Generative Design: How AI Produces 100 Compliant Floor Plans in Minutes

Generative design represents a fundamental shift in how AEC Engineers and architects approach floor plan creation. Instead of manually drafting each iteration, then checking compliance, then revising — AI generates optimised, compliant layouts based on the constraints you define. The result is not one floor plan, but dozens of ranked options, each satisfying your brief, NBC requirements, and site parameters simultaneously.

The Core Idea: Design as an Optimisation Problem

Every floor plan design is, at its root, an optimisation problem: place rooms of specified sizes within a given footprint such that circulation is efficient, natural light is maximised, structural spans are rational, and building codes are satisfied. Manual design solves this by experience and iteration. Generative AI solves it algorithmically — and can test thousands of configurations in the time it takes to draft one manually.

The philosophical shift is significant: instead of the designer producing a solution and checking it against constraints, the designer defines the constraints and the AI produces solutions that satisfy them. The designer's role moves upstream — from drafter to problem definer and solution selector.

How Ecocraft Designer's Generative Engine Works

The generation process in Ecocraft Designer runs in four stages:

1. Input definition: The user specifies plot dimensions, orientation, FSI/FAR limits, number and type of rooms, accessibility requirements, and optional Vastu preferences. This takes 10–15 minutes and replaces the briefing document.
2. Constraint application: NBC 2016/2025 requirements (minimum room areas, ceiling heights, exit widths, ventilation ratios, setbacks) are applied as hard constraints — generation only produces layouts that satisfy all of them. Zone-specific seismic requirements, ECBC window-to-wall limits, and IGBC prerequisites are included.
3. Optimisation run: The generative algorithm (a combination of evolutionary design and constraint-satisfaction methods) explores the solution space and produces 50–200 candidate layouts ranked by composite score across circulation efficiency, natural light access, structural regularity, and IGBC credit score.
4. Review and selection: The architect reviews the ranked candidates in a visual interface, comparing key metrics side by side. Selected designs are exported to AutoCAD/Revit format for further detailing and submission drawing preparation.

What "NBC-Compliant" Actually Means in the Output

This is the critical differentiator from simple room-arrangement tools. Every generated layout satisfies:

• Minimum room area requirements per NBC 2016 Table 2 (bedroom 9.5 sqm, living 9.5 sqm, kitchen 5 sqm minimum)
• Minimum ceiling height (2.75 m habitable, 2.4 m non-habitable)
• Exit access travel distance (maximum 30 m to protected exit in residential)
• Natural light and ventilation (window area ≥ 1/10th floor area for light, 1/20th for ventilation)
• Setback requirements based on plot area and road width under applicable municipal bylaws
• Staircase width and headroom requirements

The system flags when municipal bylaw data for a specific city is not in its database — in those cases, it generates against NBC defaults and prompts the user to verify locally-specific rules.

Vastu Integration: Not a Compromise, a Constraint

Vastu Shastra compliance in generative design is handled the same way as NBC compliance — as a set of directional and spatial constraints. The system encodes the classical Vastu principles (Brahmasthan central open space, kitchen in Agni corner, master bedroom in south-west, main entrance in north, east, or north-east) as soft constraints that the generation engine tries to maximise within NBC hard constraints.

When NBC and Vastu requirements conflict (which does happen — a structurally optimal column grid sometimes disrupts Vastu proportions), the system ranks options with a composite score and flags the trade-off explicitly. The architect makes the call; the AI makes the trade-off transparent rather than hidden.

For more on the five common floor plan mistakes that AI prevents, see our post on floor plan design pitfalls.

Time Comparison: Manual vs. Generative Design

Output Formats and Integration

Generated layouts export to:

• AutoCAD DXF/DWG: For traditional submission drawing preparation
• Revit RVT: For BIM-native workflows
• PDF with compliance report: For client presentations and preliminary municipal review
• JSON: For integration with quantity estimation and project management tools