Agent-based autonomous pollution source localization for complex environment
- PDF / 1,020,288 Bytes
- 9 Pages / 595.276 x 790.866 pts Page_size
- 38 Downloads / 229 Views
ORIGINAL RESEARCH
Agent‑based autonomous pollution source localization for complex environment Dehu Xiao1 · Yong Wang1 · Zhuo Cheng1 Received: 25 June 2020 / Accepted: 7 November 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Pollution source localization is important prerequisites for source control, pollution isolation, purification, and emergency evacuation. Most available studies on pollution source localization have not comprehensively investigated practical problems, such as location accuracy, obstacle avoidance, energy saving, and motion control. Besides, some multi-agent olfaction methods result in wasting resources in small detecting area. Aiming at the issues, this study presents a single agent-based autonomous approach to pollution source localization for the complex environment. In our solution, the NEAT algorithm is used to generate different networks with different weights and structures, the neural network is used to control the agent according to the concentration and distance information, and the fitness function is set to choose the optimal neural network with the best performance. In particular, the fitness function is constructed by combining the pollution concentration and the residual energy of the agent to consider both positioning accuracy and energy saving. Comprehensive simulations show that the proposed method can find the optimal neural network under energy constraints and the agent under the control of the optimal neural network can avoid obstacles and accurately locate the pollution source in different complex environments. Keywords Pollution source localization · Complex environment · NEAT · Agent
1 Introduction The control of pollution necessitates the development of effective source localization methods, which are important prerequisites for source control, pollution isolation, purification, and emergency evacuation (Zhai et al. 2012; Luo and Yang 2015; Li et al. 2015). The existing source localization methods can be roughly divided into two types: stationary wireless sensor network (WSN) methods (Boubrima et al. 2015; Luo and Yang 2019) and agent-based olfaction methods (Bayat et al. 2017; Chen et al. 2017). In a stationary WSN method, many wireless sensors need to be evenly distributed in the detected environment in advance. By applying kinds of mathematical-statistical calculations into the data collected by these sensors, the pollution source can be pointed out. The method mainly has two modes. One is that the pollution source is the central value of the data * Yong Wang [email protected] 1
School of Mechanical Engineering and Electronic Information, China University of Geosciences, Wuhan, China
collected by all sensors at a moment, including the closest point approach (Lim et al. 2008), the centroid localization algorithm (Thammavong et al. 2018), the contour localization algorithm (Luo et al. 2016), etc. Although it does not need to build any diffusion model, the localization accuracy is not high enough. The other mode is that the location o
Data Loading...
