Greetings everyone,
Digest for the first week of GSoC 2015 coding for the BeagleSat project.
Worked on:
- Hardware setup and configuration
- Basic readouts of MPU9250 via SPI
- Testing currently available drivers and available code
- Looking into what can be reused and what should be redone
- Final touches on the presentation + Intro video (awaiting upload when the VPN is stable)
- Blog space ready for tracking progress
- Received the PNI RM3100 modules via Fedex
Issues:
- The Great Firewall of China (see intro video above)
-
Hackaday.io page still pending
To be worked on:
- Getting a cohesive setup for the sensors
- Interfacing with the PNI RM3100 magnetometers
- Looking into the parameter computation ML algorithms (and how to optimally implement them in the coming weeks)
- PRU viability will be considered
-
Hackaday.io page + more blogging
Thanks,
Niko
Hi everyone,
Second week of GSoC 2015 coding for the BeagleSat project.
Worked on:
- Hardware setup mostly done
- Finished layout graphic and added it to video and some other minor edits
- Readouts from PNI RM3100 via SPI working, more testing this week
- Writing code for both sensors to measure and store a workable amount of data for ML estimator
Issues:
- Advice anyone: Recommendations for C/C++ ML library (or any optimized computational library for minimizations), links are awesome to
- Finals week, eg. a lot of distractions
- Uploading video to Youtube, if it doesn’t work until tomorrow, I’ll upload it via non google services or send it to someone
To be worked on:
- Drivers and data collector code for the sensors
- ML stuff, have to start implementing it (test with Octave, later implement in C++)
-
Hackaday.io page + more blogging (still a bit to do)
Cheers,
Niko
Forgot to link to last week’s update.
It lives here: http://nvisnjic.com/2015/06/16/gsoc-update-week3.html
C/P from the link:
Week three of the GSoC 2015 BeagleSat project, a future CubeSat development platform.
Worked on:
Issues:
- Soon to begin testing data in Octave. A possible issue is that the data should come from rotating the sensor around a stable axis, which would require building an apparatus to hold the sensors while rotating in a stable orbit. If it turns out to be a blocker, I’ll look into 3D printing such a device and rigging it to the BeagleBone
- Last weeks ML library issue should be a non issue, I’ll use a matrix computation library when porting the code to C++, it should suffice
To be worked on:
- Currently using SPIDEV for the data, transfer to drivers later in the project
- More Octave stuff in the coming week
-
Hackaday.io page soon
This week’s update (C/P from blog)
Week four is upon us, more stuff has been done, more stuff to be done.
Finished the 2D correction algorithm in MATLAB, more about that here.
Worked on:
- 2D Correction algorithm in MATLAB, works like a charm
- Parameter estimator is very fast locally, will be porting to Octave to test execution speed on the BeagleBone
- Parameter estimation on simulated data yields less than 1% error on all deformations except rotation (rotation is variable in error, but less than 2%)
Issues:
- Research was helpful, but the algorithm part could have been a bit more elaborate
- Octave was super uncooperative with the symbolic toolbox, have to look i into that and move the MATLAB code to Octave (go team open-source!)
- Sensor testing complexity still unknown (last week’s issue), will be testing in the coming weeks
To be worked on:
- Expanding the 2D correction algorithm to a 3-axis one
- Adding time-variant error compensation using telemetry and additional inputs
- Currently using SPIDEV for the data readouts, transfer to proper drivers later in the project
Weekly progress up at: http://nvisnjic.com/2015/06/30/gsoc-update-week5.html
Pasting for the no-links crowd 
Last weeks report (forgot to post it here) → http://nvisnjic.com/2015/07/16/gsoc-update-week7.html
Summary: