Machine learning

Project: The goal of this project was to find a policy for the car to race through the circuit as fast as possible. The car had the ability to steer, accelerate and brake. 

Solution: My approach to creating an autonomous racing game involves treating it as a classification problem with a discrete action space. To solve this, I employed the imitation learning method. The robot receives raw images as inputs, and I designed a feature extractor that is able to extract edge information from the pixels. After the features have been extracted, principal component analysis (PCA) is used to reduce the dimensionality. This results in a feature vector that is in a lower dimensional space, is resistant to changes in the background colour and can be applied to new, unseen datasets. To map each image to an action, a NN and a random forest algorithm were used. The random forest was able to out perform the NN.