ROS

August 2025: The ROS2 Edition

Posted on by burf2000


Hello and welcome to another Burf update! Sadly, I didn’t get a chance to publish a July post — it’s been full throttle on the robotics front. This year is really shaping up to be a robotics-heavy one, and I’m loving it.

MTC: Robotics and Automation Event
We attended the event, and the second InMoov robot did well playing Rock, Paper, Scissors. A few lessons were learned — notably, face detection struggled with different skin tones and people wearing glasses. I had (wrongly) assumed that the maturity of OpenCV would have eliminated these kinds of biases by now.

As for the Sanbot robot — it didn’t get used much, so I need to think harder about a compelling use case for it.

Gwiz Update
Back in May, the G-Wiz passed its MOT, but unfortunately, the batteries were shot. Luckily, a kind soul from the G-Wiz forum donated some replacements (seriously nice of them), and now the car is up and running. The plan — once I clear the next few tasks — is to start recording data and begin training an AI to drive it.

ROS2, the Rover Robot and the Robotic Arms
This month has been packed with ROS2 learning. I’ve completed the ROS2 Manipulator course on Udemy (and I’m nearly done with the self-driving course). I’ve successfully controlled three different robotic arms using MoveIt, Gazebo, and RViz.

Massive thanks to Compsoft for giving me time to dive deep into ROS2 — they had a cool idea involving it, which gave me the perfect excuse to learn. I still have plenty to master, but I’m feeling far more confident than I did a few months ago.

AI
ChatGPT has been incredibly helpful throughout the ROS2 journey. Sure, it’s made its share of mistakes — looping, incorrect suggestions, etc. — but it’s been a great learning tool. The bigger vision is to use AI to simplify robotic arm control, lowering the barrier to entry for others.

Burf.co
The main website has been updated to reflect my focus on robotics. The new goal is to offer consultancy and robot rentals — an exciting next step in the Burf journey!

Inmoov humanoid 3
Yes, somehow I’ve ended up with a third InMoov robot — rescued just before it was scrapped. It still needs a lot of work, but I’ve managed to piece it back together and source enough parts to upgrade the others. This one now has LCD eyes, which adds a fun new dimension.

The Plan
At the start of this year, I set myself three high-level goals:

  • MyRobotLab + The Inmoov + LLM
    (Rock paper Scissors demo)
  • ROS2 and the Rover Robot
    (ROS2-powered rover and MoveIt-controlled robotic arms)
  • Gwiz and predictive driving
    (Train an AI to drive the G-Wiz using collected data)

So far, I’ve successfully achieved the first two. The third — predictive driving with the G-Wiz — hasn’t really taken off yet, though it’s still on the roadmap.

In the meantime, I’ve also acquired a whole fleet of Sanbot Elf robots and rebranded Burf.co as a robotics-focused website.

Thanks to changes at work, I’m now spending a lot more time on robotics. That’s not only helped me hit my previous goals faster, but it’s also inspired and refined the next set of objectives I’m working towards

New Goals
These goals aren’t in priority order and will naturally evolve over time, but they represent the next phase in my robotics journey:

  • Create a Sanbot Remote Control App
    Design a controller that streams all sensor data from the Sanbot to a user interface, allowing remote movement and activation of functions.
    — Critical for the robotics rental business idea.
  • Continue to deepen my ROS2 knowledge — this remains vital for both work and personal robotics development.
    — Ongoing learning improves real-world application and problem-solving.
    • Fix bugs around MyCobot280/xArm hardware, so that API works well
    • Finish self-driving course on Udemy
    • Understand SLAM and mapping better so that a robot can travel between rooms
    • Add IMU support
  • Continue Developing the InMoov Humanoids
    —These are great for showcasing my skills and drawing crowds at events.
    • Have an active one for improving Vision and LLM use. Think about using a Neural Net to give personality. Think about gestures.
    • Create a mobile Inmoov robot, even if its remote
    • Consider merging this with the Sanbot remote control app
  • Have Fun with the G-Wiz
    —This one’s for fun, learning, and experimentation.
    • Design a remote-control system (concept-only for now)
    • Add sensors and record journey data
    • Use that data in a self-driving AI test
    • Explore ROS integration for car control

These are some pretty epic goals, and definitely enough to keep me both challenged and inspired. Each one strengthens my experience in robotics across multiple fronts — hardware, AI, mobility, UX, and control systems.

I’ll need to keep an eye on context switching to ensure I stay focused enough on one thing at a time to make meaningful progress.

The Wednesday Update: Mission Accomplished

Posted on by burf2000

Welcome to another Wednesday update, again a rare edition that’s actually on Wednesday! A bit less on content this week due to it being my kid’s half term!

GWiz Update

The mission was always to make it to my Partner’s house in Sandhurst, and guess what it did it! No issues at all, max out about 45mph but I think it has more in it!

I have 1 small issue to sort with the GWiz, when breaking the re-gen system causes a burst of over 70v to the controller which causes the car to shut down. It’s a simple setting to change but needs a special cable.

Vex Tank

So on Tuesday, I wanted to start prototyping a simple tank to use as a platform for outside. It took about 2 hours to build and works pretty well until it got to the brick test. I think I am going to make a slightly larger more powerful one. At the same time, I want to test it again on a similar platform with different types of wheels (e.g Omni) which will hopefully give me some evidence on what works best out on the pavements.

ROS Lidar Robot

So I managed to get part of my old VEX ROS lidar robot working, I had missed out on some key bits of information in the readme but managed to get parts of it working. This was a great learning platform for ROS as it was purely built from the ground up by myself. The teleops control worked to control it, the Lidar also worked, the only issue was a lot of Errors were generated

Holiday

On a short holiday with my dad to Welles and Cheddar, I managed to get an Anki Overdrive set from a charity shop, this was awesome in its day and uses AI to drive the cars.

The Wednesday Update: Gwiz Edition

Posted on by burf2000

Welcome to another Wednesday update, this one is full of achievements!! Hold on to your pants!!!

The Gwiz in all its glory!

The Gwiz did a hill!

To make sure I can drive the Gwiz to my partners’ house I needed to check it would get up a rather steep hill. It’s not very long but is quite steep. I took a run-up and managed to get to the top of it at just under 30mph.

The Gwiz hit 45mph!

Erm, I would not recommend this, it’s rather scary, imagine being in a cardboard box and thinking you may crash into a wall at high speed. It’s very noisy and very bouncy.

The Giwz overtook a skateboarder

Well, he wasn’t going very fast but made me laugh!

The Gwiz did a shopping trip!!!

Yup the car has now done something actually useful, I mean it has cost me hundreds of pounds to get to this point but I can now make it to Morrisons.

I feel that the Gwiz has achieved a lot and only had one issue which I think is an overvoltage issue

VEX Robotics

So I had a bit of spare time during the week and managed to fire up my VEX V5 and VEX IQ robotics kit and get them set up to be useful. My aim is to use VEX EDR or V5 to control the Gwiz.

VEX V5 is pretty cool and even features a camera for tracking objects

ROS and Turtlebot

I also managed to fire up my old ROS (Robotics Operating System) Turlebot robot which is super cool. It has all sorts of sensors like a Kinetic type sensor for camera and depth detection, bumper sensors, and ultrasonics. This is the perfect robot to learn ROS on. It’s powered by a Nvidia TK1 so can do a lot of machine learning.

This is a point cloud of the back of me

MakeBlock CyberPi Go Kit

Electronics Course

And to make the evenings on the sofa even more exciting I have started the electronics course on Udemy that’s is on my to-do list.

VEX Worlds 2017 – Robotics Competition

Posted on by burf2000

Sorry for the late post about VEX Worlds, I thought I would have more time after worlds to catch up with stuff, sadly (well not really), the kids have been mega active.  My eldest son played his first football tournament, had a holiday, lots of family stuff!

So, VEX Worlds, what an amazing experience, I went along for the VEX EDR side of the competition (this year it was split EDR / IQ) as I was showing off the EDR Tank.  Sadly I had to leave the US early as my son, Max was ill.  Still a very cool experience!

So, the EDR tank, well it performed really really well in remote control mode.  I mean the thing was fairly slow but must have covered MILES!  The batteries never died on me, nor did any motors!  I did kill a few Omniwheels, however, that’s to be expected.   Even though I left early, the EDR tank did not and so others drove it around.  I have not received it back yet to see how bad it is now, but I am sure it will be fine.

The autonomous side was a bit of a failure, to be honest, and looking back I had set up myself to fail and I will explain why.  The autonomous side was using ROS (Robotic operating system) which is an industry standard.  I was using a Neato Lidar system which is awesome however it only had a range of 5 meters and SLAM (simultaneous localization and mapping) to work out where I was, and where I needed to go via building up a map.  SLAM works by detecting features of the surrounding area to work out where it is.  When you’re in a hall that’s hundreds of meters wide with very little features, a sensor with a range of 5 meters is practically useless.  In the end, I just showed kids how it worked on my laptop using RVIZ.  If I had to do this properly I would need to invest in a proper LIDAR system with a much greater range.  Another aspect which makes this very hard is all the people moving around, how can SLAM pick up features if they are constantly moving!

Overall, the EDR tank was hugely popular, I gave tons of fist bumps, high fives, etc, people just thought it was cool, just a little slow.

Next year, if I did a vehicle again, I would have to make it a lot faster and forget about advance sensors etc!

Here are some videos of VEX World and the EDR TANK:

The VEX EDR Tank

Posted on by burf2000

Over the last few month’s, I have been building a vehicle out of VEX EDR with the aim of it to self-drive. This is my first big project with VEX EDR, I usually build large stuff out of LEGO Mindstorms or VEXIQ. My view was that VEX EDR would be easier as it is a) made out of metal as opposed to plastic, and b) more powerful. I thought it would take me a week to make the vehicle, and the rest of the time would be on software. I had to also learn ROS (Robotic Operating System).

It actually turned out more challenging than I expected. Due to my lack of experience with EDR, I just assumed metal would just be stronger, and the motors would just work. However the first version of the Tank collapsed under my weight, and it took around 10 versions to get it to move me (95kg) without the motors shutting down after 5 seconds. Unlike LEGO and VEXIQ, EDR motors had a protection circuit in (PTC) which shut down the motor if it gets too hot or draws too much power. This is, of course, a good feature to protect the motor, however, it made my project very difficult.

I went from 8 motors, direct drive to 14 motors geared down to 2.44. The 8 motors could move me at a rapid speed but would just shut down after a few second.

Here is the progress of how it went

Current state

So I managed to get the Tank to a point where it worked, it was not as fast as I hoped however it seems reliable.  Next is the software.  ROS is a big subject to learn, there are lot of books on it and it is not the easiest thing to learn.  I have actually made some good progress on this thanks to the community.  Stay tuned to my next post about ROS.