Device locations using machine learning

A method of determining locations of devices, the method comprising:

• receiving location data representative of a set of candidate locations in an environment;
• generating training data comprising a plurality of training data sets, each training data set comprising data representing synthetic distance measurements between respective pairs of the candidate locations, wherein the synthetic distance measurements are determined by determining geometric distances between pairs of the candidate locations and applying random variations to the geometric distances;
• using the training data and the location data to train a machine learning model that is arranged to receive a set of distance measurements as input and to generate data representative of a set of one or more locations as output;
• receiving measurement data representative of a set of measured distance measurements between respective pairs of a plurality of devices in the environment;
• inputting the set of measured distance measurements to the trained machine learning model;
• and
• determining, from the trained machine learning model, data representative of a respective location in the environment for each of the plurality of devices.