Introduction form Steve

 I think it was back in February 2018 when  James  turned up at Leicester Hackspace along with his mother and sister with an interest in building a robot and learning some electronics and software. That summer he successfully applied for PIWARS 2019, now we just had to build a robot.

Our first entry, team RATAE, was based on hardware that James already possessed, a PI0W, Pimoroni Explorer Hat and a PI Supply Controller. So James set-to designing the chassis and I said I would help with the software. We did not realise it at the start, but we were effectively building the Tiny 4WD robot. The first chassis was built out of Mecanno . Documentation on the PI Supply website led us to “Approximate Engineering’s Python Game Controller”. This is going to be simple I thought, as the software for controlling the robot was already done, I just had to interface to my Je-Vois camera, add some line following sensors and we would be started,famous last words.  James would have no problems at home controlling the robot, but as soon as he stepped into the Hackspace, everything stopped working.

After a couple of visits to the Robot sessions at Cambridge Makerspace we were making some headway. A purchase of a decent USB power bank sorted out power issues, it still was not 100% reliable.

Off we went to PIWARS 2019 and it worked on the day. Although it should not have! There was a bug in the software which explained our reliability problems, sometimes it would generate an exception and stop it working, other times it would not. We found that fault after the event!

Off to the PIWARS Mini Conference for a great day out in October 2019. James purchased a REDROBOTICS hat to use with his PI3. This was going to be the basis for PIWARS 2020, entering as ECO-AMICA.

With help from Brett at the Hackspace an improved power supply was built and James designed and printed a new chassis using OpenSCAD. The chassis had space for a camera and line following sensors, James incorporated a NERF gun on a trailer.  We were making progress , a reliable system we could stream video, (using an example for pyimagesearch.com) just had to sort out line Following.

Then lockdown happened. We had a problem with one of the Line Following sensor’s and never managed to resolve the issue.  I started to look at line following using the camera but never managed to complete it. My excuse is that I did not have a working robot at home to test it on.Still, James created a video and did everything under manual control.

 

So here we are Team “Doing a Raspberry”.

Our basic hardware is the same. I managed to build my own robot, finally succumbing to some Pimoroni Offers and buying a set of Meccanum wheels I put something together. I knew all those bits I had up the loft would come in useful, a spare PI3 , camera, two motor controllers, 4 Pimoroni Motors.

At the PIWARS mini-conference I was impressed with Brian Starkey’s approach to edge detection. I have used that rather than OpenCV as a starting point. It is still a work in progress, it follows a “bendy” line but does not follow a curve. I had this great idea of using Multi-Processing so that the Image processing script would run on one core and the robot control script on another. Gave up because I could not get the two scripts to communicate. So we can check the algorithm I stream annotated video to a webpage, then it occurred to me, why not add an API request to get the error, and that is what I did. The robot control script requests the current line error when in line following mode. It seems to work, at the moment frame rate is about 10Hz.

On my test robot the camera is low down, it is going to be fine on the actual robot as the camera is mounted higher up. If you want to see how not to write code in Python, search github for windy54, I have put all of our code there to make it easy for James to download.

 

We are currently trying to have weekly zoom sessions, GCSE course work permitting for James and    baby sitting duties for me. But finally, James is  having to modify some code, changing my robot control script to work with his Redrobotics HAT and add in the requests for line error.

 

Another member Bill has just created a base for our course, so we need to create our own “Up the Garden Path” track to test it out. Just under 4 months left to finish the software, watch this space!

 

Steve Gale