After almost a year of work, we finally got to show off our Inmoov progress at my daughter's curriculum fair tonight.    Needless to say, it was a big day for all of us.    I've got kids, so of course the robot's name is "Jarvis" (sorry Iron Man).   I use the "Peter" voice, which pretty much sounds like Jarvis.

Anyway, here is Jarvis on his first ride in the car....

He seemed a little nervous, and more than a little carsick.

Here he is at the show itself:

Jarvis (and Christine) were the center of attention all night.     I don't think anyone made it through the show without stopping and seeing the action.

Here are a couple of videos...  The noise was incredible, so it was hard to hear Jarvis.

https://youtu.be/iqvQCtrNmIo

https://youtu.be/smjnjXO9li8

Jarvis made it to the show and back home safely.    I was so drained, I slept for two hours immediately after getting home.    

However, everything wasn't without issues.    I was actually personally disappointed in how the software performed.   I haven't spent enough time with MRL and it didn't come together as well as I would have liked.   If anyone sees any egregious errors, let me know.

Here is the code I used:

https://github.com/MyRobotLab/pyrobotlab/blob/master/home/scruffy-bob/i…

Specifically, here are the issues I had:

1.  The keyboard input didn't work at all.   The Keyboard tab would show the keystrokes, but the python function was never called.   I didn't have time to debug this.   I had hoped to control the robot by the keyboard, but this didn't work for me.

2.  The OnText function was only called once or twice.   For some reason, it would stop receiving information from AcapelaSpeech.    The goal was to have Alice handle most of the responses, but also have a Python script be able to intercept incoming text, so I could control the robot by voice.   In the end, it was way too noisy to be able to do this, but it would have been nice if it had worked better.

3.  Facial recognition was marginal.    Having worked on this myself, I was let down.   I recently started a new job, so I had completely ignored the code until a few weeks ago, and it just didn't perform as well as I remembered.   I had downloaded the entire school's yearbook pictures and it did a pretty good job at trying to identify people walking by, but I had nothing but trouble training it on new pictures.   First, if I tried to train it with more than 1 picture per person, I would end up with a Mat error  [too tired to look it up now, but I'll work on it over the next few weeks].    Also, I noticed that if an image that was recognized slid off the bottom of the opencv screen, it would crash opencv. [Again, I'll see if I can track it down later]

Physically, the only issue I had was that the 805BB servo for rotneck was continously drawing about 0.5A, even when not moving.   I lubricated the heck out of the thing and it moved easily, but I still couldn't get it happy.    I was just happy it wasn't pulling enough power to burn it out over the course of a 2-hour demo, but it's something that still needs to be fixed.     I didn't have the arms running in the demo (but they both work).  I had them on in an earlier test, but since Jarvis isn't nailed down, the first thing he did was raise an arm and flip himself off the stool.   I figured it would be better to just not show the arms moving because we were confined to a very small space.

In summary, I wanted to share my first public experience with Inmoov.   For an eighth grade science fair, I think it was the star of the show.  [except for one of my friends' kid that was literally making a new star with his homemade fusion reactor]...