Behavioral Decision And Methods Research For Mobile Robot Using Data Fusion

For the research of mobile robot,the previous researches are mostly based on the ability to establish the precise kinematic model or the known,definite and structured working environment.Therefore,in the complex,uncertain and unstructured environment,the ability of reasoning and decision making has become one of the hot and difficult problems in the field of robotics.It is of great scientific significance and application value for the mobile robot behavior decision making to ensure the reliable operation and operation ability of the robot.The existing data fusion algorithms mainly consider the fusion of the same kind of sensor information,and the fusion process is complicated.The fusion model is mostly functional / structural framework model,and the data expression ability is weak.To establish an efficient and practical fusion model is difficult,resulting in the low sensitivity of the mobile robot autonomous behavior,slow response speed etc.Aiming at the above problems,this paper makes a thorough research on the difficulties and key problems of multi-source data fusion,the main research work is as follows:1)According to the fusion effect is not ideal,low reliability,data expression ability,complex mathematical calculation in the presence of data fusion,this paper introduces knowledge and domain expert,describe the information dynamic mechanism and internal interaction logic,processing of data fusion in knowledge driven.Due to the characteristics of production rules,such as naturalness,modularity and uniformity,the knowledge is represented by production rules,and the frame model of multi-source data fusion is established.2)In the establishment of data fusion model,based on the existing coupling,asynchronous,concurrent control,the characteristics of mobile robot behavior,information circulation way and fusion level and fully considered,is proposed by using the Petri net method,the data fusion of Petri network layer model to facilitate the analysis of interaction between logical behavior and independent decision-making behavior.When the data fusion model is established,there are some characteristics such as coupling,asynchronism,restriction,concurrency and so on,and the way of information circulation and the level of integration are fully considered.The method of Petri net is presented.The establishment of the data fusion of Petri network layer model to facilitate the analysis of interaction between logical behavior and independent decision-making behavior.The data acquisition and processing layer of this model by introducing the forbidden arc solves the problem of multi sensor information transmission and dynamic scheduling,and improves the ability of information processing and fusion.3)In order to improve the efficiency of modeling and the accuracy of the model,a mapping model and a mapping algorithm between production rules and Petri nets are established.The structure and parameterization of the mapping model enhances the reusability and sharing of the Petri net model.4)Considering the real-time characteristics of the sensor data acquisition and the stability of mobile robot motion,will affect the precision and reliable implementation of its own behavior.Three kinds of sensors are studied,which are ultrasonic sensor,kinect sensor and microphone array sensor,respectively for the data acquisition and performance testing.5)In order to robot can efficiently and reliably carry out independent decision-making,based on the above three types of sensor and data fusion Petri network structure model,established data fusion for mobile robot behavior decision Petri network main network and three subnet model based on the fusion process can be refined.The behavior decision of mobile robot is described in detail,and the information flow of the fusion system is analyzed dynamically,to ensure the normal work of the sensor module.6)To build an experimental platform for autonomous mobile robot,the autonomous tracking process,verify the effectiveness and feasibility of the above fusion algorithm and Petri net model for mobile robot behavior decision.It can provide methods and model support for multi-source information fusion.