Optimizing indoor air quality (IAQ) in educational environments not only can improve occupants' health and productivity but also can provide assistance to optimize energy constraints. One of the effective strategies is the optimal placement of portable air purifiers, which can help with pollutant removal and uniform air distribution. This work presents a novel approach using SIMOC-inspired sensing and networking architecture that couples a mobile robot with a dense chamber sensor grid. This will help to learn where the purifier should be positioned by experimentation. A mesh of fixed nodes and the robot continuously measured CO2, particulate matter, temperature, and relative humidity. Next, data was transported over MQTT and aggregated for live oversight. The purifier was moved over a predefined set of locations on the floor. During each movement, a set of constant sensors and a robot were employed to record the data during each placement. In order to intelligently select the most promising purifier location for each new test based on the results of previous ones, Combinatorial Bayesian Optimization was used. Then, the performance of the model was compared against simpler baseline strategies, such as random placements and locations typically recommended by the literature, to measure the effectiveness of the model. Using the mobile-plus-grid workflow, pollutant concentration fields were rapidly mapped across all trials. As a result, superior purifier placements were identified where both space-average and peak concentrations were reduced. Furthermore, even within a limited number of experiments, spatial uniformity could be improved compared to baseline operations. The system is readily extensible to additional robots on the same mesh to accelerate exploration, recognizing deployment time as a practical constraint. The results indicate that data-driven placement, including employing real-time data measurements, can be used to configure optimum placement for a portable purifier to improve IAQ levels in educational spaces.
Chahardoli et al. (Tue,) studied this question.