Paper: Surrogate Modelling for Oxygen Uptake Prediction Using LSTM Neural Network

The article can be accessed through mdpi

Test command

rosrun vectornav vn200_node _serial_port:=/dev/ttyAMA0 _target_baud:=460800 _binary_data_output_port:=2 _binary_ins_data_output_rate:=400
gdb rosrun vectornav vn200_node _serial_port:=/dev/ttyAMA0 _target_baud:=460800 _binary_data_output_port:=2 _binary_ins_data_output_rate:=400
rostopic echo /vectornav/imugps
rostopic hz /vectornav/imugps
rostopic info /vectornav/imugps
ps aux
sudo killall python
sudo killall screen
sudo killall tee
rostopic echo --filter=m.header.seq%200==0 /vectornav/imugps

Source code for devices measuring in the campaign

TC219 CPU info

Device name	HTC219-720-SPC
Full device name	HTC219-720-SPC.ad.tuni.fi
Processor	Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz   3.41 GHz
Installed RAM	32,0 GB
Device ID	bbbf3aca-47cd-4ad8-9fd0-ad23414388c6
Product ID	00329-00000-00003-AA689
System type	64-bit operating system, x64-based processor
Pen and touch	No pen or touch input is available for this display

Usage

Matlab instruction

Test repo for matab: D:\(...BackupDdrive)\(..RA)\....PAVEL\MATLAB-APP\test

• In matlab folder, create folder name ${subjectID}_out, for example O1_out

• The directory should include these files:

  • 2022-XX-XX-XX-XX-XX_s.txt (sliced ID's gaitpod file)
  • OXYGAX_marker_XXXX_s.txt(sliced ID'oxygen file)
  • run.m executable file run the everything
  • features_OK.m
  • velSeg.m
  • input_features.m
  • trainingPartitions.m

• At the begging run.mat, change subjectID and change sortedInputFeatures to true if want output input features in ordet (false by default) and run the file

• After about 5 mins it will produce input feature file

This section describes the basic usage of the device

Connecting the devices

• The Xsens IMU units are currently connected through the USB ports and should be connected before powering the device.
• GPS is connected to the UART pins. The device is enumerated as /dev/ttyAMA0 on RPi and /dev/ttyS4 on minnowboard and uses 57600 baud. This all is configured in the ROS launch files.
• It is recommended that the ethernet cable is plugged to a router which will assign an IP address to the device (unless using 3G), currently the OS has issues if you switch to a different router while the device is powered?
• The device gets time from internet, so logs are dated to 1970 or some other wrong time if no internet connection has been available before starting measurements

Powering

• The device is powered by the powerbank from the USB connector. The runtime with this seems to be around 3-4 h for RPi and 2,5 h for minnowboard.

Measurements

• The measurement mode is activated in the boot
• Primary led will flash when time is corrected from interney
• The device will start recording as soon as it loses internet connectivity
• It might take a few minutes to start the measurements, if there are errors with gps or imu. Device will try again until everything works.
• Primary led will flash when making IMU alignment, so IMU should be stationary that time and a bit after.
• It seems, the IMU wont start if it has been stationary for some time (MTw wired), so light shake might be needed before use.
• When recording, primary led will be lit.
• Recording will be stopped once the internet connectivity is restored.
• After being stopped, the files will be uploaded to Google Drive ([email protected])

Shutting down

• The power switch on top can be flipped to give shutdown commands to the unit.
• After flipping the switch and waiting some seconds, the USB cable powering the device can be unplugged.

Connecting to the device

Standard SSH conenction can be made to the device with ssh:

ssh ubuntu@[IP]
Password: ubuntu

For RPi with kinetic user: pi, password: raspberry For minnowboard user: openkin, password: openkin

Libraries utilized

ROS Driver for XSens MT/MTi/MTi-G devices.

See: http://ros.org/wiki/ethzasl_xsens_driver

ROS driver for ublox

See: https://github.com/KumarRobotics/ublox

Pozyx Arduino library

See: https://github.com/pozyxLabs/Pozyx-Arduino-library

All are added as subtrees to this repo, so updates can be fetched with commands (from git root):

git subtree pull --prefix nodes/ublox https://github.com/KumarRobotics/ublox.git master --squash

,

git subtree pull --prefix nodes/ethzasl_xsens_driver https://github.com/ethz-asl/ethzasl_xsens_driver.git master --squash

and

git subtree pull --prefix nodes/pozyx/include/Pozyx-custom-library/ https://github.com/pozyxLabs/Pozyx-Arduino-library.git master --squash
cd nodes/pozyx/ínclude
patch -p0 < pozyx-arduino-to-linux.patch

Motion Fusion

http://msthub.ain.rd.tut.fi/mobilemachines/motion_fusion/blob/master/src/KalmanImu.py

Script startup

The measurements are started by running ~/openkin/start_script.sh, which in turn calls the main script at ~/openkin/measurement.sh using locks. The script now starts from /etc/rc.local, with lockfile at /home/ubuntu/measurement.lock and as user ubuntu (or pi or openkin).

Old documentation...

... can be found in Google Docs with the [email protected] -account.

For latest documentation, please look at doc/