| This thesis takes human-robot collaborative spatial navigation task as the background.In view of that robot has difficulty in understanding the intention of human natural language spatial navigation,a method of spatial cognition modeling was proposed to establish the model of human-like spatial cognition for robots to achieve the same cognitive structure as human.Therefore,robot can understand human natural language spatial navigation instructions better to achieve natural and efficient human robot collaborative navigation.In order to construct a robot spatial navigation cognition model that is similar to human cognition,the study was carried out to find human cognitive and expressive characteristics in specific tasks.The experimental results showed human descriptive features of landmarks in a simulated lunar environment,human preferred reference frame of describing target location on lunar surface,and natural language rules human uses for representing navigation track in lunar environment.It laid the foundation for the establishment of robot spatial cognitive model,and helped to collect corpus for the robot natural language processing module.After comparing computer modeling methods of human cognition,the ACT-R symbol modeling method was chosen.Based on the research of human spatial cognition and navigation characteristics,the cognitive interaction process of human space navigation was analyzed and mapped to different modules of ACT-R cognitive structures to build a human-like spatial cognitive model for robots.The cognitive model can realize the main function such as the information perception,reasoning,decision-making and output in spatial navigation cognition consists of a variety of module.To verify the correctness and validation of the cognitive model,the Human-Robot Collaborative Spatial Navigation Interactive System consisting of a natural language recognition module,a virtual experiment environment based on Unity3 D and a spatial cognitive model was developed.This system recognized spatial navigation command of human natural language,combined it with perceived visual information of robot for decision reasoning and output the stimulated results.Based on the system,the interactive experiment of human-robot collaborative spatial navigation was carried out.Subjects' subjective evaluation data and time performance data were obtained through the comparative experiments of natural language spatial navigation mode and command language space navigation mode,which leads to the conclusion that robot spatial recognition reasoning model can reduce the workload,increase the will of the human-robot collaboration,and achieve better time performance. |
PDF Full Text Request