EN / TR
← Back to Lab
05 — COMPLEX 2024

EMERGENT COMPLEXES

Agent-Based Form-Finding: When buildings negotiate their own positions.

Emergent Complexes - Swarm-Based Urban Design
Swarm simulation: 2,500+ agents negotiating building positions through magnetic field logic

Traditional master planning takes a top-down approach: the urban planner draws the site plan, buildings conform to it. But no natural system works this way. Bird flocks, fish schools, ant colonies—all create their own order through bottom-up logic.

Emergent Complexes brings this biological principle to architectural scale. Buildings become "agents" that negotiate their own positions. German architect Frei Otto's soap bubble experiments, computer scientist Craig Reynolds' boids algorithm, and philosopher Manuel DeLanda's assemblage theory form the conceptual framework for this research.

The result: urban textures that emerge spontaneously rather than being imposed from above. Pedestrian networks form through the logic of Physarum, a single-celled organism that finds the most efficient paths between food sources.

Conceptual Anchors

01

Frei Otto

1925–2015 · German Architect
"Form is shaped by forces, not by arbitrary decisions."

Otto's minimal surface experiments—soap bubbles, tensile structures, wool thread models—demonstrated that optimal forms emerge from physical forces, not designer intent.

Impact: Otto's physical form-finding is our conceptual anchor. We digitize his catenary principles into agent-based repulsion/attraction forces.
03

Manuel DeLanda

Mexican-American Philosopher · Assemblage Theory
"A whole emerges from interactions between its parts, not from a pre-existing plan."

DeLanda's assemblage theory (2006) argues that systems—including cities—are emergent phenomena. Components interact locally; patterns emerge globally. There is no 'master planner' in a city's evolution.

Impact: DeLanda justifies our bottom-up approach. We don't design the masterplan—we design the rules. The masterplan designs itself.
04

Physarum Polycephalum

Single-Celled Organism · Network Optimization
"The slime mold finds the shortest path without a brain."

Research by Tero et al. (2010) showed that Physarum—a brainless organism—can recreate the Tokyo rail network when food sources are placed at station locations. It optimizes for efficiency without central control.

Impact: Physarum's network logic powers our pedestrian circulation solver. We simulate 'nutrients' (program nodes) and let paths emerge organically.

Research Phases

01

Site Field Generation

Every site has invisible forces: solar vectors, wind patterns, view corridors, noise sources. We convert these into vector fields that will push/pull building agents.

Output: Environmental force field mesh
02

Building Agent Deployment

Each building program (housing, office, retail) becomes an agent with properties: mass (GFA), attraction (to similar programs), repulsion (from incompatible uses), solar appetite (facade preferences).

Output: 2,500+ programmed agents
03

Swarm Simulation

Agents negotiate positions over 10,000 iterations. They avoid collisions, seek solar access, cluster by program affinity, and respect setback regulations. The final state = emergent masterplan.

Output: Converged building configuration
04

Physarum Path Network

Once buildings settle, Physarum logic generates pedestrian networks. 'Nutrients' placed at building entries; paths emerge where flow is most efficient.

Output: Organic circulation network

What We Discovered

01

Agent-based layouts achieve 18% higher average daylight autonomy than grid-based layouts on the same site (tested on 5 complex projects).

+18% daylight
02

Physarum-derived pedestrian networks are 23% shorter (total path length) than orthogonal grid systems while connecting the same nodes.

-23% path length
03

Swarm-generated building clusters show 12-15% higher retail footfall (agent-based pedestrian simulation) due to organic 'desire line' integration.

+15% footfall
04

Complex emergence is faster: converged solutions in 10,000 iterations (~4 hours compute) vs. weeks of manual iteration.

4h vs weeks

Current Limitations

Aesthetic control: Emergent layouts lack the 'signature' of authored design. Some clients prefer intentional composition.

Regulatory friction: Zoning codes assume orthogonal grids. Emergent geometries require variance applications.

Computation-heavy: 10,000 iteration simulations require GPU clusters. Not viable for small-budget projects.

Unpredictability: Same inputs don't always yield same outputs (stochastic). Difficult for approval processes requiring deterministic plans.

Interested in Emergent Design?

We apply swarm intelligence to complex-scale projects. Reach out to discuss your site.