March 2024: Robot Edition

Hello and welcome to another Burf Update, this one focuses on ROBOTS and the Apple Vision Pro.  I know, you’re probably thinking, hang on they all feature robots?  What’s different?  Well at work, we received 2 Apple Vision Pros to play with, these $3500 devices are not in the UK yet, but the boss managed to get 2.  These super cool devices allow you to do spatial computing (had a go, it really is a game changer).

So when they arrived, my company (Compsoft Creative) wanted to do something cool with them, so after a brief chat, we decided we would use them to control my Inmoov Robot.

So after a 3-day intense workshop trying to get the head/hand/finger data from the Apple Vision Pro (Good work Dave), to control the robot, we decided to take it to the next level!

So we are going to take the Inmoov robot to the Robotic and Automation Show at the NEC next week.

The Rush

As you can see in the videos above, my Inmoov robot is a bit rough, it’s had a very hard life. So I decided it was time to make it look better before the show (which is in about 2 weeks). Work was kind enough to give me time to do it as they needed the robot for the show and there was a lot of work to do.

So what did I manage to do:

  • New head with a new camera and built-in speaker
  • Completely rewired the robot, and all connecting blocks removed
  • New chest plates, sensors, etc
  • New back plates
  • Fingertips added
  • Ears added (not sure about them)
  • A new neck which is a lot stronger
  • Printed a case for the power supply
  • Setup a Compsoft profile in MyRobotLab (took hours to get this to work well)

 

One-off the list

So one of my goals for the year was to finish the Inmoov robot, and I can say pretty confidently that I have achieved this. If the robot stays in 1 piece for the show, I can also say it is classed as reliable.

Anycubic Kobra 2 Max 3D printer

Bloody game-changer, that’s all I can say! So my work bought a 3D printer to help out with the printing needed to update the Inmoov robot because 3D printing is slow as hell. It takes hours/days to do anything large.

Not now it doesn’t, this 3D printer is at least 5-6 times faster than my old one. So 16h print takes 3 hours (standard speed, there is a sports mode!). Honestly, it’s the most epic exciting toy I have played with for ages. This opens up a lot of crazy ideas now.

Goals for the year

It’s always good to look back at the goals, let’s see if we can cross some off

This year’s Targets

  • Finish printing the 3D parts, make improvements to the body
  • Rebuild / Fix the head as the eyes do not function very well
  • Learn MyRobotLab (MRL)
  • MRL: Be able to say some words, and it can perform a custom gesture
  • MRL: Be able to track or detect people using OpenCV
    Plan to do a bit more on this
  • MRL: Be able to call a custom API
    Plan to do a bit more on this
  • MRL: Learn more about AIML (there is a Udemy course)
    Started watching the AIML course
  • Fire up the KinectV2 sensor that’s in the robot
    Not done yet, however, Inmoov does not support this.
  • Improve the wiring in the Inmoov robot as it is a mess

So over half the list has been done, 1 is no longer possible (Kinect) so I may replace that with something else. That is a great start for the year.

ROS2 mentorship

So I saw an offer on LinkedIn offering free mentorship on robotics and ROS. Of course, me being me, I applied for it and got accepted into the group. It has weekly sessions and lots of useful resources etc, however, due to me being super busy with the Inmoov robot, I am massively behind. I see this as a great way to kickstart me into ROS2, even if I can’t keep up.

Vehicles Update

So Jeff the Narava has been sold as I barely used it and it seemed a waste of money. Also, the Harley is up for sale for the same reasons. My dad gave me his 250cc scooter (well swapped for some LEGO) and my Gwiz is on the road and I am LOVING IT. The Gwiz AC version is so much fun. It’s now my main vehicle

Introducing Wheeler 2.0!

A little late to announce it as it has been on Youtube for quite a while but here is version 2.0 of Wheeler.

Features:

  • Controlled via a Android device talking to a Lego NXT (Programmed in RobotC).
  • Completely redesign body from version 1.
  • Moves and rotates via 2 NXT motors.
  • Head rotates and eyes/mouth light up.
  • Arms can rise / lower.
  • Claws can open / close via 2 micro motors .
  • Uses a Mindsensors.com RCX Multiplexor

 

Video

 

PhotoStream
http://www.flickr.com/photos/lego_robotics/sets/72157630101631308/

Wheeler : Progress so far

At the weekend I got a chance to do some more work on Wheeler, the Giant wheeled robot.  I have decided that this Robot will contain a single Lego NXT with a Mindsensor RCX multiplexer.  This will give me access to 3 NXT motors and 4 old style 9v motors.

Wheeler will have quite basic functionally so that I can save enough parts for the Massive T1 and try and keep cost down.  I have tons of old 9v stuff so that helps.

The plan is for 2 NXT motors to drive Wheeler, 2 RCX motors to move arms up, 1 RCX motor to control both hands, and 1 RCX motor to turn the head.  This leaves 1 NXT motor spare.

Below are some pictures of the current progress.

Before I started!

Now Wheeler has 2 basic arms with connected motors and the neck joint has a motor attached to it.

Top view of the 2 old style 9v Motors.  The motors are connected to the turntables via a worm gear that is connected to a 24 teeth cog which then turns the outer cog of the turntable via a 8 tooth cog.

The arm is very simple and light so that the motors can move it with ease.  Remember the old 9v motors are not as strong as the Power function ones.

The hand is controlled via a old style 9v micro motor.  These are very small, slow turning light motors that have become very rare.  I have decided that this type of hand is pretty naff and will be designing a new type that works like a vice.

Wheeler : Test 1, can it move!

One of the challenges of making large stuff is making it move.  The past 2 robots had been driven by tank tracks but I thought I would try something else.  So I posted on Eurobricks asking about giant wheel designs as Lego don’t make large wheels.  I had some awesome feedback and here is the result.

Update:
I must actually update the video as it works a lot better now.  I have replace the 16 long technic studded beams on the outer rim to be 15 long studless beams so that it does not lock up (clicking noise on video).

I have also removed the front set of balancing wheels and counter balanced the back ones so that it looks more impressive.  I will post an update soon.

Say hello to George (v1 – 2.1)

After Bob, I decided to adapt the KISS approach (Keep it simple stupid)!  My objective for the Great Western Lego Show 2011 was to make a Robot that worked, and worked well!  It had to move and be able to turn on anti-slip carpet, easy to control and be strong enough to take knocks / drive in to tables without breaking.
George is a fully functional 5 foot 7″ robot. Controlled via a PlayStation 2 controller, he can move about, rotate his upper body, move his arms / shoulders and grab items. His head also rotates, moves up and down and if you get too close, his eyes will roll.
More pictures can be found here:

Communications
This time around I decided to scrap having multiple NXT’s and went with the brilliant Mindsensors wireless Playstation 2 controller sensor pack. This allows you to control your NXT via a Playstation 2 controller which has a huge number of buttons / controls on it.  Then I linked the NXT to a hiTechnic IR link so that the NXT could send commands to power functions infrared receivers (all 4 channels). This setup gave me a large number of possible functions (11 different motor channels). I used RobotC to program the NXT as I find it an amazing programming language.

Weight
This robot had a large upright body beam. However it did not rise up and down so it did not require any counter balance weight. Also, it was not plated and only carried 1 NXT. This probably cut down a 3rd of the total weight of the robot, even though it was bigger than Bob (5 ft 7” tall).

Drive
The main base of the robot was smaller than the one on the previous robot which would naturally make it move a little easier, especially being lighter. Also the 4 XL motor driving it were geared down via a 8 to 24 tooth gearing. First time around, this just snapped the gears in half. The way to get around this was to use 2 sets of 8 to 24 tooth gears in parallel on the same axle per motor. Also, as a backup, in case the robot did not work on the show carpet, I could swap the 24 tooth gears for 40 tooth gears very easily which would give it even more driving power but at a cost to speed.

On the day
George was a huge success with the members of the public. Most people did not realise it moved and so jumped quite a bit when they realised it did. This was very funny except for the odd small child that got rather scared. People even started taking pictures next to it and a few people felt compelled to try to hug it or shake George’s robotic hand.

Issues
Issues were a minimum really.  After transporting George in more pieces than I hoped, I had to do a small tweak to the neck as the gears were slipping. Sadly I only had red technic with me for another one of my models and so it looked a little odd if you saw George from behind.

The main body bearing, as shown below just did not work on the day (well 2 days).  I am not sure why it worked at home and then stopped but I could only get it to turn 1 way.  Also the footballs kept falling out! I have since swapped them out and made a better version using wheels.

The only other issue which, to be fair, I already knew about, were that the large linear actuators (x3) in each shoulder joint made horrible noises (the built in clutches) and sometimes got jammed. Lego have made improvements to these actuators, but I have not got the newer version yet. I did try using pneumatics for the shoulder but that gives you little control.

The Future
I have already started work on 2 more large robots for next year (while also keeping George).  The first will be called Wheeler and is a giant wheeled robot. The start of it can be seen below.

The second robot will be based on Bob (T1) but avoiding all of the downfalls of it. I am looking forward to seeing people’s faces when there are 3 giant robots running around.

 

It started with BOB (T1 v1)

One of the robots that really fascinated me that I thought I could build was a T1 from Terminator 3.  It was a tracked robot with twin mini-guns (another thing I wanted to build).  So I went about building a rather large Lego version for the 2010 Great Western Lego Show which is our club’s (Brickish) largest event.
My first giant robot based on T1
More pictures can be found here:
Bob Flickr Stream
Bob suffered from a few major issues.

Weight
Due to its weight (well over 20KG) it could only turn on low friction floors.  The base of it was nearly a meter square running double tracks each side powered by 4 XL motors (directly linked to sprockets). Going forward was very quick but turning just did not work well. I only actually found this out on the day of the show where I discovered the floor was anti-slip carpet.

Communications.
The robot had very clever communications in my opinion. It used 4 NXT’s (2 slave, 2 master)  which communicated via bluetooth. One set of NXT’s had an accelerometer which when moved, would move one of the robots arms so that it would mimic the users moments. The other set of NXT’s also had an accelerometer to control the other arm however it also had a mode button so that you could control the robot’s other functions (moving, body rotation and lift, head movement) via the same accelerometer but in different modes. Each master NXT also had a fire button to fire the Zamor launchers on each arm.
The problem was that the bluetooth communication just was not stable enough and also suffered from lag and loss of data. You would put it in arm movement mode and it would instead move the tracks. There there would be a large amount of lag which meant the robot was always a few seconds behind. In practice, this caused the robot to drive itself in to a table 30 minutes after the show opened causing a partial rebuild in front of many people!  Quite embarrassing.

The robot also had a large main body beam (around 4ft long) that had to be counter balanced via many boat/train weights which caused even more weight issues. This main beam actually snapped a few technic beams while it was being built. The main beam could rise up via 5 pneumatic rams. However it required over 30psi to start it off (see below). This caused other issues.

I also completely covered Bob in black plate, which had a habit of falling off and also added more weight.