Research On Socialized Navigation And Interaction Of Mobile Robot For Human-Robot Integrated Environment

The rapid breakthrough of artificial intelligence and computer software and hardware technology has brought great impetus to the research of service robots.With the improvement of human social needs,the service content of public intelligent service robots is constantly expanding.Compared with industrial robots,service-oriented mobile robots often work in a human-robot integrated environment.In this scenario,there are many unstructured objects and strong time-varying.In addition to satisfying basic navigation indicators,robots also requires anthropomorphic and natural behavior to interact with objects.At present,the mainstream mobile robot obstacle avoidance technology in the human-robot integration environment is the local path planning algorithm.This kind of method only regards pedestrians as dynamic obstacles,and does not distinguish them from ordinary obstacles,which greatly affects the interactive experience of people,which cannot meet the physical and psychological safety needs of pedestrians at the same time.Aiming at the navigation problem of the human-robot integrated environment,this paper studies the navigation method with social awareness.The specific research contents are as follows:(1)In order to give mobile robots the ability to distinguish pedestrians from ordinary obstacles,this paper proposes a real-time perception method of pedestrian multi-attribute information suitable for robot navigation process based on target detection algorithm.Based on the real-time performance and detection effect evaluation of the existing target detection algorithm on the real mobile robot platform,the depth camera is used to detect the motion information and attribute information of the interactive objects in the navigation environment,and the pedestrian detection algorithm is packaged under the robot operating system.(2)Aiming at the lack of social norm awareness in current obstacle avoidance algorithms,this paper proposes a social adaptive obstacle avoidance learning algorithm based on Conditional Variational Autoencoder(CVAE).Through the imitation learning of human obstacle avoidance strategies,the robot can autonomously select temporary obstacle avoidance points according to factors such as the movement speed,movement direction,number of pedestrians and obstacle positions of surrounding people.By moving to the temporary target point,the robot can follow the pedestrian motion specification and avoid motion conflicts with pedestrians.Design scenes with different crowd density and environmental complexity and manually mark temporary target points to form a data set,introduce a deep learning model CVAE to learn temporary target points,and design a navigation framework in combination with path planning.(3)Aiming at the problem of a single interaction mode of service robots,based on the research results of pedestrian psychology,a motion interaction method with pedestrian attribute awareness is proposed.The attributes(gender,age,body shape,emotion,etc.)that affect the size of pedestrians’ private space are encoded into the CVAE training set,and the quantitative indicators for evaluating the effect of temporary point generation are set to optimize the CVAE network.The model can adaptively adjust the navigation behavior according to the pedestrian attributes,and has stronger motion interaction ability.(4)In order to verify the social norms and pedestrian attribute awareness of the navigation algorithm proposed in this paper,we build a software and hardware platform for a social adaptive navigation system,design a human-robot integrated navigation scenario,and integrate pedestrian perception and social adaptive navigation based on Turtlebot2 and related sensors,and finally complete the navigation experiment.To sum up,this paper proposes a social adaptive navigation method from three levels:interactive object detection,social norm awareness learning,and path planning.Experiments show that the navigation method proposed in this paper has good performance.It is of great significance to the wider application of service robots.