| With the improvement of people’s living standards,more and more people are used to relying on the convenience brought by high-tech products to family life.Therefore,family service robots have always attracted great social attention.Family service robot is a kind of special robot serving for human beings,which can help human to complete a variety of tasks such as household cleaning,childcare and elderly assistance.In the long run,the research on home service robots is not only a great opportunity for future social development,but also an important channel to improve human living standards.As a key mechanism for work,it is very important for manipulator to have flexible ability of imitation learning and motion planning.Based on this point,this paper proposes a trajectory planning method of manipulator based on human body teaching and visual positioning.In order to further expand the working space of the robot,an under-actuated omnidirectional mobile mechanism is improved in this paper,and the motion control of the mechanism is completed.Finally,the robot arm learning and chassis control system are fused to build the overall learning framework of the service robot.This subject takes the humanoid home service robot independently designed by the laboratory as the research object,and carries out in-depth research from the above two aspects,expecting to make contributions to improving the autonomous planning ability and skill expansion ability of home service robot.The main research contents and results are as follows:(1)The hardware system platform of the home service robot was improved.Based on the D-H parameters of the single manipulator,the kinematics model was established and the forward and inverse kinematics equations were solved.In addition,the composition of the robot underactuated mobile chassis based on the special coupling system is emphatically introduced.Through the dynamic analysis of the underactuated mobile chassis under different motion modes,the innovation and feasibility of the design are strongly explained.In order to establish the correlation between the manipulator system and the visual recognition system,the hand-eye calibration was completed.Finally,from the software level,the control and communication methods of the robot learning system in this paper are introduced,and the idea and process of function realization in this paper are clarified.(2)A trajectory planning method for manipulator is proposed,which combines Gaussian Mixture Model(GMM)with Dynamic Movement Primitives(DMPS).Firstly,WSSS(Wireless Standalone Sensing System)motion capture device is used to replace the traditional teaching method to complete the collection of motion information of multiple human arms.Then the GMM model was applied to DMPS system forcing term learning,and the correlation model between the forcing term f and phase variable s of the teaching trajectory was established to extract the motion features.The target location information obtained by the camera recognition system is used as the input position parameters of the trajectory planning system so as to obtain a new trajectory which can converge to the target.(3)In Matlab,one-dimensional trajectory,plane trajectory and three-dimensional space trajectory are taken as the research objects.The simulation verification of the application of GMM and Locally Weighted Regression(LWR)method in DMPs is carried out,and the simulation results of the two methods are compared and analyzed.It not only verifies the ability of DMPs to reproduce and generalize the motion trajectory,but also highlights the superiority of GMM modeling method to learn the forcing term.(4)Setting up robot verification experiments.The trajectory planning experiment of the robot arm approaching to the target is designed in the accessible motion space of the manipulator arm.The planning results obtained by the upper computer are transmitted to the lower computer and converted into the robot arm motion instructions to drive it to complete the corresponding actions.At the same time,a motion compensation method is proposed to improve the DMPs by rotating the forced items,and the effectiveness of the forced items compensation is verified by the real experiment.(5)Finally,the rationality and feasibility of the design of the mobile mechanism are verified through the planning experiment of the robot arm combined with the autonomous movement of the chassis.At the same time,the DMPs trajectory planning method based on visual recognition is applied to the grasping task of the robot,which proves that the learning framework proposed in this paper is reliable in the task execution of the home service robot. |
