In this project, we aim to build a neuro-evolution agent that can play Top Gear (SNES 1992) game.

• This repo instruction asumes you are running everything on a Windows machine as the BizHawk build we provided is compiled for Windows platform. If you want to run it on Linux/Mac OS, please refer to the compile guide for BizHawk at https://github.com/TASVideos/BizHawk
• Make sure that port 37979 on your localhost is available and you have proper access permission.

We have built a customized BizHawk binary in the output folder. You don't have to build it by yourself. We have also included all the checkout points for neat training so that you can try out each individual generation stage.

Most of the programs run on Python. As a result, you need to install all the Python packages. To do so, you have to use pip. It is highly recommended to use virtualenv to separate your working environment from your system's default environment. To make things easier, we have also included pip requirements.txt. The following commands assume that you are using a Windows machine with Python 3 added to its environment path.

$ virtualenv env
$ source env/bin/activate
$ pip install -r requirements.txt

By default it assumes that your machine has two webcams. As a result, it will try to access the second webcam (index 1). If it is not the case, you need to change the webcam index in the source code. Open capture.py and find line 110. Then change cv2.VideoCapture(1) to any camera number you plan to use.

There are 2 different components to set this training process. We list those in order of running:

• Run ./TrainingServer.py
• Run executable "./output/EmuHawk.exe" to start the modified BizHawk emulator. Load up the desirable ROM. For our project, it is ./Top Gear (USA).sfc.

There are 3 different components to set this training process. We list those in order of running:

• Run ./HeuristicServer.py.
• Run ./capture.py, adjust your webcam and mark the corner points of the game screen
• Run executable ./output/EmuHawk.exe to start the modified BizHawk emulator. Load up the desirable ROM. For our project, it is ./Top Gear (USA).sfc. Load a desirable save state. In this project, F2 key is for Game Bot, F3 key is for Human player using Game Controller.

On a side note, you can play this game with your head if you use a webcam that faces directly towards you.

There are 4 different components to run this agent set up. We list those in order of running:

• Edit ./InterfaceServer.py, make sure that constant BIZHAWK is False. Run ./InterfaceServer.py.

• Edit ./NeuralNetwork.py, make sure that CHOSEN_GEN and CHOSEN_GENOME refer to the correct NEAT generation/genome. Our best performing combination is CHOSEN_GEN = 10, CHOSEN_GENOME = 1. Run ./NeuralNetwork.py

• Run executable ./output/EmuHawk.exe to start the modified BizHawk emulator. Load up the desirable ROM. For our project, it is ./Top Gear (USA).sfc. Load a desirable save state. In this project, F2 key is for Game Bot, F3 key is for Human player using Game Controller.

• Run ./capture.py, adjust your webcam and mark the corner points of the game screen

There are 3 different components to run this agent set up. We list those in order of running:

• Edit ./InterfaceServer.py, make sure that constant BIZHAWK is True. Run ./InterfaceServer.py

• Edit ./NeuralNetwork.py, make sure that CHOSEN_GEN and CHOSEN_GENOME refer to the correct NEAT generation/genome. Our best performing combination is CHOSEN_GEN = 10, CHOSEN_GENOME = 1. Run ./NeuralNetwork.py

• Run executable ./output/EmuHawk.exe to start the modified BizHawk emulator. Load up the desirable ROM. For our project, it is ./Top Gear (USA).sfc. Load a desirable save state. In this project, F2 key is for Game Bot, F3 key is for Human player using Game Controller.