پديد آورندگان :
ارگاني، ميثم دانشگاه تهران - دانشكده جغرافيا - گروه سنجش از دور و GIS , كريمي پور، فريد دانشگاه تهران - دانشكده مهندسي نقشه برداري و اطلاعات مكاني - پرديس دانشكده هاي فني , مافي، فاطمه دانشگاه تحصيلات تكميلي صنعتي و فناوري پيشرفته كرمان
كليدواژه :
شبكه هاي حسگر بي سيم , جايابي حسگر , پوشش شبكه , الگوريتم هاي بهينه سازي سراسري , مدل احتمالي پوشش , مدل رستري , مدل برداري
چكيده لاتين :
Introduction
Wireless Sensor Networks (WSNs) are widely used for monitoring and observation of dynamic phenomena. A sensor in WSNs covers only a limited region, depending on its sensing and communicating ranges, as well as the environment configuration. For efficient deployment of sensors in a WSN, the coverage estimation is a critical issue. Probabilistic methods are among the most accurate models proposed for sensor coverage estimation. However, most of these methods are based on raster representation of the environment for coverage estimation which limits their quality. In this paper, we propose a probabilistic method for estimation of the coverage of a sensor network based on raster models, and 3D vector representation of the environment. Then, the performance of global approaches are evaluated, and the 3D vector model is used as an appropriate model.
Materials and Methods
Recent advances in electro mechanical and communication technologies have resulted in the development of more efficient, low cost and multi-function sensors. These tiny and ingenious devices are usually deployed in a wireless network to monitor and collect physical and environmental information such as motion, temperature, humidity, pollutants, traffic flow, etc. The information is then communicated to a process center where they are integrated and analyzed for different application. Deploying sensor networks allows inaccessible areas to be covered by minimizing the sensing costs compared to the use of separate sensors to completely cover the same area. Sensors may be spread with various densities depending on the area of application and details and quality of the information required. Despite the advances in sensor network technology, the efficiency of a sensor network for collection and communication of the information may be constrained by the limitations of sensors deployed in the network nodes. These restrictions may include sensing range, battery power, connection ability, memory, and limited computation capabilities. These limitations have been addressed by many researchers in recent years from various disciplines in order to design and deploy more efficient sensor networks.
Efficient sensor network deployment is one of the most important issues in sensor network filed that affects the coverage and communication between sensors in the network. Nodes use their sensing modules to detect events occurring in the region of interest. Each sensor is assumed to have a sensing range, which may be constrained by the phenomenon being sensed and the environment conditions. Hence, obstacles and environmental conditions affect network coverage and may result in holes in the sensing area. Communication between nodes is also important. Information collected from the region should be transferred to a processing center, directly or via its adjacent sensor. In the latter case, each sensor needs to be aware of the position of other adjacent sensors in their proximity.
In recent years, Wireless Sensor Networks (WSN) has been studied in several applications such as monitoring and control different criteria from smart cities and intelligent transportation to land use planning and environmental monitoring. Sensor deployment for achieving the maximum coverage is one of the important issues in WSN. Hence, several optimization algorithms to achieve maximum coverage are used in the majority of researches.
Discussion and Results
In a general classification, optimization algorithms for the sensor deployment with the aim of increasing coverage, are divided into local and global optimization algorithms. The feature of global algorithms is their randomness based on an evolutionary process. In all of these algorithms, the calculation of the sensor network coverage is essential as a target function. In fact, coverage improvement is done according to the coverage calculation method. In the previous researches, a simple model was considered as the environmental model for network sensors. In this research, raster and vector modeling in 2 and 3-dimensional spaces and the optimization algorithms of global performance for optimizing the sensor layouts were compared evaluated. errorIn this study, two-dimensional and three-dimensional vector models were used as a precise environmental model. Most of the models in the previous studies considered the coverage to be binary (i.e. a point is covered by a sensor or not). For realistic modeling, this study considers the coverage as an issue, which means that the amount of coverage obtained based on parameters such as distance and angle of the sensor is expressed as a percentage between zero and one hundred. errorIn fact, all sensors are not sensed in the same way and will vary according to their various parameters. Since the purpose of this study is to compare the performance and ability of global optimization algorithms, it is therefore assumed that the study area has equal conditions. In this paper, several optimization methods such as genetic algorithms, L-BFGS, VFCPSO and CMA-ES have been implemented to optimize the location of sensors. In this study, various sensor sensing types such as omnidirectional binary sensing model, directional sensing model and probabilistic sensing model have been used and tested for the aforementioned optimization algorithms in different Raster and Vector study areas.
Conclusion
This paper was focused on comparing the performance of four global optimization algorithms to optimize deployment of sensors in environment using more spatial details compared to previous approaches. The innovation of this study was to use 3D raster and vector data and to implement the global optimization methods using probabilistic sensing model to optimize sensor network placement. Finally, promising results have been presented and discussed and future methods were introduced.
