Gaussian process based IAQ distribution mapping using an interactive service robot

Article type: Research Article

Authors: Qian, Kun^{a; b; *} | Ma, Xudong^{a; b} | Dai, Xianzhong^{a; b} | Fang, Fang^{a; b} | Zhou, Bo^{a; b}

Affiliations: [a] Key Lab of Measurement and Control of Complex Systems of Engineering, Ministry of Education, China | [b] School of Automation, Southeast University, 2 Sipailou, Nanjing, China, 210096

Correspondence: [*] Corresponding author. E-mail: kqian@seu.edu.cn.

Abstract: In recent years, how poor Indoor Air Quality (IAQ) affects the residents and staff in living homes has raised many concerns. For the upcoming ambient assisted living facilities, the monitoring of IAQ pollution levels in indoor environments is of great significance for ensuring the health and comfort of individuals. Instead of the widely used wireless sensor network or wearable sensors, in this paper we consider the use of interactive service robots as a natural way to learn the environmental map as well as the distribution of Volatile Organic Compounds (VOCs). In order to map the VOC distribution in a framework that is consistent to the room coordinate, even if the environment is unknown to the robot, we designed an Initial Guided Mapping Mode which allows the robot to effectively map the environment obstacles by human guidance using a laser sensor and a RGB-D sensor. In the following Autonomous IAQ Mapping Mode, the robot carried a tailored sensor board for sampling a number of VOC measurements for learning the spatial distribution. A key problem is to deal with uncertainties brought by mobile sensing with co-existing people. Thus we applied Gaussian processes combined with Monte Carlo sampling method for modeling VOC distribution as well as querying concentration at uncovered locations. By incorporating the method in predicting the posterior predictive statistics of the VOC distribution, more accurate VOC distribution mapping is achieved in our implemented system. We provide experimental results in several testing scenarios to demonstrate the system performance. The proposed approach not only outperforms the traditional method, but is also human-friendly and easy to install. The method is feasible for applying service robots to predict the IAQ distribution or the spatial concentrations of any interest during its long-term operation in residential environments.

Keywords: Indoor Air Quality, service robot, Gaussian process, VOC distribution modeling, human-robot interaction

DOI: 10.3233/AIS-160376

Journal: Journal of Ambient Intelligence and Smart Environments, vol. 8, no. 3, pp. 359-373, 2016