Beyond Static
Exploring Machine Learning for Adaptive Geometries in Expandable Structures

Details
Inspiration#



Module & Geometry Rules#
Basic Rules#






Revised Rules (Adapted from the Basic)#

The application of these principles in the assembly system, ensuring a thorough check of module assembly and a harmonious integration of designs with adaptability strategies.
For the next rounds of SOM iterations, create 20 varied input models using a randomizer with a Gene Pool component, ensuring diversity in architectural elements for optimized design outcomes.
Aggregation: Assembly System#




















Machine Learning – SOM#
In the ML process with SOM, cycles are performed, reducing dimensions until colors are well-defined, as seen in the iteration steps in the image, with the cycle reaching a maximum at 37. By the 36th cycle, the results have already repeated.






































After 37 cycles, 10-dimensional ‘glyphs’ condense into a two-dimensional map, resembling the image from Cycle 37. Each circle represents the central value of each dimension, and 10 models are selected based on the RGB map as representatives.
References#
- Eroğlu AK, Erden A, Akkök M. Design and Analysis of Grasshopper-Like Jumping Leg Mechanism in Biomimetic Approach. In11th Mechatronics Forum Biennial International Conference. Univarsity of Limerick, Ireland 2008. ⌃ ⌃2
- Yaheetech. (2023). 6 Seats Foldable Sideline Bench for Sports Team Camping Folding Bench Chairs Black. ⌃