Study On Cooperative Localization For Hy-Brid System Of Mobile Robot And Wireless Sensor Networks

In recent years, mobile robot and Wireless Sensor Networks (WSN) have became a focus in research, for their prominent applications in domains. Localization, as the most fundamental requirement both for mobile robot and WSN, is the precondition for robot autonomy and for WSN's monitoring and target tracking. For the inherent limits and disadvantages of mobile robot and WSN, there still exist bottlenecks to resolve localiza-tion problem in these fields. The proper combination of mobile robot and WSN brings an opportunity to overcome the disadvantage and decrease the difficulty in both fields when resolving the localization in cooperative way. Thus, more accurate localization will be achieved.On the basic of the hybrid mobile robot and WSN, and focus on three aspects: the formulation of localization problem, localization algorithm and localization error be-havior. The paper try to answer those questions: What is the localization of hybrid sys-tem? How to resolve the localization in a cooperative way? And why the hybrid mode was used to resolve localization for mobile robot and WSN?The following researches are performed and finished in the thesis to deal with the first two aspects mentioned above:(1) The localization problem of hybrid system was described as the states estima-tion in probabilistic framework, and its mathematical model was deduced from Bayes rules. By using this model, the localization differences among hybrid system, traditional robots and WSN nodes were showed clearly on the viewpoint of utilizing of reference localization information. The study proposes a novel hybrid system localization algo-rithm based on Extend Information Filtering. It uses sparse graph to show the constraint and prior information, and is easy to confuse all the localizing information in canonical form to compute all robot poses and all nodes location estimate. Moreover, the algo-rithm is processed offline and robust for avoiding communication failures and delays. In order to construct the state estimate in canonical form at low price, several methods were used to optimize the computation in the process of computing state estimate.(2) The dissertation proposes a technique for estimating the position of the nodes in WSN aided by a mobile robot by using range-only measurements, and also presents its mathematical model. Additionally it deduces the measurement probabilistic model with location uncertainty of neighboring node. A novel cooperative localization algorithm is properly designed according to the character of robot and WSN. It takes advantage of the good computing capability and mobility of the robot, as well as distributed comput-ing and communication capabilities of the sensor nodes. By combining Gaussian Mix-ture Model Filtering with Information Filtering, the algorithm deals with the reference localizing information in a central-distributed way, and assigns the computing task to robot and nodes according to localizing phases. The cooperative way makes the algo-rithm run efficiently. Accounting to the cross-correlation between reference localizing information, the study designs a collaborative mode for the system, and presents a tech-nique removing the cross correlation between conference localizing information, to lead conservative estimates by using multiple measurements,and prevent the result to be overconfidence.Then the study focuses on the error analysis:(3) Classical CRLB (Cramé-Rao Lower Bound) model used in accurate beacon en-vironment was expanded, and a more general CRLB model was deduced in case the beacon nodes location is uncertain. The localization error behavior with respect to the network setup and measurement noise was explored in the simulating experiments by using range measurements, and the availability of CRLB model expanded was proved. Then the study makes a theoretic analysis of convergent character of node localization estimation in WSN based on expanded CRLB model, and proves several theorems then deduced that: Comparing the accuracy between node location accuracy in classical node-beacon localization mode and collaborative localization mode, the collaborative way tends to lead more accuracy. We proved the conclusion above, and the influence respect to measurement types, number of nodes added was explored in the simulating experiments further.(4) On the basis of CRLB model in robot localization, we expand the classical CRLB model to the formulation which can compute the uncertainty of full state space and real-time state estimate in hybrid system. According to Gaussian noise assumption, the node location accuracy between classical robot localizing mode and cooperative lo-calizing mode was compared in hybrid system, and analyzes how system noise and ro-bot movement affect the error behavior of the system element estimates. Then the study develops a fundamental understanding of the error behavior in the system through the-ory and experiments, and indicates that in a hybrid system, sharing and using available information between all the elements can improve the accuracy of localization in both WSN and robotics. So it is replied that why the combination mode was used to deal with localization for mobile robot and WSN. At last the localization error behaviors with respect to the system noise, robot motion, etc. were explored in the simulating ex-periments, and the study also discussed how to make a balance between localization accuracy and computation cost.