An adaptive façade that balances comfort, daylight, glare and energy
A deep-learning surrogate is trained in your browser on building-performance simulation data, then deployed inside a model-predictive controller that re-optimises the shading configuration in closed loop, responding in real time to the occupant.
Surrogate model · the intelligence inside the controller
Deep-learning proxy for building-performance simulation, trained in your browser and deployed in closed loop
Initialising
Environment
Simulated sensor readings for the operating location
Operating room · east façade · Wellington NZ
Simulated sensor feedback keeps the loop closed
Current control decision
Surrogate-predicted vs simulated response
Occupant preferences
Express a desired indoor state · the controller models the next-best façade action
At the energy-efficient operating baseline.
Closed-loop performance
Recorded telemetry across control steps
Actuation log
Recent control steps · observe, predict, optimise, actuate
Waiting for the first control step…