Hi guys i recently tought of what i could run my robot of as the old laptop i used is getting really slow. I really like the idea of the tablet as gael does it. On the other hand the use of one raspberry pi would be benificial for my porte monnaie...

So I'd like to hear what you guys are using and maybe the main pro and cons you have faced with your choice. I would actually like to have a full functionaly robot with kinect etc in the end and also wonder what the minimum speccs would be to run myrobotlab with all services for inMoov.

In addition is there anything special i need for kinect or is the simple 30-40 bucks kinect360 from ebay enough to do all the work. (No special connectors etc..)

Sincerely Kakadu31.

Mats

6 years 12 months ago

I use two PI3 in my build.

The advantage is cost and that the PI has GPIO pins that can be used. So I use them for i2c commnication with different i2c devices, like servo drivers, AD-converters, IMU ( Accelerometer / gyro ).

The downside is performance. For example using OpenCV on the PI is slow even if I use the PI camera. The camera is fast, but then image processing algorithms are not optimal on the PI. The advice from the German robot soccer team was to write your own image processing.... That requires more understanding of image processing than just to use OpenCV. 

I have been able to access the Kinect  camera, but not the depth image. So the PI3 is more difficut to work with than a PC. 

I also tried MarySpeech on the PI, and the results are worse than on a PC. 

For the Kinect 360 you need to make sure that you get the "PC" version, with an additional power supply, since the Kinect use more power that a USB port can supply.

 

This is one of the primary goals of the board I'm designing.  For now, a Pi3 seems to be the favored choice, but it still leaves a lot to be desired.  Stay tuned, I'm not super active here, but I do plan on making an alpha board this year and plan to get it to a few of you guys by christmas.  I dont plan to have the neural net running yet, but it should have the basics working.

GroG

6 years 12 months ago

Kakadu !

Nice to see you surface.
I think the biggest thing to consider is Video.
Image processing takes wide bandwidth, and lots of cpu.  If there is a bottle-neck in any part - the results can be disappointing.

Camera/Kinect -> USB -> CPU + Memory -> Algorithms to do something interesting

If any one of those steps is "slow" ... saddness prevails..

So you either: 

  1. get a system which can handle that flow with reasonable performance
  2. off-load it to a different system
  3. don't do video

Moving servos and motors is trivial - use I2C or plug in another Arduino

I got an Intel Atom - and I think the general idea of getting an SBC is good, but the "FAIL" I had is the stupid mini-ITX and DC to DC converter :(

I would think that an inexpensive quad sbc with wifi, ssd & 4+ usb 3.0 exists WITH SIMPLE POWER REQUIREMENTS (ie 6, 12 or 24v without the need of a dc-dc converter) .. but don't know where they are ...

I'd search more if my hardware hubris was bigger...

good luck 

Thanks for the constructive replies. And hi GroG I am still around even if I am silent most of the time =)

I guess I will start using my MacBook for now and decide to switch on a better/"more mobile" solution later ,when there are better options (or I have more money, sigh...).

As I really want to use a lot of Video + Kinect Input but dont want to spend more than 250 €.

Thx for the tip with the Kinect Mats, I will try to hook the power supply tp a 12 V SLA that i use for my RobotBase, that should liver enough power i guess.

If anyone comes across a good alternative feel free to tell about their results!