Research On Interactive Navigation Technology For Service Robot Based On Topological Map

With the advent of a new wave of artificial intelligence,robotics continues to innovate and the applications are becoming more widespread.As the most closely related robot,service robot brings great convenience to people's life.The navigation technology of service robot has also changed from the traditional command mode to the interactive mode,which makes the interactive navigation more intelligent and humanized.This thesis focuses on the robotic wheelchair,which is a special service robot to assist the elderly and the disabled to realize mobility.At the same time,the human-computer interactive navigation of the robotic wheelchair is discussed.The specific technical research will be carried out from four aspects:multi-sensor sensing,topology map construction,shared control,and robot platform.Firstly,in terms of multi-sensor sensing technology,the multi-sensor sensing system framework and coordinate system is defined in this thesis,and the binocular camera model and distortion correction model for off-line calibration is introduced.In view of the traditional offline calibration can not solve the drift error and mechanical error of the on-line sensor system,an on-line iterative optimization calibration method based on hand eye calibration is proposed.This method can calibrates the position relationship between multi-sensors online,and effectively improve the credibility of environmental measurement information.Finally,a laser odometry based on the height image is proposed to solve the problem of slow matching of the laser frame in the traditional method,and the validity of the proposed method is experimentally verified.In order to solve the problem that the traditional map can not contain both occupied information and targets information,and can not be accessed at any time,a topological map construction pipeline based on the dynamic growth of the region is proposed in this thesis.It can divide and merge the topological regions by region dynamic growth algorithm and region merging algorithm based on metric map and robot trajectory,and then the topological map is processed by the metric map online.What's more,a map representation method based on the cone space model is proposed.And the topological information and occupied information are represented by parameter TM and parameter OM respectively.Based on the conic space model,these two kinds of information are converted into RGB information to save the topological map represented by color pictures.Compared with the traditional method,the building system is not limited by the environment structure,and can be adjusted automatically according to the actual environment structure.In addition,the topological map representation method simplifies the map representation structure.The experimental results of map construction show that the pipeline is flexible,computationally efficient and saves less resources,and the method of map representation is convenient and fast,which improves the efficiency of map preservation.In addition,in the field of interactive navigation,aiming at the problems of poor control effect and low humanization degree of traditional control methods,this thesis proposes a semiautonomous navigation method based on POMDP(Partially Observable Markov Decision Process,POMDP),which is used to assist robotic wheelchair to achieve interactive navigation.Based on the topological map,the shared controller will consider the user's history,posture of wheelchair,target point and schedule,and then the intention is predicted through POMDP.Finally,the fusion decision-maker will fuse the results of user selection and machine prediction,and give the most appropriate control signal to make the robotic wheelchair move to the desired position safely.Compared with the command control method,the shared controller can predict the user's intention,and take over the robotic wheelchair through the fusion decision-maker,which can alleviate the user's control fatigue.And the signal filtering of the fusion controller effectively guarantees the safety control of the robotic wheelchair.The experimental results show that the shared control system is accurate,feasible,security and comfort.In the technology of platform controlling and interactive navigation section,an robotic wheelchair and its navigation control system based on brain-computer interface and topological map are designed and implemented in this thesis.The control technology of differential platform is studied and the software architecture of human-machine interactive navigation is realized.The main input of the robotic wheelchair is the brain computer interface.The user sends control instructions through the human-computer interface to realize the human-computer interactive navigation.In the interactive navigation experiment,the prediction accuracy of the shared controller is 84.14%and the average user fatigue is 72.41%.and it's proves that all the systems and methods proposed in this thesis are feasible,safe and stable.