AudioNode was developed as a proof-of-concept paper to develop work in the field of experience-driven procedural content, as outlined in a paper by Yannakakis and Togelius in their paper of 2011.
The most interesting part was choosing the mappings of the experiences onto musical parameters, which involved close collaboration with the computer scientist Davide Morelli.
The paper was published as part of the Computational Creativity and Games Workshop in June 2015, available here: AudioNode: Prototypical Affective Modelling in Experience-driven ProceduralMusic Generation.
A fork of Vincent Petry's "Node Reviver" python game was used to implement the theory, with music composition written in PureData, and embedded using libpd.
Musical elements, such as tempo and pitch, as well modulating effects such as reverb and filters, were manipulated based on data calculated from emotional and strategic aspects of the gameplay, such as frustration and planning.