A Wireless Sensor Network (WSN) is composed of sensor nodes that are located within a certain target region, which are capable of monitoring different environmental phenomena such as forest fires detection, leakage pollutants, etc. In a WSN, sensor nodes are distributed across the Region of Interest (ROI). The overall coverage of the region covered by sensor nodes directly influences the efficiency of the network in obtaining useful information. Establishing global coverage in the network using the least number of sensor nodes plays a very important role in reduction of the costs associated with the wireless sensor network. However, coverage holes appear in the study region due to various reasons, such as non-uniform distribution of sensors, node failure and energy dissipation.
Existing methods to heal the holes in the study area are generally divided into global and local methods. In global methods, prior to distribution of sensor nodes in the desired region, the optimal position of the nodes in order to avoid the occurrence of coverage holes is determined using optimization functions. In local methods, on the other hand, after primary distribution of the sensor nodes in the region, coverage holes are detected and the sensors are then moved locally in order to cover the detected coverage holes. The latter methods are mostly designed based on geometrical structures such as Delauney triangulation and Vorosnoi diagrams.
This present study focuses on determining optimal locations in order to achieve global coverage in a wireless sensor network. For this purpose, the performance of an existing local method, called tree-based method, in adding new sensors in a wireless sensor network has been evaluated. The results show that, despite the generally acceptable performance of this method, it does not work well in some special configurations of sensors, which is mainly due to lack of attention to the position of the sensors when the sensors are added. Therefore, by combining the tree-base method with a proposed method, called center of gravity method, the results are improved so that it has the ability to properly cover the holes with different sizes and geometrical shapes. Moreover, in the proposed method, before identifying the coverage holes in the area, an improvement phase is applied to improve the configuration of existing sensors after they are randomly distributed in the target area. Here, nearby sensors are moved away in order to prevent the occurrence of large scale coverage holes as well as prevent overlapping coverage of the sensor nodes. Therefore, this phase has an effective operation in creating optimal coverage in the area. Moreover, the results show that the combined method has a better performance compare to the tree-base method, because it covers the region in less iterations and less number of new sensors.