The Decentralized Pianola: Evolving Mechanical Music Instruments Using a Genetic Algorithm

This paper presents computer experiments concerning the decentralized pianola, a hypothetical mechanical music instrument, whose large-scale musical behavior is the result of local physical interactions between simple elements. Traditional mechanical music instruments like the pianola and the music box rely for their operation on the separation between a sequential memory unit and an execution unit. In a decentralized mechanical instrument, musical memory is an emergent global property of the system, undistinguishable from the execution process. Such a machine is both a score and an instrument. The paper starts by discussing the difference between sequential memory systems and systems exhibiting emergent decentralized musical behavior. Next, the use of particle system simulation for exploring virtual decentralized instruments is demonstrated, and the architecture for a simple decentralized instrument is outlined. The paper continues by describing the use of a genetic algorithm for evolving decentralized instruments that reproduce a given musical behavior.