First post on the Forum.
I used Ardupilot with my APM2.x while back.
I stayed away for few years. More than I wanted.
I see that Ardupilot supports a limited amount of flight controllers and these does not seem to get updated as time goes by as other flight stacks do. I wonder the rational on that.
If you provide the HAL requirements for flight controllers manufacturers, we could have many more options with performance going up and cost increasingly going down.
For example, today one can get light flight controllers based "powerful ARM: microcontrollers e.g., F7. for ~ US30.00.
I know that this seems too simple. You would have done already.
Why is the support for affordable (US$ 30~40) flight controller for example based on F4 / F7 is not available?
Is the portability of the code primarily based on cross compiling and changing the I/O classes/functions?
I think that we already support numerous flight controler types! We got some STM32, ARM SoC, intel chip, FPGA, dsp !
There isn’t much requirement outside IMU, compass and a memory.
The main blocking point is the porting time as plenty think much be checked to ensure good flight condition. And generally cheap control cannot unsure that ! Having F7 horse power is useless if the other hardware are craps …
A good step will be the end of the dependency on the PX4 framework, after that we should allow easier portage, I believe.
The Ardupilot codebase is very portable, as it sits on an abstraction layer with an operating system underneath. That construct has enormous benefits but requires a bit of flash memory beyond what bare-metal code does (in general). The stm32f745 is “faster” than the f427 currently used, but only has half the flash memory. There is a project to evaluate chibios as an alternative to nuttx, for stm32 targets, which would result in a lighter weight rtos, and smaller flash requirement. If that works out, the f745 (and a bunch of other mcu’s) may be more feasible targets.
Thanks for the replies.
I appreciate your attention specially for question you may have to answer periodically.
I have a nice F7 Arm development boards with lots of flash that I used on my embedded systems class.
1M and 512K flash. I can run few RTOS on it.
Not that would not matter much unless someone has a flight controller around it. With accelerometer, compass, barometer on it right?
Are you saying that the flight controllers around the STM32F7 out there does not come with enough flash to have a RTOS and ArduPilot on it? I was expecting the other way around as these SOCs get more memory.
Sorry on the delay replying. I am on the road this week.
I think that we already support numerous flight controller types! We got some STM32, ARM SoC, intel chip, FPGA, dsp !
On my searches I could not numerous flight controllers.
Here is what I found on the ArduPilot website.
Selecting the right board depends on the physical restraints of the vehicle and the applications that you want to run. Broadly speaking:
Pixhawk is highly recommended for general use.
Pixracer is recommended for small frames that require no more than 6 PWM outputs.
NAVIO+ or Erle-Brain Linux Autopilots should be considered for UAV Vision applications.
I need to buy ~ 10 flight controllers for a class. I would like to find something newer and more affordable than Pixhawk…
Would you be able to point me to where I can find the several flight controller that you mentioned?
I can find several new hardware that runs Betaflight but they don’t mention Ardupilot.
I am interested in ground mobile robots. I will be connecting the flight controller with ROS eventually. Initially plain Ardupilot.
Problem maybe solved.
Someone is working on Revo board support that will open up lots of cheap AIO boards. It is already working. Mostly being tested on Plane and Copter though.
The comments on the pr associated with that thread are perhaps worth a read.
At the moment v2 builds assume 1MB flash, the v3 and v4 builds assume 2MB. The F745 can hold Ardupilot, but imho isn’t really much of a step up from the F4. Advanced features like smartRTL need flash more than clock cycles.
I had not caught the August onwards comments! Thanks for that.
Here is an idea for consideration.
I may be able to propose senior ECE Capstone projects for students at a world-class university in the US. In fact, this university was ranked recently as a #1 public university in the world.
If you like the idea, I would like to start collecting requirement to design our own robot/flight embedded controller. Assuming the SOC exists, we can design around it. Then we can send for fabrication a batch of controllers.
I have requested to be in the queue for the capstone projects for the winter quarter that starts in Jan 2018.
We have time, but not much. As you know, a good project specification takes a while to be done correctly.
What do you think?
Can we start a Thread on that?
I can setup an on-line tool for us to cooperate in the project.
I am assuming the hardware will be made available as open hardware open source. Then eventually a company will copy over for their mass production. We can probably force the cost to stay low keeping the design open and making our own batches in a place that is fully automated.
I’m really not an expert, but love the idea.
@proficnc and @apache405 are hardware gurus and @tridge has a lot of experience with academia, so if we’re lucky they might be able to give some pointers (give them time: they’re spread across the globe at present).
Philip made some new hardware announcements at the ArduPilot Unconference in China today that are relevant to this discussion. I imagine they’ll be made public sometime soon.
Thanks, I am on the road myself this week. Chile and England. Back to the US this coming week.
Looking forward further discussions.
There is an F7 based Pixhawk 2.1/Cube coming out soon. Right now the specs are not public information, but they will be released soon.
I sent @JackSilb a direct email about this and some other topics you mentioned above.