Energy-aware Contour Covering Using Collaborating Mobile Sensors Sumana Srinivasan, Krithi Ramamritham and Purushottam Kulkarni Department of Computer Science and Engineering Indian Institute of Technology Bombay, Powai, Mumbai, India Email: {sumana,krithi,puru}@cse.iitb.ac.in Abstract Environmental sensing systems are useful in the area of disaster management where the system can provide alerts as well as remediation services in the event of a disaster such as pollutant spills. Recent advances in robotics technology have led to the development of sensors with the ability to sense, move, and we show how such sensors can be deployed to perform tasks such as locating and covering a hazardous concentration contour in a pollutant spill. As mobile sensors consume energy for movement and have limited available energy, it is important for the sensors to plan their movement such that the coverage is maximized and the error in coverage (i.e., area not relevant to the contour) is minimized. In this paper, we address the nontrivial problem of continuously adjusting the direction of movement by presenting a dis- tributed algorithm where a sensor determines the direction of movement by (i) estimating the distance to the contour and perimeter of the contour using locally sensed information as well as information gathered through collaboration and (ii) deciding dynamically whether to directly approach the contour or surround the contour based on the amount of work remaining to be done by the sensor vis a vis the energy remaining. We show that our proposed algorithm has the best coverage vs. error trade-off when compared to algorithms that directly approach or surround the contour. 1. Introduction Cyber physical systems help in (a) acquiring useful in- formation about physical phenomena, e.g., spatio-temporal distributions of pollutant concentration in a spill, (b) pre- dicting their short- and long-term behaviour and (c) reacting to changes in the phenomenon, i.e., performing remedial actions. A contour or a level set comprises of a set of points of equal value of a specified parameter, e.g., concentration, temperature, etc., in a field. Contours are useful in charac- terizing features such as the areas of greatest contamination in a pollutant spill (e.g., isoplats indicate acid precipitation), mapping of air pollution, etc. Consider the scenario of an oil spill from a tanker. Subsequent to its occurrence, the most important question is: Using a network of mobile sensors can we determine the location and extent of the the most hazardous contour in the spill and cover it accurately before the energy of the system dissipates? While it is important to completely cover the hazardous contour, it is equally important to minimize the area of coverage that is not relevant to the contour (we refer to the latter area as the coverage error) for the best utilization of resources like booms and antipollutant chemicals. The problem of covering the contour such that coverage is maximized and coverage error is minimized is the focus of this paper. Today’s technology involves human hazard management teams using images obtained from remote sensing to de- termine the location and the extent of spill [1]. While a remote sensor spans large geographical regions, measurement accuracy may be limited due to the lack of proximity to the phenomenon, inclement weather conditions and high cost of deployment. The advent of wireless in-situ sensor networks capable of sensing parameters and communicating has led to the design of systems that focus on performing contour estimation with maximizing accuracy and minimizing the cost of estimation as main goals. As the static sensors lack the ability to move they can utmost perform contour estimation (determine points on the contour) and cannot be used for physically covering the contour (sensors moving to bound the contour) or to perform remedial actions. The accuracy of estimation is dictated by the density of the network [2] and maximizing coverage requires deployment of a large number of sensors. In the case of a moving oil spill, static sensor nodes have to be redeployed if the spill moves out of coverage and this may be prohibitively expensive and time consuming. These limitations are addressed by mobile sensors which have the ability to move, sense, collaborate and can be used to approach and surround the contour in time before the energy of the system is dissipated. With recent advances in robotic technology such mo- bile sensors have already been prototyped and tested, e.g., SOTAB-1 (Spilled Oil Tracking Autonomous Buoy) [3] (see Figure 1(a)): a mobile buoy is equipped with in-situ as well as image sensors which are used to locate the spill. But, since the energy consumption in a mobile sensor is mainly due to movement, the challenge in designing a mobile sensor network for contour covering lies in determining the direction of movement towards and around the contour such that the coverage is maximized and coverage error is minimized under energy constraints. The decision depends upon the amount of information available to the sensors. If only the measurement