Exploiting Bacteria Swarms for Pollution Mapping John Oyekan, Huosheng Hu and Dongbing Gu School of Computer Science and Electronic Engineering University of Essex, Wivenhoe Park, Colchester CO3 4SQ, U.K. Email: jooyek@essex.ac.uk; hhu@essex.ac.uk; dgu@essex.ac.uk Abstract—Inspired by the simplicity of how nature solves its problems, we develop a flocking controller that would enable the localisation and subsequent mapping of environmental pol- lutants. Pollutants could range from checimal leaks to invisible air borne harzardous materials. We use simulation results to validate our approach and then briefly discuss how to implement the controller onto a real robotic platform. Our motivation is to use the advantages offered by swarm robotics- simple, multiple and cheap agents- to achieve a collective complex single goal of mapping an environmental pollutant spread over a large area. We aim to make our approach as simple as possible yet highly effective in generating the map. Keywords Bacterium Inspired Algorithm, Environmental Monitoring Flocking. I. I NTRODUCTION The global warming has been a major topic in the en- vironmental sciences in the past few years, which has led researchers in various fields to find new novel ways of monitoring the environment. One of the novel ways is the use of various multi agent theories of which swarm robotics is one of them[1]. Swarm Robotics involves the use of many agents to perform a task that is not possible or that is very difficult to achieve with a single agent[2]. By using swarm robotics, multi agents could monitor the pollution spread and how it changes with time. In addition, if one of the agents fails, the mission would still carry on with little loss in performance of the entire system. Furthermore, using swarm robotics enables the system to be everywhere at once and also enables the condition of pollutants to be viewed at various locations all at once. It also enables users to view how conditions at one location affect conditions at other locations in real time. From this, a real time 3D map of the pollutant and changes in its condition can be generated. In order to build a map of environmental pollutants using multiple robotic systems, two problems have to be solved. Firstly, a robotic agent has to be controlled in such a way to make sure that it is placed at the best position in the environment to get a good reading of an environmental pollutant. Secondly, a way of building a map of the measured quantity is needed. In this paper, we address the first problem. There have been numerous research work done in this area. Cortes et al in[3] use a voronoi approach to divide an area of interest into voronoi partitions and then control the robotic agents to place themselves in the center of each voronoi partition. This they argued enables them to get the optimal reading in that sector of the environment. However, this approach requires a high computational and communication cost and can only be used in polygon derivative environments. Shucker et al in[4] use a gabriel graph theory to achieve the effective placement of agents in the environment to track a target. This approach requires that agents are able to communicate to an extent over a large distance with agents at other positions. This might not be practical in a real world scenario. Lilienthal et al achieve an effective coverage of an area by moving their agents in a predefined manner in the area to be covered[5]. This method quickly becomes ineffective if a large distance is to be covered by the single agent. Work done by Mesquita et al uses a technique based on the bacteria chemotaxis behaviour to arrange them in the environment based on the signal to be monitored[6]. However, in their work, they assumed that the structure of the signal to be monitored is known before deployment of their agents. Zarzhitsky et al. used a term called fluoxtaxis to direct a swarm of robotic agents in localizing a plume source. This term was also inspired partly by the chemotaxis be- haviour of bacteria. However, they used an artificial physics framework to achieve the flocking behaviour of the swarm which is arguably not biologically plausible[7]. In our work, we investigate the use of a bacterial chemotaxis behaviour in combination with flocking algorithms to position the agents in the environment based on the density profile of the environmental pollutant to be measured. It is our aim to arrange the agents so that areas of high environmental pollutant concentration receive more agents than areas of low pollutant concentration. We believe that by doing this, areas with more interesting data are monitored closely than areas with less interesting data[8]. This makes it possible for agents to keep searching for more interesting data if there is a possibility of any appearing in the environment. In addition, we believe that our algorithm is simple to implement and not environmental specific. In this paper, we present results of using a combination of swarm algorithm and bacterial algorithm to achieve the mon- itoring of pollutants. The developed controller is presented in Section II. Section III discusses the experimental setup while Section IV presents the results of our simulation. A 978-1-4244-4775-6/09/$25.00 © 2009 IEEE. 39 Proceedings of the 2009 IEEE International Conference on Robotics and Biomimetics December 19 -23, 2009, Guilin, China