Hi everyone, I've decided to build a multirotor for experimenting with various machine learning algorithms.
My plan is to have my own algorithms handle low-level control i.e over orientation, whilst hovering and maneuvering. That is, I don't intend to use premade autopilot software or even a standard RC handset.
As of now I am planning to use a Raspberry Pi + Arduino to control the whole machine.
I am putting emphasis on robustness (less moving parts), and the longest possible flight time with the necessary components. Agility is not of any concern, but I would like for the machine to be able to operate indoors (standard home/office enviroment) and outdoors.
With these considerations in mind my current specification looks like this:
- 4 Rotors: most efficient yaw-capable number of rotors without using other mechanisms.
- Square, ~25cm width plastic frame, able to fit through British doors
- GPS
- 1 movable camera, about webcam quality or slightly better
Since I am a noob in the multicopter world, I'd like to ask for your advice on:
- What sensors I should use to gather orientation information. The ArduIMU seems the obvious solution to me at this point.
- If my frame size estimate is correct for indoor and outdoor use, and where I can source the frame.
- How much power each of the motors I am going to use should have.
- What battery capacity I should use given the other variables, and what flight time I can expect.
I am aware that the small size requirement, combined with the amount of electronics I plan to cram onto the machine, will make this a difficult build. With careful component choice I think it will be possible, if at worst by sacrificing flight time.
I think that covers everything; of course let me know if I've missed something obvious, keeping in mind my custom means of control.
Thanks in advance for all your help!
Replies
It might be worth looking at the Navio board for the PRI
Below is the official link.
http://www.emlid.com/
I think they might be at production stage already but just check it out. I think it contains pretty much everything you need.
Also check out the RPI forums. There are various individuals that have already done the same kind of thing.
Which model of the RPI are you going to use. The new Model B+ has a 40 pin GPI which I suspect will com in handy if you don't want to wait for the Navio to be released. RPI boards have a lot of hardware on them you will not need. I suggest de-soldering the parts you wont need like the sound port, HDMI port, etc. This will save a couple of grams.
To answer your questions:
1. You will need to decide which sensors you need for your purpose. Looking at the specs of FCs like the APM2.6 will give you a good idea of what is included in standard FCs
2. Size is up to you . 250mm is considered a microquad and normally uses 5-6 inch props. You will need to investigate what your complete payload ( all electronics + battery) will weigh before you decide on frame size, motors, ESC, prop size, etc.
3. See answer to question 2
4. See answer to question 2. In general ,the higher the capacity of the battery, the more weight it has, Unless you use Panasonic Li-on cells. You'll need to research that though. Flight times are impossible to predict without knowing the specs of the hardware you are going to use, and even then it's going to be more a guess than anything else.
Regarding the statement of " I don't intend to use premade autopilot software or even a standard RC handset.". I STRONGLY advise you to a a normal RC RX and TX unit to your build. This way you can take control of the craft if something goes wrong and you have a fly-away or erratic behavior from the craft. Going fully autonomous with RC backup is just plain dangerous!!