Optimal Sensing In Sensor Networks

The sensor networks are composed with many sensor nodes, which are deployed inthe scene. The sensing data are obtained by the sensors, and are transmitted to the serverfor processing, for example data fusion, object detection,etc. The sensor networks arewidely used in many application fields, such as traffic management, security monitoring,agricultural production, and virtual reality, etc.The optimization computing plays an important role in many key aspects of sensorsnetworks, such as nodes deployment, routing protocols, and localization, etc. In this pa-per, we aims to solve nodes deployment in visual sensor networks and the localization forgeneral sensors, which are both originally non-convex optimization problems. The maincontributions of this paper are listed as follow:1. We firstly focus on the visual sensor networks. The multi-agent genetic algorithmbased sensor layout algorithm was designed to optimize the observation quality inthe scene. As a visual sensor has a bounded field of view and easily be obscured,a random deployment of network sensors cannot solve this issue. Aiming at theproblem above, the imaging specialty of visual sensor is analyzed. An anisotropicbounded observation field sensing model of visual sensor is proposed firstly. Thismodel can describe the sensing feature of visual sensor. On the basis of this mea-surement, the sensor placement methods are devised by means of multi-agent ge-netic algorithm (MAGA) to optimize the deployment of sensor nodes. The coor-dinates of sensor nodes is firstly coded and inputted to the MAGA scheme. Theenergy function measures the distribution quality of observation source, and theoptimal placement is obtained aiming to maximize the energy function. The posi-tions and poses of sensors which can enhance the coverage can be worked out withhigh efficiency. Thus the visual network’s capability of obtaining information canbe ensured.2. Based on the first part job, we further propose a sparse representation based sen-sor deployment algorithm. Moreover, a more accurate sensing model was designedbased on the physical imaging process. Nodes deployment is a critical issue af-fecting the quality of service of visual sensor networks. The deployment aims at adopting the least number of visual sensors to cover the whole scene with expectedobservation quality. This is generally formulated as a non-convex optimizationproblem which is hard to be solved in polynomial time. In this paper, we proposean efficient convex solution for deployment guaranteeing satisfactory observationquality based on a novel anisotropic sensing model of visual sensor. We constructthe sensing model by exploring the relationship between visual sensor parametersand the imaging quality. This model provides a reliable measurement of observa-tion quality for the directional nonuniform sensing field of visual sensors. In addi-tion, the deployment is treated as selecting the most sparse subset of nodes from aredundant initial deployment with numerous visual sensors. Based on this charac-teristic and the sensing model, we formulate the deployment as an minimizationproblem which is further relaxed to a convex form. Therefore, the high qualitydeployment is efficiently obtained via convex optimization.3. Finally, we propose an efficient localization algorithm, which is also an critical is-sue in sensor networks. The high speed and accuracy localization is achieved basedon a special kind of linear combination (convex combination). Both distance mea-surement based and angle of arrival based localization scheme are designed. Thelocalization is generally formulated as an optimization problem to tackle the noisymeasurements. However, the objective is non-convex, and thus localization is diffi-culttosolveinitsoriginalform. Inthispaper,aconvexobjectivefunctionisderivedbased on a linear combination scheme, within which the target position is expressedas a linear combination of positions of virtual anchors around its real position. Inaddition, the linear combination provides a highly accurate approximation for thecomputationofthedistance/anglefromtheanchors. Thus,thelocalizationisformu-lated as a convex problem to find the optimal coefficients of the linear combinationand is solved efficiently by the linear least square method.