Study On Walking Companion Robot And Control Technologies

With the growing needs of rehabilitation assistance devices due to an aging society and for low limb dysfunction,walker robots aimed for helping the eldly and the disabled to walk has attracted many attentions from many scholars.Most of the current walker robots acquire walking intent by adopting various force sensors against movement of users’ s upper limber.Against the shortage of the available robots,most of which were too complicated to operate,a walking companier robot with a user-friendly human-robot-interface is proposed in this thesis,with corresponding motion control methods.The major research contents and results are as follows:1)Addressed the shortages of the available walker robots and according to the needs of the target group,based on user-friendliness,the construction method and technology realistic way of a walking companion robot are proposed.The robot system includes: mechanical structure,human-robot interface,control system,driving system,etc.The system has the function of walking-assist,safety protection and medical data acquisition.2)Due to the important influence of user walking intent to the robot motion control,two elemental walking states are proposed,that is,go straight and rotation.The identification method of the human-robot interface for the different walking states is established.For the elemental walking state I and II,the mathematic models between force sensor output and human-body displacement or angle are deduced.The desired displacement or angle for the robot can be calculated based on the mathematic model.The mathematic model based control method is proposed to solove the motion control of robot.Further more,in order to optimize the mathematic model,force control method is proposed by the analysis of the deviation signal between the sensor output in initial state an in real-time.In that way,the noise deviation derived from the imprecise mathematic model and parameters were resolved.The validity of two control methods is proved in simulation.3)For a combined walking state with user both made displacement and and rotation,a non-linier mathematic model is presented to solve from counter-force between human and robot to displacement and angle.The characteristics of the modelis analyzed by linearization.The dual-input and dual-output different force control systemis built.The simulink results show robot can meet the needs in displacement.With the angle information detected by angle sensor,the displacement and angle of user in real-time can be calculated based on the mathematic model.To improve the user-comfort,we concept displacement fluctuation and angle fluctuation.A smooth trajectory for robot is constructed which use multistage splicing bezier curve.4)To avoid the obstacles effectively and ensure the user-comfort,obstacle avoidance of robot is studied.A shared control method is proposed which based on reinforcement learning.An obstacle avoidance trajectory is planned by fuzzy control.The control weight of user is adjusted on-line by reinforcement learning.Finially,the motion control method is established which can avoid the obstacles effectively and ensure the user-comfort.5)The walking companion robot is experimentally studied,which include: mathematical model based control method and force control method in elemental motion state I and II,difference force control system,multistage splicing bezier curve trajectory planning,share control method based on reinforcement learning.The expermantal data are analyzed,and according to the questionnaire survey,the efficiency and rationality of the robot are verified which include robot struct,human-robot interface,servo control and safety.Finially the walking companion robot is achieved which is more safe,stable,flexible,low-cost and productization.