While most of my project updates have historically been posted on the Cemetech forum, Cemetech has over the last twenty years transitioned from being primarily about my projects with a small ancillary community, to being solely a community hub and no longer focused on my projects at all. Therefore, although I still post about my projects along with all the other talented Cemetechians in that community, I am also making an effort to post more detailed information for easy reference here in a blog. Completed projects are documented in the Projects section, to the extent that I make time to keep it up to date.
I have long played Train Simulator, and have wanted something a bit more realistic to control the game. The first part of this log discussed my plans and motivations, and detailed the first steps along the journey: experimenting with connecting a speedometer and AWS acknowledgement plunger to the game. The core of controlling any train is making it go and stop, so naturally that was the next component to which I turned my attention.
Can AI run on a calculator? Machine learning and computer vision algorithms can certainly be run on a calculator albeit slowly: I ported a convolutional neural network (CNN) to a TI-84 Plus CE, making it capable of using “AI” to identify handwritten digits. As an added challenge, I implemented this in a single three-day train ride, including solving several interesting systems problems and making the code equally useable on a computer.
For quite a few years now, I have been an avid Train Simulator player, operating virtual trains from the United States, the UK, and beyond. I’ve played Dovetail Games' Train Simulator 2015 to 2022 and now “Train Simulator Classic”, FIFA-style annual upgrades that have incrementally improved the core game - and I’ve spent a decent amount of money on the much-memed Train Simulator DLC. The game is pretty immersive, but since part of my life also includes running real-life rail vehicles, I naturally want to make it feel even more real.