Here are the 5 entities supporting the BeaglePilot project (a.k.a. push the AP_HAL_Linux of ardupilot):
- BeagleBoard.org
- 3D Robotics
- ardupilot
- Erle Robotics
- Google (through the Google Summer of Code)
To this we should add many developers doing a great work. All of it on the behalf of the Open Source Drone community (btw, anyone heard something else about the DroneCode Foundation?).
So how far are we on the road? Well, in the last weeks we've had good progress. We fixed several issues with the SPI bus managing to get better data form the sensors. We even compared it with the APM2.
It seems we have good enough data to run it on a Rover (and maybe on a plane) but it seems like it's still no good for a copter.
https://www.youtube.com/watch?v=QoPwKTos7B4
We will keep pushing it forward. Thanks to everyone for the support and interest.
Edit: Thankfully, i was pointed out of my mistake speaking solely of the initial 4 entities. Obviously the ardupilot team is one of the most important component. I sincerely hope that no one got offended and if so, my apologies.
Comments
Thanks for the reply, I'll be following your work.
Hola Winston, gracias por tus amables palabras.
El MPU9150 es un IMU de InvenSense que utilizando los algoritmos de ardupilot se acaba convirtiendo en un AHRS. En BeaglePilot estamos trabajando con varios IMUs, el MPU9150 sea probablemente el menos reliable de ellos.
En cuanto a tu sugerencia, en primer lugar, todavía, en BeaglePilot estamos trabajando es aspectos de bajo nivel. Ideas como la tuya deberán llegar en fases posteriores. De cualquier modo, lo que sugieres a grandes rasgos es tratar de hacer odometría utilizando únicamente la información del IMU. Es una idea bonita y me encantaría verla hecha realidad pero me temo que los sensores con los que trabajamos no tienen la precisión necesaria para que esto se pueda llevar acabo.
Si revisas la literatura, verás que esta clase de comportamientos se están empezando a implementar utilizando cámaras estereoscópicas (como el Kinect) de modo que el vehículo tenga información 3D de su entorno y esto le permita estimar el desplazamiento y corregirlo posteriormente.
Interesante idea sin duda! Si no es molestia te pediría que si deseas continuar con la discusión lo hagamos en inglés para que muchos y muchas más puedan participar en ella.
----
Briefly what Winston was asking is whether it was possible to make odometry using the IMU values. I replied that the sensors we use are not accurate enough (not actually sure if such an accurate sensor does exist). Also I informed that nowadays, stereoscopy-state of the art techniques are being used for this matter. This is an interesting topic and i'm sure many of you are much more experienced than I am. Looking forward to hear some other insights.
@Victor Mayoral, buenos días y gracias por el gran aporte que has estado haciendo a la comunidad, desde el lanzamiento del iPhone 5S con su sensor de Movimiento (Co_Processor M7) he venido haciéndome la idea de si podrías integrar un sensor como ese en el controlador de vuelo que usted esta desarrollando (no estoy claro de si el sensor MPU9150 hará lo que imagino) y la idea radica en que ese sensor registre el desplazamiento ocasionado por la fuerza que ejerce el viento sobre el Drone en "Stabilized mode" obligándolo a desplazarse en una dirección, de manera tal que pueda medir la cantidad desplazada en un eje y contrarrestar esa fuerza para volver a su posición inicial, seria como el position hold usando un GPS, en este caso será un con el Sensor..
@Gary thanks for your kind words. Indeed the challenges are big but that makes it just more fun :).
@Swift, thanks for the suggestion. Indeed we could do what you suggest. Actually there's already an MPU9150 driver coded that i did over the first weeks so it wouldn't even be necessary to use an additional microcontroller. We have a lot of computing power within the BBB processor (+PRUs). Still, the IMU is a relevant part of the autopilot (the gyros act as the metronome of the overall system) and we definitely would like to avoid I2C for such an important matter. Still i'd be interested in seeing the results of your suggestion! The new "Arduino Tre" sounds like fitting perfectly for your idea ;).
Great to hear of the excellent progress Victor,
It is no small task to get a complex cycle time critical autopilot program to run properly in a Linux operating environment.
I'm really excited with the moves towards more powerful on board computers! Way to go guys!
Victor, we have a project that we had been using the Beaglebone black for the video so that we could coordinate EMF and frequencies used for RC and craft data, which has been very successful
We did make a custom board for our controller however would be interested and willing to port over Rover control to the Beaglebone, let us know one you need to give it a out of lab test, as we have a extra Black to load it on.
@Victor,
Thank you for the update, much appreciated.
I have been following your progress posts on the drones-discussion Google Group and will continue to do so.
As soon as you have a stable hardware configuration and the initial bugs worked out I will be glad to test the hardware on one of my rovers.
Regards,
TCIII ArduRover2 Developer
Hi Thomas,
Thanks for your words. Unfortunately the hardware is not ready yet. We are having sensors and booting issues. We are looking into it but we are starting to think it's hardware related thereby it might take some time until we isolate the problems and come up with solutions.
We are eager for others to join unfortunately the manufacturing costs of the boards are big. We'll keep you posted :).
@Victor,
Great work and nice progress:-)
I have a rover and a BBB that I can use to test the rover version of the code.
Will an IMU Cape be available soon?
Regards,
TCIII ArduRover2 Developer