Recent advances in wireless communication and electronic has enabled the development of low-cost wireless sensor networks. This type of network is especially useful in tasks of target detection and environment monitoring. Current researches cover topics from the application layer to the network layer, and the physical layer.

It has been realized that the sensor placement in the physical layer is a fundamental element of higher level performance. The placement bounds the detection performance of sensors, further limiting other aspects of the entire network. In reality, many sensor networks are randomly deployed, for example, by throwing sensors from an airplane to a tropical jungle. It is inevitable that those sensors are unevenly distributed, leaving uncovered areas (holes). These holes may degrade the detection performance of the entire sensor networks.

The Majority of previous works dealt with this defect in two ways: either tried to adapt it in higher level protocols, or required all sensors to he mobile to improve the performance. The latter approach is extremely expensive, because more battery power is required and the optimization is time consuming.

Our research takes a different approach. We believe, by correcting the lower layer, i.e., the physical layer, we will be able to optimize the network performance. Our work is one of the first works to propose the Partial Mobile Sensor Network Framework, where only a small number of mobile sensors are introduced to solve this problem. These added sensors are designed to adjust the existing sensor topology and improve network performance.

In this research, we focused on the behavior of mobile sensors. There are two major problems with mobile sensor movement: destination selection and path planning. To select the optimal destination of mobile sensors, we proposed various strategies in different scenarios. In a connectivity-aware scenario with single access point (AP), we place mobile sensors to connect isolated sensor islands to the AP. In mobile targets detecting scenarios, we select specific points which can block the paths and free walking lines of mobile targets as the destinations of mobile sensors.

For mobile sensor path planning, we introduce a terrain model into the system. Our terrain-aware strategy studies the power consumption and loss ratio of mobile sensors when they are moving on different terrains. These two metrics are utilized to select the best destinations and the best possible paths. This work is also the first attempt to address the constraints of terrain on a mobile sensor network.