Abstract

Morpho Design Explorer supports the investigation of alternatives within a user-defined parametric design space. Implemented for Rhino and Grasshopper, the software combines parameter generation, local data collection, fitness-based parent selection and population visualization. Its non-destructive population model retains saved alternatives, allowing quantitative evaluation and qualitative design judgment to inform subsequent exploration. This documentation describes the method, the current component interfaces and the interpretation of recorded design populations. It distinguishes implemented functionality from the broader research concepts that motivate the approach and identifies the information required to document a computational study.

Keywords: computational design; evolutionary exploration; dynamic population; parametric modeling; qualitative assessment; visual analytics.

1. Introduction

Computational design exploration involves more than the evaluation of a single optimum. A design question may include measurable objectives, competing constraints and spatial or visual qualities that are difficult to reduce to a numerical expression. An exploration environment must therefore support both the production of alternatives and their interpretation.

Morpho addresses this task through a recorded population of designs. Parameter values generate alternatives in a Grasshopper definition; analysis outputs and images describe those alternatives; the designer then identifies parents or numerical conditions for further generation. The archive remains available as the direction of inquiry changes.

The approach relates to work on genetic algorithms and the Theory of Inventive Problem Solving (TRIZ) in conceptual design [1]. The manual describes the software’s population model as a Non-Destructive Dynamic Population Genetic Algorithm (NDDP-GA) [2]. The software’s operational scope is specified separately from that broader research context.

2. Representation of the design space

Let a design be represented by a vector of input parameters x. In the current implementation, Grasshopper number sliders specify the name, bounds and precision of each input. The parametric definition determines the geometric family that can be expressed.

x = (x₁, …, xₙ) ∈ 𝒳, ℓⱼ ≤ xⱼ ≤ uⱼ(1)

Notation is explanatory. The realized set is also constrained by slider precision, integer steps and the geometry encoded in the definition.

Table 1. Notation used in this overview.
Symbol	Interpretation	Software correspondence
x	Design input vector	Genes generated from the connected sliders
y	Recorded numerical outputs	Named values supplied to Data Aggregator
m	Associated media	Tagged images and optional textual files
𝒫	Saved design population	Project records in the local database
𝒮	Eligible parents	IDs selected by the connected fitness filters

The analysis model supplies the numerical outcomes. Morpho does not contain a structural, daylight or environmental simulation solver. The validity and units of an output must be established through the model and evaluation tools used in the study.

3. Exploration procedure

The process alternates between generation, recording and review. With no fitness filter, the generator samples new parameter values within the input bounds. When eligible saved parents are selected, their values inform new combinations; mutation and spread settings control aspects of that variation.

Method diagram: parameter domain feeds generation and model evaluation; the saved population supports visual review and parent selection; selected parents return to generation — **Figure 1.** Conceptual organization of the Morpho workflow. The designer supplies the parametric model and its analysis. Saved alternatives form the record for visual review and subsequent parent selection. This diagram summarizes the workflow; it is not a timing or performance model.

Procedure 1. A designer-directed exploration cycle

Inputs: a working definition, slider bounds, named evaluation outputs, project directory and generation settings.

Define the design variables and verify the evaluation model.
Generate an initial sample without a fitness filter.
Record input values, numerical outputs and relevant images.
Review the saved population using images and numerical representations.
Select eligible parents by solution ID, numerical criteria or a combined expression.
Generate additional alternatives and append their records to the population.
Reassess the selection criteria and document the reasons for continuation or termination.

Output: a retained population and a documented sequence of design decisions.

3.1. Population retention

Non-destructive exploration keeps previously saved alternatives available for review. Conceptually, a completed save adds a record containing an ID, input values, recorded outputs and associated media:

𝒫_k+1 = 𝒫_k ∪ {r_k+1}, r = (ID, x, y, m)(2)

A conceptual description of successful record accumulation, not a guarantee against external modification or data loss. Back up the complete project folder.

3.2. Parent selection

Selection may express numerical conditions, direct inspection of design images, or both. Filter ID identifies specific saved alternatives; Filter and Filter Expression apply named comparisons and rankings. A selected set is local to the connected project.

ID IN (46, 73) AND Weight < 40000

Illustrative expression. The IDs, parameter name and threshold above are examples, not a recommendation or a reported experimental result. A valid selection with no matching records falls back to random generation; invalid expressions stop generation and report an error.

See the expression-language reference for precedence, ranking behavior and treatment of missing values.

4. Observation and interpretation

Each representation supports a different question. The image gallery relates visual form to recorded values; scatter plots reveal distributions across two or three parameters; parallel coordinates support the comparison of multiple dimensions. Their interpretation depends on objective directions, units and analysis assumptions.

Parallel coordinates representation reproduced from the Morpho manual — **Figure 2.** A multi-parameter representation in the Morpho Visualizer. Each line corresponds to an alternative across the selected axes. Source: Khodadadi (2026), *Morpho 2.0 User Manual* [2]. Reproduced with attribution; consult the current visualization reference for version 2.1 behavior.

4.1. Quantitative and qualitative assessment

A numerical filter makes a selection rule explicit, but does not by itself establish design quality. Visual inspection may identify proportion, spatial character or other concerns absent from the recorded metrics. The rationale for both kinds of selection should accompany the record of a study.

4.2. Non-dominated alternatives

A solution is non-dominated for a specified objective set when no other solution is at least as good in every objective and strictly better in at least one. This concept is discussed in the manual. The current bundled viewer does not provide automatic Pareto-front extraction or highlighting; an exact non-dominated set requires separate analysis of the saved values.

Changing the objective set or its directions can change which alternatives are non-dominated. A visually promising region of a scatter plot should not be presented as a calculated Pareto front unless that calculation has actually been performed.

5. Implementation and documentation

Morpho 2.1 provides seven core components for generation, capture, storage and visualization, and nine components for fitness selection and logical composition. The following documents describe the implemented behavior.

Table 2. Documentation map.
Document	Coverage
I. Methodology	Research context, population model, application domains and project credits
II. Component specification	All 16 active interfaces, filter syntax and selection semantics
III. Visual analysis	Gallery, plots, parameter filters, import and export behavior
IV. Implementation protocol	Installation, model connections, execution and data management
V. Study protocol	Research questions, assumptions, run records and reporting requirements
VI. Materials & references	Software, original manual, references and BibTeX records

6. Scope and limitations

Representation. Exploration is restricted to the family of designs expressible by the supplied parametric definition.
Evaluation. Recorded outputs inherit the assumptions, uncertainties and errors of the analysis model.
Selection. Numerical rankings describe the current recorded population and the selected criteria; they do not guarantee global optimality.
Research concepts. The current software contains neither an automated TRIZ problem-solving engine nor automatic Pareto-front extraction.
Repeatability. Random generation and external dependencies can affect repeated runs. Preserve the actual population and settings; do not assume deterministic replay.

The study protocol specifies a practical record for making these assumptions visible and distinguishing saved observations from their interpretation.

7. Acknowledgments

Morpho is led by Anahita Khodadadi, with development support from the University at Buffalo Research Foundation. The manual acknowledges Krishna Sivakumar, Jagruthi Kelavalli Jagadish and Babak Askari for computational development; Ryan Mellen for initial documentation and formatting; and Kimia Ghaderian for the logo. Further project background and contact information appear in the contributor record.

8. References

[1]
Khodadadi, A., and von Buelow, P. (2022). Design exploration by using a genetic algorithm and the Theory of Inventive Problem Solving (TRIZ). Automation in Construction, 141, 104354. doi:10.1016/j.autcon.2022.104354.
[2]
Khodadadi, A. (2026). Morpho 2.0 User Manual. First edition, October 2026. 35 pp. Full text [PDF]. doi:10.5281/zenodo.23477542.

The manual’s published title is retained. Software distributed on this website is version 2.1.0. Download bibliographic records [BibTeX].