Turing Learning Framework (TL) applied to the problem to generate Trajectories
- Java 8
- Java app implemented for Predicting Individual Trip Destinations With Artificial Potential Fields.
- Java app to load GraphML from file
- Java app to load trajectories from file
Modelling Human Behaviour with Turing Learning
Abstract. Modelling human behaviour is still an ongoing challenge that spaces between several fields like social science, artificial intelligence, and philosophy. Since the research of a metric able to define all the aspect of the human nature is still an ambitious task, most current studies use concepts like social forces or handwritten rules to model it Following the growing trend behind a new branch of Artificial Intelligence called Gen- erative AI, this paper presents the application of Turing Learning on the problem of modelling human behaviour. Turing Learning is a generative model that uses evolutionary algorithms as a way to learn behaviours without the need of predefine metrics, and using deep learning models it is able to produce human-like trajectories. We show how the system is able to infer the behaviours of the trajectories in the ETH dataset, fore- casting the next points with the truthfulness of being a possible human movement.
Paper currently under review at ECML 2018