This group is for those who are interested in the challenges and benefits of pure [1] Linux-based Multirotors.
[1] "pure" means: the whole control loop required for flight stabilization is carried out in a high-priority Linux user space task.
Multicore FCU based on Parallela
Hello guys,I am new in this forum but I have been flying multi copters for 2 years now and I was wondering if anybody had looked into using this wonderful board called the Parallela board. This would provide a very flexibly solution for a flight controller unit like:redundancy: use say 3 cores for orientation reconfigurability: allocate cores to visual processing tasks, navigation tasks, trajectory optimization, mission planning etc.energy efficiency: use only core you requireIs there something…
Read more…
Comments
FlytPOD runs linux ros based FlytOS. See https://igg.me/at/flytpod/x/14288731
https://www.youtube.com/watch?v=aUUj03Tm9LQ
dear all,
is there any way or APIs to make a process(ex: AHRS loop) 100% use in one of the CPU cores without OS time slicing? is that possible?
Thanks
Ubuntu-based drone delivery. With 4Gmetry (4gmetry.voltarobots.com)and Volta (diydrones.com/group/volta)
We want to develop a fully autonomous octocopter which uses computer vision for a fully automated operation, We have a patent which defines the conceptual framework and are now in the process of deciding what would be the best platform. In principle we were thinking on havina a linux system sitting on top of a microcontroller Arducopter or alike system, the Linux system executing the top level intelligent control functions and commanding the autopilot through communications. Now we see the possibility to have all together on a signle Linux machine. What do you think would be the best architecture?
An elevation map flight achieved by using different programming languages (C/Python) on a Linux-based UAV controller: http://diydrones.com/profiles/blogs/quadrotor-srtm-elevation-map-al...
GDAL is used to load and transform the SRTM data from GeoTiff.
Some impressions from our Raspberry Pi copter platform: http://diydrones.com/xn/detail/705844%3ABlogPost%3A1733275
For those of you who are interested in fast Python MAVLink serialization: https://github.com/PenguPilot/swig-mavlink
These results have been acquired with PREEMPT. The UARTS can generate quite a lot of load, but this is hardware-/driver-dependent.
- We see nearly no load on the Raspberry Pi
- Gumstix, however, creates a lot of load
Tobi,
with the PREEMPT option it is somewhat better with regards to jitter on the time axis.
but you are going to kill linux on heavy interrupt loads (multiple uarts).
on you flying monster you are having way to much cpu power :-)
also when looking at arm you have to carefully verfiy the operations of the internal driver. they may not really utilize the dma for example.
all uart drivers i have seen so far have a buffer size defined which is a fraction of the real physical hw.
Robert, is this your experience with Linux?
An example investigation: For a 200Hz control loop, in which a MPU6050 and a HMC5883L are read synchronously via 100KHz I2C and PWM's are set via ioctrl, I get the following dt measurements:
Minimum | 1st Quartile | Median | Mean | 3rd Quartile | Maximum
0.004578 | 0.004913 | 0.005005 | 0.005000 | 0.005096 | 0.005402
The main impact on these timings is caused by two reader tasks, which concurrently access the same 100KHz I2C bus for reading the MS5611 and I2CXL sensors at a lower rate (appoximately 10 Hz). Besides concurrent I2C access, serial ports are read for GPS and remote control input.
We have recorded the time deltas at the full rate of 200Hz and sent them via ZeroMQ to a low-priority python task, which writes the logfile to sd card.
-
1
-
2
of 2 Next