Approach and contribution
30-second case brief
An optimized façade with personalized daylight control.
- Problem
- Balance daylight, glare, energy demand, and user preference in a home-office façade.
- My contribution
- Project lead: developed the models, simulations, optimization, preference-based controller and digital interface; first author of the published case study.
- Outcome
- Five strategies compared, a Pareto-front geometry selected, and a personalized controller evaluated across four stages with a demonstrated phone-to-model interface.
ApproachCommon simulation model → five strategies → NSGA-II geometry search → preference regression → controller → staged results.
Project description and responsibility
- Context
- M.Sc. thesis · Politecnico di Milano · 2021
- Role
- Project lead · modeling, simulations, optimization, preference modeling and digital control
- Case
- Southwest-facing home office · Lecco, Italy
- Methods
- Rhino/Grasshopper · Ladybug/Honeybee · Radiance · EnergyPlus · Wallacei/NSGA-II · regression · OSC
- Thesis team
- Ramyar Tajik · Saeideh Soltanmohammadlou · Amir Kianfar
- Publication
- Journal of Green Building 19(1), 2024 · first author
Interactive tool · façade controller
Control daylight from inside the room.
Move the sun, set the glare response, and tune shade transparency. This demonstrator explains the control relationships; values are illustrative.
Venetians 49% closed · Roller 47% deployed
The design problemA home office with competing needs
Keep useful daylight while controlling direct sun
A large window can make a home office bright and inviting, but the same opening can admit direct sunlight onto a desk, create glare in the occupant’s field of view, and change the heating and cooling demand. Closing a blind solves one problem while potentially creating another: less daylight, more electric lighting, or a reduced view outside. This project approached those competing effects as one design and control problem.
The case was a third-floor home office in Lecco, Italy, with a southwest-facing exterior wall and two tall windows. The room measures 3.8 × 4.6 × 3.0 m, with approximately 7 m² of glazing. The objective was to develop a retrofit that could respond to the sun and the occupant rather than relying on a single fixed shading position.
Develop the façade and its operating logic together
As project lead, I developed the parametric model, simulations, optimization, preference model, control implementation, and the results and discussion. I received support with the literature review and parts of the NSGA-II methodology. The thesis and published article retain the full collaborator and coauthor credits.
The resulting concept combines exterior Venetian blinds over the upper part of the opening with an interior roller over the lower part. Slat rotation controls the upper zone, while roller deployment controls the lower zone. This division provides separate responses to direct sunlight, diffuse daylight and glare. The development therefore needed to establish both a suitable physical configuration and a rule for deciding how it should operate.
Geometry → simulation → comparable outputsConstructing a common comparison model
Hold the case constant and change the shading
The room was rebuilt in Rhino and Grasshopper so the same geometry could drive daylight and building-energy analysis. Ladybug and Honeybee connected the model to weather and simulation inputs; Radiance provided the daylight calculations, while EnergyPlus provided the thermal and energy calculations. Each shading strategy was evaluated against the same case, making its performance a consequence of the shade configuration rather than a different room or envelope.
For the strategy comparison, the simulations sampled the 21st day of each month from 9 a.m. to 5 p.m., giving 108 hourly cases across the year. These common dates captured different seasonal sun conditions. The unshaded room served as the reference for the five-strategy comparison, and the later controller evaluation used February 21 and August 21 to examine the daily response in winter and summer.
- 01Rebuild the case
Room, glazing, envelope and shading geometry in Rhino/Grasshopper.
- 02Simulate each candidate
Daylight through Radiance; thermal and energy analysis through EnergyPlus.
- 03Compare the outputs
Heating, cooling, lighting, useful daylight and probable glare.
- 04Resolve geometry and control
NSGA-II trade-offs, preferred illuminance and responsive shade positions.
| Case | Third-floor home office · Lecco, Italy · southwest façade |
|---|---|
| Room / glazing | 3.8 × 4.6 × 3.0 m · two 2.7 m-high windows · approximately 7 m² glazing |
| Envelope | Wall U-value 0.25 W/(m²·K) · window U-value 3.5 W/(m²·K) |
| Strategy comparison | 21st of each month · 09:00–17:00 · 108 sampled hours |
| Final controller comparison | February 21 and August 21 · hourly daily profiles |
| Daylight / glare | Radiance · published UDI band 150–1,500 lux · later directional DGP |
| Thermal / energy | EnergyPlus through the Rhino/Grasshopper workflow |
Choose the system before refining its dimensionsComparing five shading strategies
Separate the effect of each shading device
The first comparison tested five alternatives: the existing unshaded room (ST01), fixed exterior Venetian blinds (ST02), a dynamic interior roller (ST03), fixed Venetian blinds combined with a dynamic roller (ST04), and dynamic Venetian blinds combined with a dynamic roller (ST05). This sequence isolated what each device contributed before combining them.
Cooling, heating and electrical-lighting loads were evaluated separately. Blocking solar gains can reduce cooling demand but also remove useful winter gains; a shade that suppresses daylight can increase lighting demand. Looking at the three outputs together made those consequences visible instead of judging a configuration from one favorable metric.
Read the seasonal loads, then the combined result
The cooling comparison shows the benefit of exterior protection during the warmer sampled periods. The heating comparison reveals the competing effect of reduced solar admission, and the lighting comparison shows when shading increases the need for artificial light. These plots explain why the choice could not be made from cooling demand alone.
Across the source’s sampled comparison, the fully dynamic combination offered an acceptable energy balance and retained the ability to change both devices. That flexibility was important for the next stage: the geometry and operating rules could be improved while keeping independent control of the upper and lower window zones.
Select a balance of daylight and glare
The published comparison defined useful daylight as average illuminance between 150 and 1,500 lux. It also counted sampled hours above 3,000 lux as a probable-glare indicator. This early indicator provided a common screen for all five strategies; the later controller used the directional Daylight Glare Probability (DGP) calculation at the occupant’s viewpoint.
Reading the daylight and glare comparisons together shows why the choice needed to consider both outcomes. The study selected ST05 as a controllable starting configuration that allowed both devices to respond. The next question was which combination of slat dimensions and roller transmittance would improve its balance of energy, useful daylight and glare.
Make the trade-offs and the selected design traceableFinding the geometry through optimization
Turn four physical properties into search variables
With ST05 selected, I parameterized four properties: roller visible transmittance, slat width, the distance from the topmost Venetian slat pin to the bottom slat pin, and the number of slats. Changing these inputs rebuilt the shade geometry and reran the performance calculations. The roller’s transmittance was particularly important because it had to control glare while still admitting useful daylight.
Wallacei applied NSGA-II within Grasshopper to search combinations of those variables. Rather than assigning one arbitrary score to every design, the search compared competing performance objectives. A Pareto-front solution represents a trade-off: improving one objective would require giving up performance in another.
| Property | Tested values | Selected |
|---|---|---|
| Roller visible transmittance | 30%, 45%, 60%, 75% | 45% |
| Slat width | 8, 12, 16, 20 cm | 16 cm |
| Topmost-to-bottom slat-pin distance | 50, 70, 90, 100 cm | 90 cm |
| Slat count | 4, 6, 8, 10 | 8 |
Refine the objectives to expose useful alternatives
The initial search treated cooling, heating, electric lighting, useful daylight and probable glare as five separate objectives. I then combined the three energy terms into a primary-energy objective, using the study’s conversion indices of 2.5 for cooling, 1.1 for heating and 2.5 for electrical lighting. The refined search minimized primary energy and glare while maximizing useful daylight.
The parallel-coordinate plots show how each candidate moves across three fitness-ranking axes: F01 is glare, F02 is illuminance and F03 is primary energy. Generation color progresses from light pink to dark blue, and the second plot overlays the Pareto-front solutions in black. These are the optimizer’s ranking scales. Reading the trade-offs alongside the Pareto set made the selection explainable: a practical compromise among energy, daylight and glare rather than an isolated minimum on one axis.
Decode the selected Pareto solution into a physical configuration
The selected Pareto-front configuration uses a roller with 45% visible transmittance, eight Venetian slats, a slat width of 16 cm, and 90 cm from the topmost slat pin to the bottom slat pin. These are the physical values recovered from the selected solution, so the optimization result can be read as a specific façade configuration.
This geometry became the basis for the personalized-control stage. It established what the system could physically do, but did not by itself determine the best position at each hour. That required connecting the available movements to the sun, the occupant’s preferred illuminance and the view-dependent glare calculation.
Measured preference inputs → two regression modelsTurning preferences into control thresholds
Record a preferred range rather than one fixed target
A fixed illuminance threshold does not describe how an individual wants to work under different daylight conditions. I therefore collected preference observations in my home office over four months in 2020. For each observation, the procedure recorded the default condition and the highest and lowest acceptable illuminance conditions, together with the activity and position.
The measurement procedure used a phone light-meter application, with measurements taken 30 cm perpendicular to the surface. Five work-plane points provided an average work-plane illuminance (WPI), and a further wall or forward-facing reading described the visual surroundings. This created paired observations linking the available daylight condition to the preferred upper and lower limits.
Predict both boundaries of the preferred range
Two linear regressions related default WPI to the preferred lower and upper illuminance thresholds. The default reading was the input; the two preferred readings were the outputs. This produced a range that changed with the observed condition instead of imposing the same pair of limits at every hour.
I implemented the regressions using LunchBox components in Grasshopper and passed their predicted boundaries to the control routine. The scatter plot shows the observations and fitted relationships, while the component view shows how those relationships were connected to the model. The next stage used that range as guidance while also handling direct sun, glare and view.
A controller with explicit priorities and a working phone interfaceFrom preferred daylight to shading actions
Intercept direct sun at the relevant part of the window
The occupied-mode algorithm first checked whether direct sunlight reached the work plane. A sun vector was traced from the affected position back to the façade, identifying which shading zone needed to respond. An intersection in the Venetian zone closed the relevant slats and those above it to −90°. An intersection in the roller zone closed the Venetian blinds and deployed the roller to the corresponding height.
This geometric test made the response specific to the source of direct sunlight. It also separated direct-sun protection from the subsequent adjustment for diffuse daylight: once the direct path was addressed, the model could search for a position that better matched the preferred illuminance range.
Coordinate illuminance, glare and the view outside
Venetian angles were evaluated in 10° increments. In summer the routine began from the closed state and opened the slats toward the preferred range; in winter it began from the open 10° state and moved toward closure. These candidate positions allowed the model to connect daylight admission with the season’s thermal preference.
The roller was then evaluated at 30 cm deployment intervals using DGP at the modeled occupant viewpoint. The routine sought the least coverage that brought DGP below 0.30. A view target of at least 70% of the possible outside view guided the decision, with glare protection taking priority. The preferred illuminance band therefore guided the balance among several requirements, rather than acting as the only condition.
When the modeled occupant was absent, an energy-oriented branch used outdoor temperature and departure from adaptive thermal comfort to select the shade state. The controller therefore changed its purpose with occupancy: visual conditions and user control when occupied, solar and thermal management when unoccupied.
Keep the decision sequence inspectable
The controller combined the branches into one decision sequence: presence, direct sun, preferred WPI, directional glare, view, and shade movement. Each branch corresponded to a calculable condition or a user command. This made the operating logic traceable from an input to the final position of the two devices.
Manual override provided another route to the same geometry. The user could select a preferred state directly, with the associated condition and preference available as an additional observation for the preference workflow. The automated and manual paths therefore shared the model rather than operating as unrelated demonstrations.
Demonstrate the phone-to-model control path
I built a phone-based OSC/MIDI interface and connected it to Grasshopper through gHowl. An automatic/manual switch selected the operating mode, a fader controlled the roller, and a radial fader changed the Venetian angle. The interface was tested with the digital model so its geometry responded to the phone commands on screen.
The original video shows this connection in operation. It is the final link between the parametric design and an occupant-facing interface: the user can see and change the same two shade elements that were assessed in the simulation and optimization stages.
February 21 and August 21 · four modeled development stagesResults: from the existing room to final control
Compare the same room at four stages
The final evaluation compared the existing room, static shading, the initial automated system after geometry optimization, and the final automated controller. February 21 and August 21 provided winter and summer comparison days. Keeping the room and dates consistent made it possible to see what changed when geometry selection was followed by the personalized control logic.
The work-plane curves show the progression from high daylight peaks toward a more controlled daily profile. The final controller reduced the peak illuminance while using the changing preferred range to guide its response. The plotted upper and lower preference boundaries explain why a suitable position depends on the condition at that hour, while the fixed thresholds provide a second reference for reading the result.
| Stage | Configuration | What the comparison shows |
|---|---|---|
| 01 · Existing room | Unshaded reference condition | Daylight and glare peaks to address |
| 02 · Static shading | A fixed shading configuration | Effect of solar interception without responsive control |
| 03 · Initial automation | Control applied to the optimized geometry | Effect of linking the selected geometry to movement |
| 04 · Final automation | Preferred illuminance, direct sun, DGP and view coordinated | Final daily response across visual, thermal and lighting outputs |
Suppress glare peaks through viewpoint-based control
The same four-stage comparison was then read in terms of DGP. The existing, static and initial automated configurations show periods with elevated glare probability; the final configuration controls those peaks using the roller-position evaluation. This provides a direct result of adding the viewpoint-based glare check after the geometry optimization.
The illuminance and glare plots should be read together. A daylight peak can be reduced through slat movement, but glare from a visible light source may require additional roller coverage. The final sequence coordinates those actions, with the outside-view target and manual override maintaining the occupant’s role in the decision.
Read the thermal effect alongside the visual outcome
Shading also changes how much solar energy enters the room. The thermal comparison plotted the total thermal load and operative temperature for the same four stages, alongside the solar context. The resulting profiles show smaller changes in these outputs than in the glare peaks, which makes the balance among daylight, glare and thermal performance visible.
This stage connected the façade decisions back to building performance. The selected geometry and final controller were assessed as part of the room’s daily thermal response, while the visual results explained when their movements were useful to the occupant. The February 21 and August 21 results retain that two-day scope.
Use available daylight to reduce the electric-lighting schedule
The lighting-control routine combined modeled presence with the daylight available at the work plane. Artificial lighting compensated when daylight was insufficient and dimmed when it could provide the required contribution. The final schedule comparison shows this operating pattern against a conventional office-lighting schedule for the two evaluation days.
Taken together, the development produced a selected dual-zone geometry, a preference-based controller, and a demonstrated phone-to-Grasshopper interface. The staged results connect each addition to its purpose: the strategy comparison established the device combination, optimization resolved the physical trade-offs, and the controller improved the daily handling of illuminance and glare while coordinating the lighting and thermal response. The five-strategy comparison and geometry optimization were subsequently published in the Journal of Green Building; the complete preference and controller development is documented in the thesis.
The complete project record
Full book
Read the complete report, thesis, or presentation here. Use the page controls to explore the original figures and methods.
User-Specific Automated Kinetic Shading Design
The original document, page by page.
The reader loads when you open the book.
