Research On Human-Robot Interaction Technology Based On Humanoid Intelligence

Human-robot interaction(HRI)problem is to understand and shape the interactions between humans and robots with special sensors and interfaces under the supporting of control technology and interaction technology.The objective of HRI is making robots be harmonious with humans.Then,the robots can accomplish interactive tasks naturally and provide operators with effective feedback.The development of HRI will raise human's working efficiency and satisfy human's demands of a life.Currently,HRI has become a hot research topic in robotics applications.With the continuous expansion of the application fields of HRI,the environment of HRI is becoming more and more complex,and the user's demand for HRI is more natural and intuitive.Due to the different and complex application environments,it is hard to describe in precise mathematical model.The traditional human-robot interactive control based on unified models is difficult to take into account the performance of human robot interaction systems and the needs of users.Therefore,the key to overcome the bottleneck of HRI technology is to integrate human-simulated intelligence into the main technologies of HRI.This paper mainly studies human-robot interaction technology based on human-simulated intelligence.The main research work and innovations of this dissertation are summarized as follows:1)A humanoid behavior model of human-simulated intelligent human robot interaction system is established.The muscle training behavior model(MTBM)is applied to optimize the user's gestures in the human robot interactive system.The pre-touch Surrounding Detecting Behavior Model(PTSDBM)is applied to the real-time detection of the robot's pre-touch information in human robot interactive system.The Dual Hands Cooperation Behavior Model(DHCBM)is applied to multi-robot cooperation in human robot interactive system.2)Propose an optimization method for the user's gesture instructions based on MTBM.On one hand,a double directed clustering algorithm(Double Directed Clustering Algorithm,DDCA)is proposed to train the user's gesture data,which is in order to generate the core intention of the user's gestures.The method aims at the operator's invalid gestures derived from fatigued or unfamiliar operations.The DDCA can optimize and filter the operator's original gestures,then the redundant movements of the robot are reduced.On the other hand,a memory responding algorithm(Memory-Responding,Algorithm,MRA)is proposed to construct the mapping of the desired motions of the robot and task requirements,and to realize the reusability of human-computer interaction experience.3)Propose a human robot interaction pre-touch information detection algorithm based on PTSDBM.In view of the multi-robot human robot interaction system,the perception of the blocking information is studied.Proposed a virtual frame construction method for the robot arm end-effectors to avoid the robot arm end-effectors from blocking each other.The method can predicts pre-touching positions and relative distance between the two robot arm end-effectors.On the one hand,it is helpful for the real time feedback of the motion states of the robot arm end-effectors.Construct the mapping matrix between the robot's coordinate and the coordinate of the depth information sensor based on ICP algorithm when one of the robot arm end-effectors is blocked by the other.The algorithm can predict the positions of the robot arm end-effector.4)A human robot interaction method based on DHCBM for multi-robot is proposed.The method consists of centralized leader-follower control and closed-loop constraint control.The centralized leader-follower intention control according to the user's gesture manipulation to control the expected motions of the common object,and gives the virtual driving capability of the common object.Multi-robots closed-loop constraint control is the control instructions of the robot that are according to the desired motions and the movement feedback of the common object.The result of the method is that the movement of robots changes the motion of the common object under the closed-loop constraint chain composed of the robot and the operating object.The proposed method is helpful to solve the problem of kinematics inconsistency in multi-robot cooperative system,and to achieve the purpose of reducing the accumulation of motion error in human-robot interaction system.5)The prototype experiment of human-simulated intelligent human robot interaction is tested and verified.On the basis of the above theoretical study,the overall architecture of this system is created.On the interactive experimental environment with multi-degree of freedom robot arms,the optimization algorithm of the user gestures instructions based on the MTBM,the detection algorithm of the robot's pre-touch information based on PTSDBM and the human robot interaction multi-robot cooperation algorithm based on DHCBM are experimentally tested and analyzed.The experiments show that: the proposed human-simulated intelligent human robot interaction methods reduce the influences of the robot's redundant motions derived from the user's non intentional instructions.The methods reduce the error accumulation of the robot,improve the performance of human-robot interaction system,and enhance the user's natural and intuitive interactive experience.It lays a solid foundation for the further research of the human-simulated intelligent human robot interaction.