Abstract
Collective motion behaviour such as the movement of swarming bees, flocking birds or schooling fish has inspired computer-based swarming systems. They are widely used in agent formation control, including aerial and ground vehicles, teams of rescue robots, and exploration of dangerous environments with groups of robots. Collective motion behaviour is easy to describe, but highly subjective to detect. Humans can easily recognise these behaviours; however, it is hard for a computer system to recognise them. Since humans can easily recognise these behaviours, ground truth data from human perception is one way to enable machine learning methods to mimic this human perception. Hence ground truth data has been collected from human perception of collective motion behaviour recognition by running an online survey. In this survey, participants provide their opinion about the behaviour of 'boid' point masses. Each question of the survey contains a short video (around 10 seconds), captured from simulated boid movements. Participants were asked to drag a slider to label each video as either 'flocking' or 'not flocking'; 'aligned' or 'not aligned' or 'grouped' or 'not grouped'. By averaging these responses, three binary labels were created for each video. This data has been analysed to confirm that it is possible for a machine to learn binary classification labels from the human perception of collective behaviour dataset with high accuracy.