Research On Vision Based Robot Manipulation Technology

Vision is an important means for robot to perceive the environment,and it is the main way to improve the robot’s working ability.Robot operation based on visual guidance has been widely used in object sorting,handling,stacking and other tasks.However,in the parts assembly scene,there are some difficulties in the vision-based robot assembly technology due to the variety of parts,irregular shape,uncertain attitude,and high requirements on positioning accuracy and alignment accuracy.In addition,considering the high cost,large size and complex hand-eye calibration of 3D vision,this paper takes monocular vision guided robot operation technology as the research object to carry out research work.The main research content of this paper includes the following parts:Aiming at the assembly requirements of parts with irregular shape and random position,this paper proposes a robot assembly system structure based on initial positioning of monocular vision,precise positioning of mechanical fixture and precise alignment of force compliance control.Firstly,a hierarchical pyramid search and matching method based on monocular vision was used to realize the rough positioning of three-dimensional attitude of irregular parts based on monocular vision.Then the auxiliary positioning fixture is designed based on the potential energy criterion to realize the accurate positioning of irregular parts such as eccentric shaft motor and L-shaped parts.Finally,the precision assembly of shaft-hole alignment in the above parts is completed by force compliance control technology.The work results of this paper were partially applied to the industrial robot assembly competition of the World Robot Summit.The robot automatic assembly system built based on vision and force perception successfully realized the automatic assembly of the transmission unit containing a variety of parts.Further,in view of the robot’s precise operation requirements based on monocular vision,this paper studies the robot control method based on visual servo,proposes the robot’s precise visual servo operation method based on model-free adaptive control,and solves the problem of robot operation error caused by the uncertainty of robot hand-eye calibration error,robot model error and so on.The proposed method firstly establishes the pseudo-inverse Jacobian matrix model of the visual servo system,and then uses the historical data of robot input and output to adjust the pseudo-inverse Jacobian matrix parameters according to the optimal tracking error criterion,so as to improve the ability of the robot control system to deal with the parameter disturbance problem and make the system converge more quickly and stably.And the positioning accuracy has also been improved.Finally,in order to verify the effectiveness of the visual servo control algorithm based on model-free adaptive control,the simulation environment and the physical environment of the experiment are built respectively,and the hardware and software of the two environments are introduced respectively.The comparison between the model-free adaptive control method and the model-free adaptive control method is done in the experiment.The effectiveness of the method can be proved by analyzing the experimental data.