This report investigates the relation between the accuracy of localisation methods and their cost on the underlying communications network. The considered localisation methods are all based on the concept of integrated sensing and communications, where the location estimates are made using data extracted from existing communication. The report details the development of three machine learning based methods. Two methods which use machine learning as an intermediary step in angle of arrival based fingerprinting, and one method which uses machine learning to perform location estimations directly. The first two aim to improve on the localisation accuracy of a considered fingerprinting-based baseline method, while the third aims to reduce the required setup cost of gathering large amounts of fingerprints. It was found that despite the reduced cost of the third method, it was not able to perform in the considered system. For the two fingerprinting methods, both outperform the baseline while imposing little to no additional strain on the communications network.