Research

To investigate the empirical implementation procedure and applicability of simulation automation, an automatic simulation system is developed using an intelligent human object database based on the residential units. The human object database includes the three modules of a human object, activity, and behavioral rule that can be input via the web server and connected with the spatial data from BIM. In the BIM module, the BIM add-in extracts the semantics of the given space by generating 3D bounding boxes for architectural components so that the agents perceive the room tags and furniture. 

Both input data are transferrable into the simulation system. The simulation system allows the users to measure the proposed design performance with visualized  (circulation, heat-map) and quantitative analytics (privacy, work efficiency, household efficiency, socializing, autonomy). Through this process, this study proposes the automated simulation generating BIM alternatives and simulating the alternatives to search for the optimum one by measuring building performance. All of the performance measurement and analytic image data are saved into the analytic database.

This project developed two different agents' behavior schedule models for integrating semantics  of BIM and human object data. The first model is a manual input model, which allows simulator users to adjust residents' behavior schedules through schedule customization. The second model is an LLM-based schedule generation model that integrates ChatGPT with the simulator.

Epilogue |   we conducted empirical tests to practically evaluate the simulator, involving both undergraduate students and industry practitioners.