Intelligent Shading System

Conventional shading is often optimized for one performance target or controlled with a fixed threshold. This M.Sc. thesis developed a user-specific automated kinetic shading workflow that connects façade geometry, daylight, glare, energy, thermal conditions, and observed preference. The project was completed with Saeideh Soltanmohammadlou and Amir Kianfar at Politecnico di Milano.

A home-office case study established the building, climate, façade, occupancy, and simulation assumptions. Daylight, energy, glare, and adaptive-comfort baselines provided a common reference for later strategy comparisons and keep the resulting performance claims traceable.

Five strategies were compared: no shading, fixed exterior shading, a motorized interior roller, a combined motorized roller with fixed exterior shading, and a combination of motorized interior and exterior shading. The comparison shows why no single position can maximize daylight, limit glare, reduce loads, and satisfy every user at the same time.

The workflow used Rhino and Grasshopper to define the façade and shading geometry, then connected that model to Ladybug/Honeybee, Radiance-based daylight analysis, and energy simulation. Parametric control made it possible to vary design genes consistently and send comparable alternatives into the optimization process.

An NSGA-II evolutionary search explored candidate geometries across competing objectives. Instead of declaring one mathematically perfect answer, the workflow produced a family of trade-off solutions and a transparent selection process. The resulting Pareto set connects candidate performance with the geometry genes and objectives used to produce it.

A four-month preference dataset connected environmental conditions with user choices. Regression translated those observations into a case-specific control input, allowing the automation concept to respond both to modeled performance and to the person occupying the space.

The control logic combines view factor, work-plane conditions, presence, direct sun, average illuminance, directional glare probability, user preference, and manual override. A decision tree determines how the system responds when the user is present or absent and preserves a clear path for user control.

The project archive connects the parametric definitions, simulation logic, control workflow, and automated-system demonstration. A short video records the responsive sequence, while selected Grasshopper views make the relationships between geometry, performance inputs, optimization, and control legible without exposing the full working directory.

The project demonstrates a reproducible path from façade geometry to simulation, optimization, preference modeling, and control. It identifies candidate high-performing configurations and shows how user-specific logic can be incorporated into adaptive shading. Results are comparative rather than a full-year field validation. A peer-reviewed article was published in the Journal of Green Building in 2024, DOI 10.3992/jgb.19.1.123.