DOMS: Dynamic Occupancy Management System - a space management technique for smart buildings

Space management in smart buildings requires expert knowledge and demands manual intervention. Analysing the space, choosing spots to deploy sensors are manually done by experts even in this modern era. Though planned by experts there are inefficiencies in sensor placement and deployment. To address this issue, we propose a framework for better space management in smart buildings called as DOMS (Dynamic Occupancy Management System). In this framework, we have integrated computer vision, language model and graph based intelligence for achieving the smart space management solution. We have utilized YOLOv5 for extracting the structural and geometric features from the floor plan. These extracted features are used to build the Spatial Knowledge Graph (SKG) where SKG helps in finding semantic relationship between the identified objects in the floor plan. This SKG is then processed by a fine-tuned LLaMA 3.2 model that performs semantic reasoning to recommend suitable sensor types, optimal placement strategies, and communication modules based on contextual constraints such as room purpose and occupancy flow. This hybrid reasoning enables DOMS to achieve adaptive, explainable, and automated configuration of IoT-based smart environments. Experimental results demonstrate improved coverage accuracy, reduced manual intervention, and enhanced scalability compared to conventional occupancy systems.