Back
rAIcer: Self Driving AI Simulation Environment
Project: Game Design Certificate CapstoneOrg: Arizona State UniversityThe goal of this project was to develop a simulated virtual environment for training a self-driving game AI. The simulation can recreate a multitude of different environmental factors including different times of day, weather conditions, etc. The integrated data capture system uses custom depth, segment, and grayscale shaders to minimize file size while optimizing the quality of captured training data. AI models are trained from the data using the TensorFlow deep learning model running on the Nvidia CUDA parallel computing platform. Once an AI model is created, it can deployed and raced against by the user.