Emergent Complexes
Local placement rules are useful when their spatial consequences can be inspected.
- Study period
- 2024
- Note updated
The research question
A cluster of buildings contains competing relationships: proximity and quiet, shared routes and service access, growth and protected open space. We investigate whether local placement rules make these relationships easier to explore and revise.
The browser example changes attraction strength across three fixed starting layouts. A larger experiment would retain the same programme, boundary and rules while recording every seed, including collisions and unusable arrangements. The test must also ask what happens when the brief changes after an initial layout.
The architectural contribution is a legible rule set and a record of its consequences. A recognisable emergent pattern is only a starting point. Access, phasing and programme fit need explicit checks before an arrangement can inform a project.
One rule, three starting layouts
With weak attraction, separation spreads the blocks apart. The three starting layouts remain visibly different.
Deterministic toy model with the same three seeds and iteration count in every state. Crosses mark overlapping blocks. It calculates local placement only, not access, daylight or design quality.
Ideas and methods
Reynolds’s Boids explains how local rules can produce collective patterns. In architecture, a rule also carries a design choice: what attracts, what separates and what is protected? We retain that distinction when moving from the geometric example to a proposed settlement study.
OMA’s Seattle Central Library organizes distinct programme platforms and the spaces between them; ZHA’s KAPSARC describes a cellular arrangement of departments and shared public spaces. These precedents make the architectural question concrete: which activities belong together, how do people move between them, and where can the arrangement grow? They inform the programme and route checks that a geometric flocking model cannot supply.
Read the sources ↓Current work and the next experiment
- On this page
- A repeatable local-placement example that exposes collisions at different attraction strengths.
- Planned output
- A programme matrix, fixed seeds, a conventional reference plan and a revision test after a brief change.
How we will test it
- Keep the brief visible
- Use the same programme areas, required adjacencies, setbacks and site obstacles in the reference and generated layouts. Record hard constraints separately from preferences.
- Retain the whole family
- Repeat runs with recorded seeds and include congested, disconnected and unbuildable outcomes. Compare the distribution of results rather than the most attractive image.
- Change one condition
- Add a programme unit or close a route, then test how much rearrangement is required. Inspect pedestrian/service conflicts and route redundancy outside the generation model.
What remains open
- No verified comparative simulation outputs accompany this note.
- A simplified agent cannot represent the full brief, accessibility requirements or planning constraints.
- Stochastic models require repeated runs and explicit selection criteria.
Next output
The next output is a reproducible cluster study with a programme matrix, rule set, reference plan and small atlas of generated alternatives. Each sheet will show which relationships improve and which fail. This can inform an architectural workshop on phased growth without presenting the browser simulation as a validated planning tool.
Sources and method notes
- Craig Reynolds, Flocks, Herds, and Schools: A Distributed Behavioral Model (1987) ↗
The foundational paper on flocking through local rules.
- Craig Reynolds: Boids ↗
The author’s account of the local rules behind the flocking model, with links to the original work.
- OMA, Seattle Central Library (1999–2004) ↗
Official project account of programme platforms and the spaces between them. A precedent for translating a brief into spatial relationships.
- Zaha Hadid Architects, KAPSARC ↗
Official account of the cellular organization of departments and shared spaces, with provision for expansion. Informs the programme and growth questions rather than validating the placement rules.
Research connections to our products
- Swarm layouts as scenario input Exploration
Bringing reviewed layout alternatives into Archly is an exploratory direction, not an available masterplanning feature.
Archly → - Swarm rules as Grasshopper definitions Exploration
Falcon is a research tool. Parametric agent-rule experiments would need a defined implementation and evaluation.
Falcon AI → - Studio concept work
The 50-hectare campus concept is a related context for studying clusters and circulation; it is not evidence of measured gains.