This project is part of the Computational Intelligence course of the MSc Ai. In this project a driver is created that is capable of racing in a racing simulator called TORCS. Multiple machine learning models are applied to create a well performing driver.
MLPR.p is the MLPRegressor model with default parameters.
MLPR_alt.p is the MLPRegressor model with optimized parameters.
MLPR_no_opponents.p is the MLPRegressor model with 3 opponents as training features added.
MLP_regressor.py is the training of the MLPRegressor model.
RNN.py is the training of the RNN model, with the use of TensorFlow.
Swarm_MyDriver.py is the driver file for the swarm optimization model.
Swarm_optimization.py is the optimization of the evolutionary model with the use of Swarm Intelligenc.
communication.py attempt to communicate between two drivers.
dense.h5 is the default deep neural net trained model.
evolutionary2.py the evolutionary algorithm with mutation and crossover.
latecurrent_best_run.p the best trained model with both evolutionary and swarm intelligence.
lstm.h5 is the optimized trained LSTM model.
my_driver_mlp2.py the driver file of evolutionary model.
my_driver_mlp.py the driver file of the MLPRegressor model with finetuning.
my_driver_RNN.py the driver file of the Deep Neural net model.
read_data.py reads in the training and test data.
run.py runs a driver file.
start.sh starts the driver on the server.