So, I am now training things on tier 3, and although I think I probably won't get astonishing results before the deadline, I am quite happy to see my drones trying without using the imo-cheat noisy positional information :)
During my work with my approach on this project (end-to-end Deep Reinforcement Learning), I missed a few API functions. It would definitely be nice to have these in the future:
A way to directly specify the (reduced) size of the images from the drones without touching the json file (because I assume the competition setting will not let us modify this file ?)
A way to synchronously get the disqualification information
A way to synchronously get the gate-passing information (with gate id). Actually there is already a way to do this and I use it, but it is a bit geometrically intense.
A way to really control time (basically changing simContinueForTime() for something that is simulated-time based instead of cpu-time based should do the trick)
Hello everyone,
So, I am now training things on tier 3, and although I think I probably won't get astonishing results before the deadline, I am quite happy to see my drones trying without using the imo-cheat noisy positional information :)
During my work with my approach on this project (end-to-end Deep Reinforcement Learning), I missed a few API functions. It would definitely be nice to have these in the future: