Making InMoov Autonomous

hey, I'm looking to create an Inmoov project to study robotic perception and interaction for my final year project. I would like some advice or tips on programming the robot to be more autonomous. In particualr I would like the robot to be able to recognise objects and if not the ask questions to be able to learn what new objects are. 

I would like to know what programing language to use and maybe some basic code to get me started if possible.  

Any help would be greatly appreciated.  

Many thanks :)

kwatters's picture

Welcome !

Ahoy & Welcome Kevonline!

You have come to the right place...  There are so many topics to cover...  We already have the ability to train the InMoov to do facial recognition.  This is based on OpenCV.  We have the ability to hold down conversations based on ProgramAB.  For the more general case of general object recognition, you could probably begin creating some haar cascades classifiers for those objects.

By using webkit speech recognition + program ab, you can script up the interactions that would allow a user to chat with the robot to teach it.

MyRobotLab is written in Java.  It is scriptable via Python.  MyRobotLab also uses JNI to call into C / C++ (native) libraries for supporting things like video processing and getting depth map data back from the Kinect.

You've got an ambitious project, where are you going to school, what is your field of study?

Best & Good luck!

  -Kevin

 

Kevonline's picture

Thanks!

Hi Kevin, 

 

Thanks alot for your reply, the information and direction you have pointed me in have been so helpfull and has guven me more of an overview of the project which I have before me. 

I currently study Mechanical Engineering at Coventry Universitym UK. I Have some knowledge of electronics and take a module in coding. Yes it is a very big project, but somehthing which I have always been curious about and this is the perfect challenge for myself. 

Was also wondering about the different methods which you can use to operate the InMoov, Some people use a tablet and some use the Pi 3. What do you think is the best method? 

 

Thanks, 

Kevon

 

kwatters's picture

See you on the Dark Side of the Moon

I'd recommend using your normal desktop connected via USB to a couple arduinos.  Download an instal "Eclipse".. check out the MRL source code.. get it all going so you can debug it and step through things.

When you've got your stuff working, you can build it and deploy it to run on something like a RasPI3 or other.  (Personally,  I like the RasPI3, it's cheap, works pretty well, and a few of us here have the same hardware.

 

Reality is, when you debugging and testing out new functionality, you'll want to use a computetor that is comfortable to use (and fast)