Occupancy Estimation in Semi-Public Spaces using Sensor Fusion and Context Awareness

Accurate occupancy estimation is of great importance in managing indoor comfort demands and energy costs in semi-public buildings like academic institutes, offices, etc. Large spaces with multiple occupant seating are common in such buildings. Sophisticated sensors used for occupancy detection and estimation in these spaces, like cameras and motion sensors, may prove expensive and intrusive. Contrarily, simple devices like a door sensor or a Passive Infrared (PIR) sensor, if employed individually, might not be accurate enough. In this work, we exploit the unique advantages of each of these low-cost, off-the-shelf sensors to propose a novel heuristic algorithm for occupancy estimation. We develop an analytical model to probabilistically derive the uncertainty associated with a door sensor. We propose an effective and unique way to enhance the capability of the door sensor by fusing it with a PIR sensor and context information in terms of occupancy memory of the space. The hardware test bed was deployed in an academic institute, that mimics a work space in semipublic building. The performance comparison of the sensors and proposed algorithm is presented in terms of probability of detection and errors.