[HH] Rethinking LinuxPC Robot -- the Linux PI Robot

markw at mohawksoft.com markw at mohawksoft.com
Sun Apr 7 23:01:00 EDT 2013


> Hello Mark,
>
> On Apr 7, 2013, at 7:24 PM, markw at mohawksoft.com wrote:
>
>> The PC is still swamped with I/O, especially with the cameras.
>
> I am wondering, do you need any drivers for the cameras? Switching to ARM,
> you should check you have that covered in the R.PI distros.

I'm using the standard v4l2 stuff and I carefully select cameras that
perform well. You'd be surprised at how badly cheap USB cameras perform.

>
> How many cores do you have? I presume it's an older Atom, so 1? you should
> be able to swing 4 cores in an AMD chip within the budget we described.

I have a dual core hyperthreaded Atom, so it looks like 4 cores, but as
you know, hyperthreads suck. I originally had an old VIA 3 800MHZ mini-ITX
board, which, lets be honest, is probably matched by a PI today.


I try to keep the robot as something that has modern technology. Its one
of my hobbies. I just think that the current model design is getting
really old. I/O is a HUGE limitation it slows everything down.
Distributing that I/O does make sense. So, a PI for each camera, and each
PI being a camera sever. On top of that, if I can implement the motor
control on a PI, a PI for that as well otherwise keep the Arduino
controller and port my interfacing to PI.

Using an ethernet switch as a backplane and a number of PIs dual
functioning as I/O processor and distributed processor. I could even use
the MPI interface to do parallel processing on the PIs with the cameras so
that a central path planner only gets processed images. It starts to feel
more like a brain, don't you think?

The current design is done. It has its limitations. I really do think the
future of all computing is by scaling across processing units and what
better place to experiment with that than robotics?


>
> Best -F
>
> _________________________________________
> -- "'Problem' is a bleak word for challenge" - Richard Fish
> (Federico L. Lucifredi) - flucifredi at acm.org - GnuPG 0x4A73884C
>
>
>
>
>
>
>
>





More information about the Hardwarehacking mailing list