Reinforcement implementation of udacitys self-driving car simulation: https://github.com/udacity/self-driving-car-sim Coding changes are in the Assets/1_SelfDrivingCar/Scripts folder As well as this, there are some cosmetic changes, including adding a road divider between two separate roads
The development of a reward function: Penalty for being stuck, driving off the road. Reward for driving further. More or less penalty based on how the vehicle is stuck.
Road Detection system: Checks when any of the tires are off the road, used for resetting as well as data collection for road detection algorithms.
Segmentation system: Duplicates the RGB world, and copies movements of the car. Used for data collection & giving a reward based on location in lane.
The reward functions are transferable from simulation to real-world with transfer learning.