Minimal motions: analysis of human motion tracking in machine sensing for filament construction in architecture

Addressing the need for resourcefulness in architectural design and construction, this research project bridges two key aspects: augmenting human craft with minimal technology to elucidate and leverage craft-related tacit knowledge in architecture, and developing a comprehensive analysis framework for these structures. Current augmented reality (AR) approaches in architectural construction predominantly rely on visual feedback systems—primarily head-mounted displays—followed by hand-held devices, that guide craftspersons according to pre-designed digital models with real-time updates during construction. In this context, it is investigated how motion capture can be used to develop a framework for machine learning during the prototyping of architectural filament structures at room scale. Minimal wearable technologies, in the form of motion trackers, record information from the physical model to create corresponding virtual structures. In order to employ statistical analysis models for generating these representations—and to develop a recommendation system for future iterations—it was initially necessary to ascertain what nature of information to provide, particularly for partially ordered systems in which tool-paths emerge during the construction process as opposed to being determined in advance. This research was conducted in three phases: first, gathering and visualising motion data from craftspersons during construction using wearable trackers; second, recording and conducting statistical analysis on the filament structure and corresponding graphs at the macro-scale; third, focusing on isolated edge creation and effects on the local and global structure. Through motion capture, statistical analysis and generative design processes, the process of making these structures is elucidated, their topology is described on a micro- and macro-scale and a preliminary prototype capable of generating future structures is presented. Thus, this research project introduces a multi-scale AR methodology with non-visual interfaces that minimises interference with craft processes while enabling the simultaneous observation of multiple craftspersons. Unlike conventional approaches, the Unity-based framework uses physical structures to inform digital models through iterative statistical analysis models, analysing emergent filament structures at both macro- and micro-scales. As a future perspective, these systems can be analysed to provide future craftspersons with real-time recommendations as an example of integrating tacit knowledge into the design of architectural material systems.