Observation Space Improvements and Headless Run

I wanted to write more about my reinforcement learning experiments with Unreal Engine's Learning Agents plugin. And also wanted to complement my previous post with some more details. Please check my previous post if you have not read it yet or need a refresher.

Write some introduction

Start with observation space improvements but only the first part where we add more observation points on the spline

Then write about the headless run, how it improves iteration, training speed and so on

Add comparison of training times, screenshot from tensorboard

And then try to find more improvements for observation space

Add comparison of cars' driving performence, speed etc with videos possibly