About the project
This project will use artificial intelligence (machine learning) techniques to establish new models of pedestrian, cyclist and scooter rider path prediction for use in complex urban environments.
Established models of crowd movements predict the paths that pedestrians take based on a cumulative set of forces resulting from forecast future interactions. Increasingly however such models are thought to be inaccurate at predicting paths when the pedestrians are sharing urban spaces with other users, such as cyclists, scooters, café tables or advertising hoardings.
Standard models such as the common mathematical "Social Forces" approach fail to take into account both the momentum of the object (a key constraint for cyclist and scooter riders who cannot stop or change direction instantaneously) or the ability of the human mind to predict the future paths of moving objects that are themselves capable of independent movement decision making (e.g. other pedestrians respond to you as well as other pedestrians).
Such limitations can produce unrealistic predictions of how crowds of mixed pedestrians, cyclists and other micro-mobility users interact, leading to suboptimal design of urban spaces and building interiors. The opportunity therefore arises from developments in computer vision (to track movements of people within real life crowds) and machine learning algorithms to approximate human decision making, to develop the next generation of behavioural prediction models.
