@hazikkhalid

So after changing the reward math again and tweaking the train script we got a big jump, i trained the model for 1lack steps and it came to 3k frames, my next goal is to increase the difficulty of the environment to make it challenging for the model

I am finally seeing results, my model is starting to show result, i checked the 20k steps one and its worked fine for 400 frames which is better that the last model with 50k steps, LETS GOOO

It has been a long time since my last dev log. I tried a few methods to create a custom model, but I think we should save that for later. I tried Keras, then returned to Stable Baselines and improved the training method.

Finally done with adding the notepad app, and its functions

I added functions like opening apps and dragging them, now will work on notepad

So i changed the Desktop, its better now and i like it more, i am still using simple css to style it

Made my own custom os, did javascrit after a long time and was really fun

So i had to take a break for 2 days as i was sick, but i am back on grind and now i am working on the ai the first train was trash and now i did some changes and am training 2nd one

Implemented a pre-made algorithm and trained my model on it

Working on the Ai part now getting interesting

Changed some things in the environment like changed obs from cars to tires

Started to learn Reinforcement learning in games, will start adding it to the game from tmrw

Made the environment better, added some png’s and also updated the collision part, now next step is to learn and add RF

Got to basic driving mechanics and obstacle generation, now i will either go for ai or make it look better first with asserts

So i am making a reinforcement learning system that learns driving entirely on its own. today started my first making the environment in pygame