Hello everyone,
So I am a undergraduate mechanical engineering in my senior year. I have not really had the time or money to invest into side projects but now I do and robotics has always interested. I have no idea where to start besides continueing reading books and tutorials, however I would like to just jump into things. I believe I have enough patience and knowledge to at least not frustrate myself.
What I am wanting to build is a robot based around the Microsoft Kinect and the first thing I am trying to figure out is what microcontroller I should use. I want to have something powerful enough that I can utilize a very large portion of the kinect's potential but am unsure of the minimum requirements.
Sorry for the lack of information, I really don't have an idea of what this robot will be used for, because, really, I would like it to do everything.
Thank You
Kinect
The Kinect needs to be connected to a complete (and decently powerful) PC. Interpreting and using 3D data points is not the easiest project to start with. A mobile robot which can support a full PC and Kinect is also larger and consequently more expensive. If you’re really new to robotics, consider a construction kit:
https://www.robotshop.com/en/robot-construction-kits.html
If you’re really confident in programming and want to use the Kinetct, mount it on a simple pan/tilt and try to do something interesting.
min CPU
Hi
I use a intel I3 no my robot the kinect uses 70% of computer resourses I’m running ubuntu 14.4 with ros indigo its the same if iuse navigation or face recognition I have tried to run this setup on a beagle board XM but its to slow to be useful. I just use the micro controller to control motors and such
Peter
http://escaliente-robotics.blogspot.com.au/
Intel NUC?
I agree with CBenson; it needs to be a full PC rather than a microcontroller/SBC board.
I’m not sure what the specs are but have a look at the intel NUCs; they are full PCs but in a tiny form factor that could assist in mounting full PC power it to a project without the huge footprint.
Another long shot could be the lattepanda, which mtriplet is using to run windows in his AVA project.
Thanks
Potentially Raspi
Are you planning to use the Kinect V1 or V2?
I have seen some really good results from people using Raspberry Pi’s with V1 Kinects (see link:) https://www.youtube.com/watch?v=DMXKp98BY-w
However if you plan on using the V2, perhaps this might not be viable, as the V2 has much higher detail, and therefore more points to calculate. If you are planning to use the V1, I would suggest looking into “OpenKinect/freenect”, it’s an open source Kinect driver project (https://github.com/OpenKinect/libfreenect). Windows 10 is also available compiled for ARM chips (and runs quite nicely on Raspberry Pi, I tested it when it was in beta), so you may even be able to use the proper Microsoft drivers, but I can’t confirm this.
Personally I think the V1 would be a good choice, it’s good enough quality to perform most tasks, such as avoiding obstacles, and using the Kinect “fusion” it’s possible to project the RGB data (which is very good quality) onto the point data to allow for more complex tasks, such as object recognition and the like.
I’m not entirely sure OpenKinect plays nicely with Raspbian distro’s, so you may need to install Arch or something similar:
https://archlinuxarm.org/platforms/armv8/broadcom/raspberry-pi-3
Whatever you do, I hope you plan to share it here, would be interesting to see what projects you come up with.
Cheers,
Lexisonfire