Sound Swarm: A Synthetic Ecology of Embodied Mesh Synthesizers for Emergent Soundscapes

Mikhail Mansion, and Yasuaki Kakehi

Proceedings of the International Conference on New Interfaces for Musical Expression

Abstract

Sound Swarm is an embodied, multi-agent system that functions as a self-organizing mesh synthesizer, generating soundscapes through distributed sensing and environmentally mediated interaction. As these systems scale, traditional musical concepts—individual voice assignment, score-based synchronization, and deterministic event mapping—break down because the acoustic environment introduces propagation delay, masking, reflections, and spatial heterogeneity. We argue that composing for such systems requires a shift from event-specification to condition-setting: designing interaction rules, constraints, spatial arrangements, and time-scales under which sonic organization can emerge. This paper details a three-tier architecture—Perception, Behavior, and Expression—that decouples real-time audio synthesis from low-rate “colony cognition". We introduce a graph-based method for collective spatial sense-making: pairwise RF/acoustic measurements populate a relational tensor which is interpreted by a lightweight neural relational model to infer topology in situ. The framework is grounded in a biological taxonomy of coordination behaviors: honeybee-inspired role election for hierarchy, ant-inspired stigmergy for environmental coupling, and tree frog models for rhythmic coordination. By treating the swarm as a synthetic ecology, we demonstrate how emergent musical form arises from the triadic interaction between user-defined conditions, the physical environment, and agent-level biological heuristics.

Citation

Mikhail Mansion, and Yasuaki Kakehi. 2026. Sound Swarm: A Synthetic Ecology of Embodied Mesh Synthesizers for Emergent Soundscapes. Proceedings of the International Conference on New Interfaces for Musical Expression. DOI: 10.5281/zenodo.20784382 [PDF]

BibTeX Entry

@inproceedings{nime2026_122,
 abstract = {Sound Swarm is an embodied, multi-agent system that functions as a self-organizing mesh synthesizer, generating soundscapes through distributed sensing and environmentally mediated interaction. As these systems scale, traditional musical concepts—individual voice assignment, score-based synchronization, and deterministic event mapping—break down because the acoustic environment introduces propagation delay, masking, reflections, and spatial heterogeneity. We argue that composing for such systems requires a shift from event-specification to condition-setting: designing interaction rules, constraints, spatial arrangements, and time-scales under which sonic organization can emerge. This paper details a three-tier architecture—Perception, Behavior, and Expression—that decouples real-time audio synthesis from low-rate “colony cognition". We introduce a graph-based method for collective spatial sense-making: pairwise RF/acoustic measurements populate a relational tensor which is interpreted by a lightweight neural relational model to infer topology in situ. The framework is grounded in a biological taxonomy of coordination behaviors: honeybee-inspired role election for hierarchy, ant-inspired stigmergy for environmental coupling, and tree frog models for rhythmic coordination. By treating the swarm as a synthetic ecology, we demonstrate how emergent musical form arises from the triadic interaction between user-defined conditions, the physical environment, and agent-level biological heuristics.},
 address = {London, United Kingdom},
 articleno = {122},
 author = {Mikhail Mansion and Yasuaki  Kakehi},
 booktitle = {Proceedings of the International Conference on New Interfaces for Musical Expression},
 doi = {10.5281/zenodo.20784382},
 editor = {Benedict Gaster and João Tragtenberg and Anna Xambó and Tom Mitchell},
 issn = {2220-4806},
 month = {June},
 note = {},
 numpages = {13},
 pages = {993--1005},
 title = {Sound Swarm: A Synthetic Ecology of Embodied Mesh Synthesizers for Emergent Soundscapes},
 track = {paper},
 url = {http://nime.org/proceedings/2026/nime2026_122.pdf},
 year = {2026}
}