Estimate position from inertial data

Question

Swapnil Sayan Saha 2020-1-14

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/500172-estimate-position-from-inertial-data

评论： Swapnil Sayan Saha 2022-5-12

Can someone provide me an example of how kalman filters can be used to estimate position of an object from 6DOF/9DOF IMU data. All examples I have seen just seem to find orientation of the object using ahrs/imufilter.

2 个评论
显示无隐藏无

BOUCHILAOUN Yacine 2022-5-5

Hello

I have a question if you don't mind, I use data from an IMU sensor to determine positioning in the civil engineering field. I am using MATLAB code to do this. For the moment I don't have the right result compare by the real case.

Did you have a satisfactory result according to your studies?

Swapnil Sayan Saha 2022-5-12

Yes. https://www.researchgate.net/publication/360075622_TinyOdom_Hardware-Aware_Efficient_Neural_Inertial_Navigation

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Bhargavi Maganuru 2020-4-1

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/500172-estimate-position-from-inertial-data#answer_423353

Hi,

You can go through following links.

https://www.mathworks.com/help/fusion/inertial-sensor-fusion.html

https://www.mathworks.com/help/fusion/ref/imufilter-system-object.html#mw_8d99f490-b04c-46e5-b3ae-e34474c79714

https://www.mathworks.com/help/fusion/examples/estimate-orientation-through-inertial-sensor-fusion.html

3 个评论
显示 1更早的评论隐藏 1更早的评论

Bhargavi Maganuru 2020-4-5

You can find below links helpful!

https://www.mathworks.com/discovery/kalman-filter.html

https://www.mathworks.com/help/control/getstart/estimating-states-of-time-varying-systems-using-kalman-filters.html

Swapnil Sayan Saha 2020-7-16

https://www.mathworks.com/help/fusion/ref/insfiltermarg.html << This one to be more precise. Thanks.

请先登录，再进行评论。

Answer 2

Martin Seyr 2020-4-24

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/500172-estimate-position-from-inertial-data#answer_428068

hello,

you need to integrate the accelerometers if you want to calculate linear positions. this will be subject to quadratic error propagation over time, so it is necessary to periodically reset the integrator.

it works like this: you use the orientation calculated from the fusion algorithm (kalman filter or some other algorithm) to rotate locally measured accelerations into the world frame. then you subtract nominal gravity, then you integrate twice.

good luck,

Martin

7 个评论
显示 5更早的评论隐藏 5更早的评论

Linda Scarzanella 2020-7-8

在 MATLAB Online 中打开

Hi, I've tried to write a code as you explained. The porblem is that the values of displacement and velocities for the z are completely wrong. Any idea why? I'll attach here the code:

clc 
clear all
close all
% pos = [0 0 0];
delete(instrfindall);
%serialportlist("available")'
a = serial("COM7", "BaudRate", 9600);
gyroReadings = [0 0 0];
time = 0;
fopen(a);
g = 9.81
accelReadings = [0 0 -9.81];
Fs = 200;
tic;
for i=1:21
data = fscanf(a);
[ax ay az gx gy gz] = strread(data, '%f %f %f %f %f %f ','delimiter',';');
toc;
gx = gx*pi/180;
gy = gy*pi/180;
gz = gz*pi/180;
accelReadings = [accelReadings; ax ay az];
gyroReadings = [gyroReadings; gx gy gz];
time = [time; toc];
end
decim = 1
fuse = imufilter('SampleRate',Fs,'DecimationFactor',decim);
% [orientation,angularVelocity] = FUSE()
q = fuse(accelReadings,gyroReadings)
time_1 = (0:decim:size(accelReadings,1)-1)/Fs;
plot(time_1,eulerd(q,'ZYX','frame'))
title('Orientation Estimate')
legend('Z-axis', 'Y-axis', 'X-axis')
xlabel('Time (s)')
ylabel('Rotation (degrees)')
%convert Acceleration to World Ref Frame
rotatedAcceleration = rotatepoint(q,accelReadings)
AccelWithoutGravity = [rotatedAcceleration(:,1) rotatedAcceleration(:,2) rotatedAcceleration(:,3)-g];
fs = 200; % Sampling Rate
fc = 0.1/30;  % Cut off Frequency
order = 6; % 6th Order Filter
%%Filter  Acceleration Signals
[b1 a1] = butter(order,fc,'high');
accf=filtfilt(b1,a1,AccelWithoutGravity );
figure (2)
plot(time,accf,'r'); hold on
plot(time,AccelWithoutGravity)
xlabel('Time (sec)')
ylabel('Acceleration (mm/sec^2)')
%First Integration (Acceleration - Veloicty)
velocity=cumtrapz(time,accf);
figure (3)
plot(time,velocity)
xlabel('Time (sec)')
ylabel('Velocity (mm/sec)')
legend('Z-axis', 'Y-axis', 'X-axis')
%%Filter  Veloicty Signals
[b2 a2] = butter(order,fc,'high');
velf = filtfilt(b2,a2,velocity);
%%Second Integration   (Velocity - Displacement)
Displacement=cumtrapz(time, velf);
figure(4)
plot(time,Displacement)
xlabel('Time (sec)')
ylabel('Displacement (mm)');
legend('Z-axis', 'Y-axis', 'X-axis')
 

Andrea Gusmara 2020-9-1

HI linda , have you finally found the solution of the wrong position problem ? Beacause I found myself in the same problem for my thesis.

Swapnil Sayan Saha 2020-9-1

It's best to not dead-reckon in z axis. I have observed similar phenomena from real-world data. In my MS thesis I am discussing the implications of inaccurate z axis localization. Better option is to use a pressure sensor or some other sensor (e.g. GPS if available, acoustics, Radar etc.)

请先登录，再进行评论。

Estimate position from inertial data

2 个评论
显示无隐藏无

采纳的回答

3 个评论
显示 1更早的评论隐藏 1更早的评论

更多回答（1 个）

7 个评论
显示 5更早的评论隐藏 5更早的评论

另请参阅

类别

标签

产品

版本

Community Treasure Hunt

Estimate position from inertial data

2 个评论 显示 无隐藏 无

采纳的回答

3 个评论 显示 1更早的评论隐藏 1更早的评论

更多回答（1 个）

7 个评论 显示 5更早的评论隐藏 5更早的评论

另请参阅

类别

标签

产品

版本

Community Treasure Hunt

2 个评论
显示无隐藏无

3 个评论
显示 1更早的评论隐藏 1更早的评论

7 个评论
显示 5更早的评论隐藏 5更早的评论