Skip to content

Latest commit

 

History

History
15 lines (9 loc) · 955 Bytes

README.md

File metadata and controls

15 lines (9 loc) · 955 Bytes

selfdrivingcar2

In this project, I use reinforcement learning to train a self driving car in AirSim. This model's framework is based off of the cookbook made on AirSim.

Navigation

rl_model.py The raw model for reinforcement learning. The machine learning and CNN portion of the project is located in this file.

trainRLmodel.py This is the script you need to run if you want to train the model.

airsim_client.py This file includes many of the commands we use in distributed_agent.py so we can connect our results to the AirSim client.

distributed_agent.py This file converts the results of our decision-making neural net into executable commands in the AirSim environment.

Generator.py This file is used to generate additional data for our model to train upon.

pretrain_model_weights.h5 Usually, this model would take days to train. Using preloaded weights for our model, the training converges after a couple of hours.