Occupancy and Direction Estimation Based on Grid Search Approach

Abstract

The goal of this research is to maximize the use of energy in buildings using sensor-based occupancy detection, a process in which the objective is to determine the number of occupants in a room. In this research, sensors are installed to identify the number of occupants and the route they map while moving across a room. Occupancy data was also collected manually (no. of persons in 10-minute increments) to revalidate the sensor data. To determine estimated direction and occupancy, a grid search approach is the most effective to implement. Grid Search is a machine learning tool used to search through sets of data and train the model to output certain combinations of results based on the needs and behaviors of an individual. To estimate occupancy and direction, first, sensor plot matrices were integrated into Python scripts. Then, Python dictionaries were created to distinguish the values of the sensors and sort out the data. Lastly, a Grid Search was conducted on all of the unsearched coordinates, generating needed coordinates which were, then added to an external output list. To determine different sensor paths (connected islands of activated sensors), an algorithm was developed in this research. The algorithm consists of a series of steps; validating only sensors that are part of a motion path, excluding sensors that are not, and calculating the start & end position of the path. Using this information, an animated motion trail can be produced in order to visually depict the movement and number of the occupants.
Keywords - Occupancy Detection, Grid Search