Towards Locating Time-Varying Indoor Particle Sources: Development of Two Multi-Robot Olfaction Methods Based on Whale Optimization Algorithm
Yibin Yang,Boyuan Zhang,Qilin Feng,Hao Cai,Mingrui Jiang,Kang Zhou,Fei Li,Shichao Liu,Xianting Li
DOI: https://doi.org/10.1016/j.buildenv.2019.106413
IF: 7.093
2019-01-01
Building and Environment
Abstract:Source localization is crucial for controlling indoor particle pollution. Locating indoor particle sources is challenging because the dispersion of particles is more complicated than that of gases, and the release rates of particle sources usually change with time in real-world applications. This study presents two multi-robot olfaction methods based on the newly emerging whale optimization algorithm (WOA), namely, the standard WOA (SWOA) and improved WOA (IWOA) methods, for locating time-varying indoor particle sources without and with airflow information, respectively. By combining experiments and CFD simulations, the presented methods were validated and compared with two particle swarm optimization (PSO)-based methods, namely, standard PSO (SPSO) and improved PSO (IPSO) methods. Four typical scenarios, including two time-varying source types (decaying source and periodic source) and two ventilation modes (displacement ventilation and mixing ventilation), were simulated and exported as virtual environments to test these methods. The methods were evaluated by the success rate (the number of successful experiments divided by the number of total experiments) and the average localization time of the experiments. The results showed that the SWOA method outperformed the SPSO method with a higher success rate (SWOA: 66.00%, SPSO: 52.00%) and a less average localization time (SWOA: 65.48 s, SPSO: 69.65 s) for all four scenarios. The IWOA method performed slightly better in success rate (IWOA: 97.75%, IPSO: 97.00%), while the IPSO method performed slightly better in average localization time (IWOA: 42.18 s, IPSO: 39.18 s) for all four scenarios. In addition, the most cost-effective anemometer was also determined.