Inspired by the cloud formations and weather events, this mathematized cloud plays with visitor’s perception. Movement through the structure generates a series of dynamic interference views in its deep fabric, drifts and ruptures in visibility. A sea of redirecting vectors is pulling the visitor like an invisible gravity force through the fabric.
For the first time in history, design and architecture can start to approach extreme resolutions found in material systems in nature. Recognising that architecture is as fundamentally informational as it is material, with an intrinsic structure of information and concepts on which to operate, architecture’s logical core can be rewritten, by using accelerating capacities used to process information. Architecture is opening to concepts and techniques utilised in broader extra-disciplinary informational ecologies. It relates to the logics of formation of the contemporary world at large. Through large data, computer and material science, as well as new methods of building, matter can be accessed in its native resolution, both in vitro through simulation, and in vivo through robotic fabrication. We can now design and construct across many orders of scale, from material science to large scale design ecologies, from micro to macro. What could be called “data materialisation” is opening up the potential for architecture to finally resonate with the complexity of ecology.
Finer scale building blocks of matter and energy are now becoming available for increased resolution in manufacturing, construction and design. We can imagine a new scale of structures—microstructures capable of finer blending of material states, micro-precision design engineered for massive scale applications—increasingly malleable, plastic and intricate, primed for superperformance and unseen aesthetics.
Cloud Pergola is coming out of a lineage of research projects (Andrasek, Biothing and Wonderlab research, UCL), on high resolution structures, designed with algorithms, and built by robots. It belongs to a family of structures that are information packed, unprecedentedly intricate, lightweight yet strong and resilient.
Through Wonderlab research, numerous examples of proto-microstructures were developed using combinations of various algorithmic strategies. Its various instances are introducing applications of design and programmable multi-material printing with a high level of intricacy, resulting in superperformance (structural, thermal, acoustics and saving material) and expanded aesthetic possibilities. Robotically 3D-extruded lattice structures are one such example, designed with micro-precision for large scale applications in architecture and product design. Potential benefits of 3D printing by robots, are seen in faster construction, less material required to build in additive processes without unnecessary waste, highly specialised design fabrics, lighter structures as opposed to the current incredibly heavy buildings, and a high degree of tectonic and structural heterogeneity and local adaptation.
However, during the printing of thermoplastics into such complex spatial lattice, an issue arises with accumulation of tolerances and resulting imprecisions or even collapse of the structures; this is caused by the nonlinear behaviours of materials being printed, and a high number of connection points that very quickly accumulate errors—particularly in the case of extruding 3D lattices through the air where the material is only supported and connected through the nodes, while being suspended in the air the rest of the time. The problem for deep learning (AI) becomes finding the exact nodes to connect to after an accumulation of material tolerances, which inevitably mismatches the original computer simulation. To resolve this, the 3D printing path is trained to adapt real-time to the unpredictable material behaviour, by using the NVIDIA Jetson card on an industrial robotic arm. This enables path generation, real-time visual tracking of material and recomputing of robotic targets, thus increasing the speed and accuracy of such printing, and the overall stability of resulting lattice structures.
Croatian pavilion structure was designed by using the algorithm for multi-agent systems (MAS), whereby agents can be understood as active discrete elements whose behaviour is determined by a collection of rules, often based on stimulus-response logic. When agents act collectively in large populations, they are capable of producing complex behaviours and emergent effects.