Supervised training data generation for interior environment simulation

A method of automatically generating supervised training data for a control system for an interior environment, comprising the steps of:

• performing, by a processor, simulation calculations using a three-dimensional parametric computer-aided design (CAD) model of a virtual interior environment to determine spatial and temporal distributions of environmental conditions within the virtual interior environment;
• simulating, by the processor, virtual sensors at specified positions within the virtual interior environment and calculating predicted environmental conditions at the specified positions;
• generating, by the processor, output control signals for representations of virtual environmental control hardware within the virtual interior environment;
• and
• training an artificial intelligence (AI) based controller for the interior environment using the predicted environmental conditions and the output control signals.

A method of automatically generating supervised training data for a physical controlled environment control system comprising the steps of:

• obtaining measurements from a physical sensor array in the environment;
• calculating, using the obtained measurements, control outputs for physical control hardware that controls the environment;
• generating a set of spatially dense virtual sensor array values based on the obtained measurements, wherein the spatially dense virtual sensor array is more dense than the physical sensor array;
• and
• training a physical artificial intelligence based controller in the physical controlled environment control system using the obtained measurements and the generated set of spatially dense virtual sensor array values.