Week 5: Dts updates, edX course, battery issues, gym simulator and

07/10/2024

I came bright and early and to no avail, there was still no light flashing from the battery. I went ahead to take a new DB21J battery and charger. So now, I have the DB21J battery, HUT and charger with my DB21M components. Fingers crossed everything works out smoothly. It takes about 5 hours to fully charge and now I see the lights blinking on the battery indicating it’s percentage of charge.

I left it to charge and went to the weekly departmental meeting where I met a colleague who is currently trying to use the floor as a display in the hallway to detect hand gestures.

My colleague Yu Wati also found an edX course on the duckietown which I believe will help with a more in-depth analysis and understanding of the duckietown and duckiebots. Currently, I am going through 3 chapters – Introduction to self driving cars, planning I and planning II which dives into motion planning with graphs – explaining some algorithms such as Dijkstra’s, A* and D* – which talks about path optimizations.

I came back from the weekly departmental meeting to see that the DB21J battery I put to charge started showing some light and it was currently on orange – a good sign. It took longer than expected to charge – should be around 5 hours – but it was more than that so I had to wait for about 10 hours.

During the waiting period, I decided to disassemble my current DB21M to make room for when I insert the DB21J battery. After about 10 hours, I saw the 4 lights I needed to see and proceeded to assemble skyduck with this new battery. This took about 2 hours to fully assemble and I left it to charge overnight.

07/11/2024

Today, my goal was to make sure that skyduck turned on, was updated properly without any issues and to test all hardware components.

When I started the process, everything went on smoothly, even the dts desktop update which my colleagues had some issues with earlier. I was super excited. I got it to boot, went into the dashboard and tested all the components. One thing I found interesting was that the front bumper was not detected but nonetheless, it hasn’t caused any problems yet!

I started going through the other steps which were to make it move, make it shine and to make it see. Making it move involved using the arrow keys to navigate it. This was fun but something that came to an end because one of the wheels came off which resorted to the navigation keys not working. I tried to test the hardware component of the wheel and it did not work so I shut it down and turned it back on again and that seemed to work. My understanding is that once a part comes off, you will need to reset it for everything to start working well.

I then moved to making it shine. This consisted of manually changing the front and back LED lights and adjusting the intensity. There are 4 main colors – red, green, white and blue.

Moving on to “make it see“. This was pretty fun. I had set skyduck on the mat we put together earlier on, moved it and saw that it was capturing some good moments. As and when skyduck moves, the camera footage lagged a bit but when it was stable, it was pretty fast at capturing everything.

7/12/2024

After I had finished with the making it move, shine and see. I moved on to the next step which was the calibration of the wheels and the camera. This process was a bit more tedious because I had to get a certain level of accuracy to move on. The first thing I did was to power on skyduck to see if everything was working alright and it was. I proceeded to work on the calibration for the camera.

One difficult aspect of the calibration of the camera was that it was hard to get the Y axis to become green. I had to put skyduck on the table and moved the intrinsic camera calibration tool towards the floor to get the Y axis moving and that helped. Another thing I noticed is that the calibrate button was not a bit slow to react and that triggered a “not responding” which led to redoing the process(twice) but in the end, I did not get the popup and I was able to commit and viewed the stats on the dashboard.

For the calibration of the wheels, it was a bit tricky with the left and right drifts because just 0.1 made a huge difference. From the video below, you could see that it was initially going in a straight line and then it drifted very far away from the red tape. I also tested it by adjusting the left and right drifts but until I got one that was nearly perfect. I decided to try and test it off the tape because I suspect the tape did make it drift more. I did about two or three off the tape and I saw that the drift was not as bad and it was indeed moving in a straight line for a longer period of time. I guess time will tell when I test it in the physical/simulated environment.

07/13/2024

The focus moving forward is to have the simulator working, then move to the physical environment. The main goal of today is to get the duckiebot simulator -gym- running. A good amount of the day was spent doing some research, following steps and seeing where it landed me. I ran into the first problem when I tried running a file which was supposed to give me some results and it never did because I kept getting an error. This may have stemmed from the fact that there were some incompatibility issues arising from pyglet and gym. I tried downgrading the pyglet version and that gave me a different error which I debugged for most of the day but it was not as fruitful as I had hoped. Upon looking at the GitHub which I pulled to work on from the instructions given, it seemed the github had not been updated in the past 2/3 years so that may be a reason for our errors and compatibility issues. Some questions have been posted on the duckietown stack overflow and currently waiting for some feedback.