Complex musical behaviours via time-variant audio feedback networks and distributed adaptation: a study of autopoietic infrastructures for real-time performance systems
The research project presented here is a study of the application of complex adaptive systems (CASes) for live music performance and composition by fully autonomous or semi-autonomous machines. The fundamental artistic concept and the motivating idea behind this project are that complex systems are an optimal model for creative music practice as they operate at the edge of chaos: that is, a condition where there is an interplay between order and disorder, or patterns and surprise. Arguably, this is an essential element found in music regardless of its genre or style. The central research questions addressed by this project are: how to realise music systems with an abstract yet structurally coherent and contextually complex output that display organicity and resemble aliveness? How to design music organisms so that artificial expressiveness and formal developments emerge and generate musical behaviours that contribute to the ongoing exploration at the edges of new music practices? How to create audio networks that are responsible for their structure and organisation where a substantial autonomy expands the paradigm of human-machine interaction? The methodology used in this project for the implementation of the music systems applies theories from complexity and adaptation, biology, and philosophy within nonlinear time-variant self-modulating feedback networks. The structural coupling between system and performer is realised by following a cybernetic approach in human-machine interaction and human-machine interfacing. The music systems developed for the creative works in this research are a combination of interdependent algorithms for the processing of information and the synthesis of sound and music. A technique formulated to improve the complexity of music systems is referred to as distributed adaptation, related to the notion of evolvability in biology and genetic algorithms. Distributed adaptation consists of making the adaptation infrastructure itself adaptive and time-variant by employing emergent sensing mechanisms for the generation of information signals, and emergent mapping functions between information signals and state variables. This framework realises the idea for which information and information processing recursively determine each other in a radical constructivist fashion with the important consequence that the machine ultimately constructs its reality as a self-sensing, self-performing, and context-dependent entity. This research includes seven music performances, each implementing CASes with or without human intervention. Also included is a library of original software algorithms for low-level and high-level information processing written in Faust. Chronologically ordered, the performances depict the progress of the study, starting with systems having basic adaptive characteristics and eventually revealing the more advanced ones where the distributed adaptation design is applied. Through self-reflection and post-hoc analysis, case studies illustrate that the combination of CASes and cybernetic performance and interfacing, and particularly distributed adaptation systems with or without human agents, produce a sophisticated musical output whose evolutions are complex, coherent and expressive. These results suggest that the emergent behaviours of such systems can be deployed as a means for the exploration of new music in practice. Furthermore, the autonomy and contextual nature of these systems suggest that promising results can be achieved when applying them to other fields involving audio, especially where interactivity with the surrounding environment is crucial, or when using them as musical instruments for users with special needs.