CONSTRAINT-BASED COMPOSITION IN REALTIME
Torsten Anders, Eduardo Reck Miranda · 2009
This paper proposes an approach for constraint-based algorithmic composition in realtime. To our knowledge, constraint programming – which performs a search – has not been used for music composition in realtime before. The main contribution of this paper is a meta-solver with a timeout. We decompose the music creation process into one sub-constraint-problem and solver call per time step. The meta-solver is given a maximum search time: if no solution was found after the maximum search time elapsed, then the search is terminated. The user defines what to do in this case. That way, the system is never busy for too long and always responses to new realtime input. This research extends our music constraint system Strasheela with realtime capabilities. Strasheela is based on the multi-paradigm programming language Oz. Our metasolver exploits its paradigms of declarative concurrency (concurrent constraint programming plus first-class procedures) and constraint programming based on computation spaces.