In the middle of developing Syntrino - Syntro for the Arduino - and I was starting to pull my hair out while waiting endlessly for the Arduino IDE to complete a compilation. Then I discovered the Arduino Eclipse plugin and life is so much better now! The screen shot shows it working. There a few annoyances – for some reason it complains about Serial because HardwareSerial is supposedly not supplying a write() function to cover a pure virtual function in Print but whatever! It really is the only way to develop a program of any size. Anyway, rapid progress is being made – the Arduino is already swapping Syntro messages with SyntroControl, publishing its directory of services and doing other Syntro housekeeping functions.

Just about to start work on integrating an Arduino with Ethernet shield into Syntro. This could be an ideal way of controlling simple robots and yet providing them with significant intelligence with the help of off-board servers in the Syntro cloud. Since the Arduino can be and is being used for so many different types of sensing and control applications, it is an ideal Peripheral Class Processor (PCP) for use with a Syntro Lightweight Compute Cloud. It will be really easy to integrate large numbers of Ethernet connected Arduinos with storage and intelligent processing using this technology – truly ubiquitous computing (UbiComp) technology. I’m hoping that the initial open source release of Syntro will include the code that runs on the Arduino and also a sample Syntro app to show how to use it what it can do.

 //Check to see if the graph is running, if so stop it.
   if( (pControl) )  
   {
       HR = pControl->Pause();
       if (HR != S_OK)   
       {     
           if(verbose)printf("ERROR - Could not pause pControl\n");
       }
       else
       {     
           HR = pControl->Stop();     
           if (FAILED(HR)) if(verbose)printf("ERROR - Could not stop pControl\n");
        }
     }
   //Disconnect filters from capture device

First thing is to install nfs-kernel, nfs-common and portmap on the development PC. Then add something like this line to /etc/exports on the development PC:

/mnt/emlinux  192.168.5.5(rw,no_root_squash,no_subtree_check)

where 192.168.5.5 is the IP address of the embedded Linux system.  You may not be able to see the whole line but whole thing can be selected by triple-clicking on it.

You need to create the /mnt/emlinux directory with:

sudo mkdir /mnt/emlinux

Not a bad idea also to change the permission to make things easier on the development PC:

sudo chmod a+rw /mnt/emlinux

To get this working, either restart the PC or else do:

sudo /etc/init.d/nfs-kernel-server restart

That’s it for the devlopment PC.

On the embedded Linux system do:

mkdir /mnt/shared

This will be the mount point for the shared directory. Then, add this line to the /etc/fstab file on the embedded Linux system:

192.168.5.2:/mnt/emlinux  /mnt/shared  nfs  nolock,noauto,rw  0 0

where 192.168.5.2 is the IP address of the development PC. Again, triple click to get the whole line. You’ll need nfs-kernel, nfs-common and portmap on the embedded Linux system too, along with busybox. Then, to mount the shared directory, enter:

busybox mount /mnt/shared

or to unmount:

busybox umount /mnt/shared

Once this is done, files in /mnt/emlinux on the development PC will be visible also in /mnt/shared on the embedded Linux system.

This is my interpretation of the recipe – the original worked so any errors are mine!

Looking at the OpenCV source code, it really doesn’t want to use the native JPEG from the webcam so I slotted in my own uvc camera interface instead which does allow the JPEG mode to be selected. The results are impressive – a full 30fps now at 640 x 480 frame size and only roughly 12% CPU utilization! Not bad at all.

I always like to get SyntroWebCam running on new boards just to see what happens. This took about 15 times longer than expected for one reason or another and always due to me doing things wrong. Because I am more used to Ubuntu than Angstrom, I set up Ubuntu 11.10 for it from here. That worked just fine. I wanted to use OpenCV-2.3.1 so I decided to native compile that and SyntroWebCam on the BeagleBone itself for simplicity. The problem was that it would not recognize the existence of the camera!

At first it seemed as though it might be that libv4l-dev was not installed. So I did that and rebuilt but still no luck. Took me far too long to realize that there was a permissions problem and the quick fix was to use sudo to run it. I now know a lot more about the internals of OpenCV than is strictly necessary though.

As can be seen, the Bone was pumping out around 18fps - it was configured for 30fps at 320 x 240 sized images. In this case, OpenCV was running the JPEG compression but in principle the webcam I’m using can generate MJPEG frames directly which would certainly give much better performance. Haven’t got around to working out how to get OpenCV to use that mode as yet…

Just ordered my BeagleBone (photo obtained here) as it’ll make a nice addition to the devices supported by the Syntro system. The BeagleBone will look like what the late and lamented RCP (Robot Control Processor) was intended to be – a peripheral interface allowing other SyntroApps to drive robot platforms and interface to pretty much anything. It’ll fit in nicely with the SyntroRobot family of Syntro apps.