we had a real test-bed with model trains, circuits, and switches on a table with its own dedicated room. This test-bed was designed for students to achieve their research-oriented goals like their thesis or dissertation. Now, the issue was that every time a student would want to test their algorithms on it and if they were to reset it and test it again, the whole cycle without the testing would take them 30 minutes, just to set everything up again.
I solved this problem by creating a Train Simulator in Unity3D which can run on Windows or Linux and students can test their algorithm the same way they would test it on the real test-bed. This involved creating the same communication protocols with the trains in the simulation as the research test-bed along with exact same mapping of tracks, trains, and their speeds in the simulation. I also worked on a reinforcement learning algorithm for autonomous trains to automatically slow down on sharp curves and utilized an optimization algorithm to prioritize train tasks and reach the destination timely with minimum cost. This simulation was and is still being used by several students to test their algorithms before porting them to the real test-bed.