| Autonomous localization of mobile robot in unknown environment is an important topic of robotics research.Researchers had made many theory and application achievements about this topic in recent years.However,there are still many challenging research works need to be done for realizing truly autonomous localization.With an unknown environment,there are dynamic objects,environmental disturbances and other uncertain factors,etc.These factors have brought many difficulties to accurate and robust autonomous localization of mobile robot.In this paper,some issues of this problem are studied in-depth,and the corresponding innovative solutions have been proposed.The other major theme of this paper is multi-robot cooperative localization.This theme is an extension of the previous subject.The multi-robot system has more observation information than the single one.How to integrate and coordinate those observation information while avoiding the potential information conflicts,is a challenging problem in multi-robot cooperative localization research community.In this regard,complete information static game model and maximum Information entropy game model for multi-robot cooperative localization have been proposed in this paper.Specific research works finished in this paper are as follows.Firstly,perception of structured environment has been studied.An incremental clustering algorithm for feature extraction based on laser radar data has been proposed.Then,on the basis of this algorithm,a robot localization experiment with constructed feature map is presented.Secondly,for data association in the environment with many uncertainty and ambiguity factors,such as large noise,dynamic occlusions,and so on.A novel data association algorithm subtle conjoins Reweighted Random Walks and Shape Contexts has been proposed.The algorithm can utilize the geometric topology information contained in the observation data of the mobile robot more effectively,and obtains accurate and robust data association results.It can effectively guarantee the accuracy and consistency of the autonomous localization of mobile robot in an unknown environment.Thirdly,the problem of simultaneous localization and mapping in unknown dynamic environment is studied.A novel algorithm for SLAM in dynamic environments using landscape theory of aggregation has been proposed.The algorithm innovatively uses landscape theory of aggregation into dynamic SLAM model constructed by probabilistic Robotics.The dynamic and static objects in the observation data can be clearly separated without a prior dynamic object model by this algorithm,then creates a dynamic environment map with high consistency.Fourthly,this paper study on how to integrate and coordinate relative observation information of multi-robot,in order to realize the more accurate and robust cooperative localization.For recognizing and eliminating the relative observations conflict between two meeting robots,a novel cooperative localization algorithm which merges complete information static game model into EKF(extended Kalman filter)has been proposed.Since relative observations can be classified into two types,i.e unilateral relative observations and bidirectional relative observations,two sets of EKF cooperative localization formulations have been deduced respectively.The proposed algorithm makes sure the two meeting robots only share coherent relative observations between them,and improves the cooperative localization accuracy as well.Finally,this paper focuses on the problem of the potential observations information conflicts when an object is detected by many robots simultaneously.A novel cooperative localization algorithm based on maximum entropy gaming and EKF is presented.With the basic principle of the information theory and the game theory,the proposed algorithm optimized of the co-observations between the multi-robot system,and improves cooperative localization performance while avoids the potential information conflict problem. |
PDF Full Text Request