Modeling, Planning And Controlling Of Active Sensor Networks Based On Multi-robots

With anticipated capabilities of self-organization, target tracking, less sensors and adaptation to changing environments, active sensor networks have many potential applications in both military and civil areas.The dissertation studies the problems of modeling, planning and controlling of the active sensor networks based on a group of mobile robots. It is focused on the development of a model that can simultaneously represent robot motion, communication constraints and dynamic interactions among robots so that new algorithms of motion planning and control can be developed on a unified basis for achieving the self-organization and redeployment of the active sensor network.The major research work and contributions of this dissertation can be summarized in the following five aspects:Firstly, aimed at the complicated environment application, a model has been developed for modeling active sensor networks based on multi-robots to uniformly represent communication topology and dynamic interactions effectively.Secondly, to minimize the energy consumption and the change of topology, we studied the problems of the coverage and the connectivity of the active sensor network with initial randomized uniform distribution. This research provides a theoretic basis for design of the sensor node and the selection of network scale. Firstly the probability distribution of the connectivity degree of the active senor network is analyzed based on coverage processes theory. Accordingly the relational expression of the communication radii and the quantity of sensor node is deduced on condition that the probability of desired connectivity degree is maximal. With the necessary condition of the full region coverage, the relational expression of communication radii and the sensing radii of sensor node is inferred. Lastly the influence on the active sensor network with initial randomized uniform distribution caused by boundary effect is analyzed by simulation.Thirdly, in order to solve the network fragmentation in sparse networks, a new algorithm is presented to maintain the connectivity of the sparse network. In the algorithm a virtual node is inducted as the collective object, which can make all the subnets connected at first. Then the GGA (Greedy geographical algorithm) is used to find the nearest node to the collective object in the subnet, and to judge whether the condition of full connection of the network is realized using the distance among the nodes. When a subnet needs move, the node nearest to the collective object is regarded as leader of this subnet. By using the DLF (Dispersed leader-follower) method that enforces the nodes in the subnet to form a tracking chain, it is possible to maintain the connectivity of the network during the motion.Fourthly, an algorithm based on Triangle Grid Coverage (TGC) is developed to realize the full coverage of an area by a dense network. The algorithm first calculates the triangle grids with the minimum number of triangles based on the communication topology of the network. By analyzing the overlapping among the sensing coverage of the three nodes located at vertices of a triangle grid, the sensing coverage ratio of the triangle grids can be calculated. Therefore, it is possible to judge whether there are any holes in the triangle grids and if there are, the holes can be filled to achieve a full coverage of the area.Fifthly, a distributed self-organization motion planning method is presented. Self-organization of an active sensor network always requires a group of mobile robots to move from an area to a desired area in the environment with obstacles to reconfigure the network topology according to the scheduled layout. During the self-organization, it needs to be ensured that each mobile node maintains the wireless link to the network. By optimizing the preserved connectivity of the mobile robots, a distributed motion planning algorithm based on a single-step location prediction and collective potential field is presented to deploy and reconfigure the active sensor network, which reduces the complexity of implementation. The stability of this method and the preserved connectivity of the network are analyzed. Simulations are conducted for a group of more than 40 robots with and without obstacles in the environment to validate the proposed algorithm. The results show that the proposed algorithm is effective for reconfiguration of a large scale active sensor network, as well as the networks with different size.