracking
In this part, you need to take the initial orientation calculated in part 1, and perform gyro integration for each timestamp onward. We will provide you with a trace of gyroscope data, in CSV format. There are three columns in the file, representing the gyroscope readings in three local axes (unit:  rad/s ). The gyroscope is sampled at 100Hz. Your task is to track the phone’s 3D orientation and output the end direction in which the phone’s local X axis is pointing at in the global frame.

One way of solving this problem can be:

Assume the gyroscope’s sample interval is  Δt .
Get the phone’s instant rotation axis and rotation angle in the local frame  (l⃗ ,Δθ)  for each time stamp  ti,  where  l⃗ =(ωx,ωv,ωz)  and  Δθ=(ω2x+ω2v+ω2z)−−−−−−−−−−−−√⋅Δt
Project the instant rotation axis  l⃗   into the global frame using the phone’s  3D  orientation matrix  Ri  at time  ti .
Convert the instant rotation axis and angle in global frame into the form of rotation matrix  ΔRi .
Find the total 3D rotation matrix for time  ti+1:Ri+1=ΔRi⋅Ri
Implement both algorithms in track_orientation(orientation_accl_file, gyro_file) below. This is because the initial rotation matrix needed for calculating final orientation is a by-product of calculating initial orientation. Do NOT change the function signature. You are, however, free to define and use helper functions.

# This function takes 2 arguments:
#     – orientation_accl_file (string) – name of file containing a single accl reading
#     – gyro_file (string) – name of file containing a sequence of gyroscope data
# It returns two arguments: an array representing the initial global direction
# in which the phone’s local X axis is pointing at, and the final.
def track_orientation(orientation_accl_file, gyro_file):
# Your implementation starts here:

return [
[0.0, 0.0, 1.0],
[0.0, 0.0, 1.0],
]  # [initial orientation], [final orientation]
Run & Test
Use the cell below to run and test Part 2.

def get_deviation(calculated, expected):
calculated = np.array(calculated)
expected = np.array(expected)
with np.errstate(divide=’ignore’, invalid=’ignore’):
dot_prod = np.dot(calculated, expected) / \
np.linalg.norm(calculated) / np.linalg.norm(expected)
return np.degrees(np.arccos(dot_prod))

if __name__ == ‘__main__’:
gt_init = [0.9999, -0.0020, 0.0120]
gt_final = [-0.0353, 0.9993, 0.0076]
stu_init, stu_final = track_orientation(
‘data/orientation_accl.csv’, ‘data/gyro.csv’)

output = [[‘Test’, ‘Dataset’, ‘Expected Output’,
‘Your Output’, ‘Deviation’, ‘Result’, ‘Grade’]]
init_state = ‘FAILED’
final_state = ‘FAILED’
init_grade = 0
final_grade = 0
init_dev = get_deviation(stu_init, gt_init)
final_dev = get_deviation(stu_final, gt_final)
if(init_dev < 2):
init_state = ‘PASSED’
init_grade = 10
if(final_dev < 2):
final_state = ‘PASSED’
final_grade = 10
output.append([‘Initial Orientation’,
‘orientation_accl.csv, gyro.csv’, gt_init, stu_init, “{:2.2f}°”.format(init_dev), init_state, “{} / 10”.format(init_grade)])
output.append([‘Final Orientation’, ‘orientation_accl.csv, gyro.csv’,
gt_final, stu_final, “{:2.2f}°”.format(final_dev), final_state, “{} / 10”.format(final_grade)])
output.append([‘<i>👻 Hidden test 1 👻</i>’,'<i>???</i>’, ‘<i>???</i>’, ‘<i>???</i>’, ‘<i>???</i>’, ‘<i>???</i>’, ‘<i>???</i> / 10’])
output.append([‘<i>…</i>’, ‘<i>…</i>’, ‘<i>…</i>’, ‘<i>…</i>’, ‘<i>…</i>’, ‘<i>…</i>’, ‘<i>…</i>’])
output.append([‘<i>👻 Hidden test 4 👻</i>’,'<i>???</i>’, ‘<i>???</i>’, ‘<i>???</i>’, ‘<i>???</i>’, ‘<i>???</i>’, ‘<i>???</i> / 10’])
display_table(output)
Rubric
Step Counting (100 points)
You will be graded on the 5 sets of provided data (5 points each), as well as 5 sets of hidden data (15 points each). For each test case, the grade depends on how much the result deviates from the groudtruth. A 2-step error for the provided data is tolerated. A 4-step error for the hidden data is tolerated. For results greater than the error threshold, your score will be scaled proportionally.

Orientation Tracking (100 points)
You will be graded on the provided data as well as 4 addition sets of data. They are each worth 20 points. A 2-degree error is tolerated. For results greater than the error threshold, no points will be rewarded since we provided a detailed algorithm to follow. The test data also include the simple case where the phone’s initial local frame is aligned with the global frame, and phone will only rotate along Z axis onwards. (In case you find the MP too difficult, only doing 1D integration on Z axis should at least give you some points.)

Submission Guideline
This Jupyter notebook is the only file you need to submit on Gradescope. If you are working in a pair, make sure your partner is correctly added on Gradescope and that both of your names are filled in at the top of this file.

Make sure any code you added to this notebook, except for import statements, is either in a function or guarded by __main__(which won’t be run by the autograder). Gradescope will give you immediate feedback using the provided test cases. It is your responsibility to check the output before the deadline to ensure your submission runs with the autograder.

MP4: IMU PDR
$30.00
Open chat
Need help?
Hello
Can we help?