Wireless Sensor Networks Localization Algorithm Based On Constraint Particle Swarm Optimization

Wireless sensor network is a new platform for information access, which can realizes real-time monitoring, sense and gather information within the network, in order to achieve detection and tracking of target in the specified area. For most applications, the information collection and node locations are closely related, so node localization is the key technology of sensor networks.The common wireless sensor node localization technique contains distance-independent algorithm and distance-dependent algorithm. Through researching and analysising the existing technique, we found that the accurate node positioning generally includes two steps, localization and calculation whether it is independent with the distance, such as the DV-Hop algorithm, or dependent with the distance of the positioning algorithm. When the ranging data satisfys the conditions, we can calculate the position of the nodes. When we have measured two or more AOA measurements of anchor node, we can use triangulation method. When we have measured 3 distances of the unkown node and the anchor node, we can use trilateration method. When we have measured more than 3 distances, we can use the maximum likelihood estimation method to calculate.However. AOA ranging techniques are susceptible to external environmental and equipment. The localization error of trilateral positioning algorithm is also large. This is because while solving, the triangulation method uses only two anchor nodes, and trilateration uses only three anchor nodes. In order to improve positioning accuracy, and the full use of existing anchor node location information, you can use maximum likelihood estimation, using multiple anchor nodes to locate an unknown node. However, the solution accuracy of this approach is subject to an specified anchor node distance. If the ranging error is large, the end result in a great error.In this paper, I talk about the shortcomings of traditional positioning method, and then I propose a localization algorithm based on constrained particle swarm optimization. First, convert the node localization problem into a constrained optimization problem. Then, search space is constituted of a feasible solution and infeasible solution. Feasible solution satisfies all the constraints. Infeasible at least violates a constraint. While solving, we use restrict function to search in order to quickly find the feasible solution space, and then search for optimal solutions based on the objective function. PSO has two methods for processing these two search procedure, they are solution based on double fitness value and based on penalty function.PSO solving method based on double fitness value sets the constraint to be the first fitness function, and the objective function to be the second fitness function. Through the first fitness value, we can clear feasible region and infeasible region. While searching, the particles first converge to the feasible domain, and then compares the second fitness value. That can reduce some unnecessary computation. PSO solving method based on penalty function compares all possible solutions in the two-dimensional space. Through the penalty term, it can control the particle away from the infeasible region, and closer to the feasible region.This search method can reduce the possibility of local optimum. After determining the solution method, we can search using particle swarm optimization. Current research on the particle swarm algorithm has been very much, such as particle swarm with weights, particle swarm with the shrinkage factor, dissipative particle swarm and discrete particle swarm, etc.. Dissipative particle swarm algorithm which principle is simple, and is easy to implement, has a wide range of applications. In this paper, we made improvements based on original dissipative particle swarm, and proposed an improved dissipative particle swarm optimization (IDPSO), which finally applied the solution for node localization problems.Finally, simulation experiment shows that compared with maximum likelihood estimation method the two CPSO-based localization, can get more precise solution in different ranging error, different distance radius, different number of anchors and different number of nodes. This shows that CPSO-based localization algorithm has more robust against errors, better convergence and less hardware investment, etc; In addition, effects in the sparse network node localization are superior.