Build

Computer Aided Floor Planning.

14-Sep-2019

About

One day, a friend of mine (an architect) told me, that one of their daily tasks is to estimate the value of a building site by creating a rough design of a hypothetical building and calculate its net internal area (NIA). This design does not have to be fancy but it should be reasonable. This is quite obviously a task to be given either to a trainee or even better to a PC. I took that as a challenge and started studying methods to automate the design of floor plans, hoping to find a proper solution.

Goals

1. Complete review of literature on the subject of computer aided floor planning and related topics.
2. Breakdown of the features and methods required by a system for computer aided floor planning.
3. Development and evaluation of a reasonable concept for computer aided floor planning.
4. Estimation of the overall work effort for the implementation of such a system or useful parts of it.

Results

Literature Summary

Reviewed 104 papers and books on the subject.

Wrote a brief summary on each item.

Feature Breakdown

Identified common features of the systems/methods found in literature and summarised them in a break down (see Chapter 1 in Computer-gestützte Grundrissplanung (de)).

Design of a Concept

Deduced design decisions from analysis work and developed a concept for a system, which assists in floor planning.

Concept Evaluation

Evaluated the use of such a system, the effort to implement it and the risks in respect to yet unsolved issues with the approach, which may reduce its practicability.

Conclusions

Studying the subject I made the following conclusions: There are lots of interesting approaches, but none which can actually generate reasonable floor plans without significant support by the user. Finding an optimal solution to a given set of constraints for a floor plan is a so-called NP-Complete problem, which means that the processing time will just be impractical for the task. Therefore, I went for a semi-automated approach: The computer constantly tries to improve solutions and the user can interactively select sets of solutions for further improvement. Once satisfied, the user can stop the process, export the result and work on that.

The practicability of such a semi-automated system seemed questionable to me. Usually the architect has a library of building designs from recent projects and could just take one of those floor plan designs and adapt it to the given conditions. This might be just faster then working with such a semi-automated planning system.

A rough estimate for the implementation of the outlined system suggests about a year for a prototypical implementation. Considering my concerns about its applicability, the risk was just too high and I set the project on hold until I find a better solution.