Go to Post I may be an engineer, but I am not smarter than a 5th grader. - Josh Drake [more]
Go Back   Chief Delphi > CD-Media > White Papers
portal register members calendar search Today's Posts Mark Forums Read FAQ rules




New Motion Sensor

By: sailorjoe
New: 09-08-2018 05:48 PM
Updated: 09-08-2018 05:48 PM
Total downloads: 239 times

This paper describes how to make the Flow Breakout motion sensor from Crazyflie work with a RoboRio.

This white paper collapses about six months of work done by Team 4579 to get a new motion sensor, called the Flow Breakout Board from Crazyflie to work correctly on a RoboRio based robot. It describes the sensor, two key issues we had to overcome to make it work correctly, some math for correct usage, basic Java code for a class library and test robot program, and important links to other resources.

Attached Files

  • doc FlowSensorWhitePaper.doc


    downloaddownload file

    uploaded: 09-08-2018 05:48 PM
    filetype: doc
    filesize: 82kb
    downloads: 237

Recent Downloaders


view entire thread


09-09-2018 01:12 PM


Unread Re: paper: New Motion Sensor

This is pretty cool! I have a few questions and some comments.

Did you compare the measured displacement to any ground truth measurements? IE did you measure the accuracy of the reported accumulated displacement?

Are the accumulated displacement measurements repeatable? At one point I think the paper says that the reported distance varied with the speed. To what % did this affect the reported measurements?

You briefly mentioned that this has benefits over encoder based odometry, can you expand a bit more on that?

I would say label your register addresses, but then I looked at the Arduino library, then tried to find a register map, and realized that apparently it doesn't exist. Ah well.

Based on the register size, I think you have a theoretical resolution of (maximum speed) / 2^15, for ~0.006 in resolution at 183 in/s. I'm not sure however if the sensor itself can manage this resolution.

09-09-2018 02:23 PM


Unread Re: paper: New Motion Sensor

Great work so far, thanks for sharing! I too would be interested in the accuracy of this sensor compared to standard encoder odometery - especially when the robot is rotating.

Bitcraze_PMW3901::initRegisters(){ ... }
This is what some in the industry like to call "Job security".

09-18-2018 06:33 PM


Unread Re: paper: New Motion Sensor

Thanks for your questions, Nick.
To date we have not calibrated and tested the sensorís repeatability over long distances on the robot. Resolution is a function of height since the sensor has a fixed number of pixels in its viewing window, but neither the number of pixels nor a resolution specification is published. You can see in the paper what we found experimentally.
You can see the short range repeatability numbers at the Bitcraze Forum links in the paper. I sent them a spreadsheet with graphs of the results. As I recall, it was a bit better than 10%, but I think that can be improved with filtering. There is a graph showing the variation with speed, as well.
I think the benefit over encoder sensors are two-fold. One, easier mounting because itís just one sensor, it mounts anywhere with a small white LED, its small. Two, itís not subject to wheel slippage, the bane of encoders.
Iíll try to post my test results spreadsheet tonight or tomorrow, to make it easier to find.
Right now we are doing our recruiting drive and not testing just yet, but in a few weeks we should have more data on full scale performance.

09-18-2018 06:34 PM


Unread Re: paper: New Motion Sensor

Ha! Yes, maybe, but eventually it will all leak out.

view entire thread



loading ...

All times are GMT -5. The time now is 04:47 AM.

The Chief Delphi Forums are sponsored by Innovation First International, Inc.

Powered by vBulletin®
Copyright ©2000 - 2018, Jelsoft Enterprises Ltd.
Copyright © Chief Delphi