I have a situation where I'm controlling my robot, which has a Kinect, via Matlab. I also have matlab code that gets depth information directly from Kinect. The robots actuators behave in synchronous fashion, i.e., the Matlab code won't return unless the action is completed. Thus, during action execution, I'm not getting depth information.
I would like to keep getting depth info from the kinect (perception) in thread/process 1 and would like to send signals to actuators, if necessary, in another thread/process. The tricky part is that the perception and action may or may not communicate with each other, i.e., they may or may not be independent. This being said, parfor etc in Matlab is not a good option for me since this is not parallel computation, rather two parallel processes that are completely different.
I'm using Freenect wrapper for perception and iCreate_toolbox wrapper for actuation, in Matlab.
Thank you, that helps a lot. My aim initially was to program in Matlab because these actuation/perception components are a very small part of the robot brain that I'm trying to build, and Matlab would've shrunken the coding time substantially. But I guess I can also try Python or something to see if that helps my cause. :)
Yes, you can do this in MATLAB, using the Parallel Computing Toolbox. You can use parfeval to execute a function asynchronously on a parallel pool worker.
http://www.mathworks.com/help/distcomp/parfeval.html
Anshul, the Parallel Computing Toolbox is a very powerful tool for programming parallel applications. You can use automatic parallelization (parfor), SPMD or GPU programming. However, in your particular case where you want to program completely different processes I would say that maybe Matlab is not the right choice for you. This paper gives an overview of you can do using the Parallel Computing Toolbox: http://www.europment.org/library/2014/santorini/bypaper/COMPUTERS/COMPUTERS1-05.pdf, excepting the GPU case. I hope the article be useful.
- Badges
- Science topic