Research On Trajectory And Motion Planning Of Assembly Manipulation For 6R Industrial Robot

Industrial robots are widely used in manufacturing because of their strong universality,programmability,and high security.Assembly manipulation,welding,and 3D printing technologies of the industrial robot become the hotspots of the current research.Among them,assembly manipulation technology has become a key technology for high-end equipment manufacturing in fields of aerospace,automobile,etc.Point-to-point motion is the main action of the traditional assembly robot.However,with the technical advance,industrial manufacturing gradually transforms into intelligent manufacturing.Traditional industrial robot can not eliminate the vibration of the body with high-speed motion.The motion are not continuous in singular regions.The flexibility of the dual robot cooperation is poor.Perceptual aliasing of the robot is easy to happen in target recognition.The mationed limitations restrict the industrial robot to tend to intelligent assembly.Therefore,intelligent assembly has been a great challenge to the exiting trajectory and motion planning technologies,and is of great significance to carry out deep research.A great deal of valuable investigations have been done at home and abroad.However.the available trajectory and motion planning technologies for industrial robot assembly can not fundamentally reveal and solve the problem of the inaccuracy or unavailability of the trajectory optimization results caused by the singularities,and it is difficult to develop a suitable optimization algorithm.Lack of planning and control of the cooperative path,trajectory and motion of the master and the slave robot,which resulted in low efficiency and poor flexibility in the dual robot cooperation process.The influence from the environment interfere with the robot's recognition on the target,which resulted in the low success rate and poor precision of the autonomous assembly.This doctoral dissertation has studied on the trajectory planning,cooperative compliance control of the dual robot,and the robotic online motion planning for 6R industrial robot,and proposed reasonable trajectory optimization and motion planning strategies to improve the planning efficiency,cooperation flexibility and assembly success rate for the robot.The main contents are as follows:The differential transformation method was used to obtain the Jacobian matrix of SR10C robot.With the singularity conditions analysed,the partitioned inatrice which lead to singularities were saperated form the Jacobian matrix which can locate the positions of the singularities,and reduce computation complexity for finding the singularities of the joint curves.In view of the trajectory planning problem and the control characteristics of the robot's lower computer,a singularities avoidance method based on the cyclic damped reciprocals was proposed and simulated.To improve trajectory planning and optimization methods,time-optimized trajectory planning method using improved adaptive elite genetic algorithm(IA-EGA)was proposed under predefined assembly path.The singularities avoidance method was integrated into IA-EGA to avoid singularities while optimizing trajectory curves of the joints.The singularities avoidance method can improve the optimal time identification capability of IA-EGA.The quintic polynomial curve was used for the end effector motion curve model for smooth joints motion.Real time motion relationship between the Cartesian and joint space was deduced according to the kinematics.We set dynamic and static threshold for each joint and made the movement time as the objective function.The joint curves were optimized by the singularity avoidance integrated IA-EGA.The final objective function value converged.According to the optimal time,the end effector trajectory curve was planned then the motion curves of the joints were obtained.The joints curves were input into each degree of freedom of the robot to perform the motion.The proposed method improved the trajectory planning efficiency on the basis of ensuring the smoothness and continuity of the robot motion.Based on the previous research,the assembly manipulation trajectory planning for dual robot collaboration under optimal assembly path was studied.Traveling salesman problem(TSP)model of the cooperation assembly path was built based on the Markov decision process(MDP)framework.This paper proposed a shortest path planning method based on policy memoried adaptive dynamic programming(PM-ADP)algorithm.Comparing with typical path planning algorithms,the PM-ADP was more efficient and accurate for low dimensional tasks.Master-slave robots spatial relation model was built based on the basic coordinate system of the master robot.According to the assembly task features,the two kinds of cooperative motion modes were discussed and the superposition motion was selected.Under the shortest assembly path,the time-optimal trajectory of the master robot was planned,and the master robot trajectory was obtained.The master robot dominated the cooperation.Based on the spatial transformation relation between the master and slave robot,the following of the slave robot was resolved with the theoretical derivation of the dual robot.The experiment showed that the proposed method made the dual robot independent and flexible.Moreover,the smoothness and continuity of the robots motion was ensured.The planning of the path and the trajectory is offline and lower layer planning.For realizing the intelligent assembly,further research is the motion plannig of the online assembly manipulation of the robot.Aiming at intelligent assembly of industrial robots under complicated enviroment,an uncalibrated motion planning of peg-in-hole assembly method with integrated human skill and convolutional neural network(CNN)fusion driving was proposed.A LeNet-5 CNN network model was established to conduct the theoretical derivation of feedforward and backpropagation pass and programmed.The manual assembly skills were summarized,simplified and classified.CNN network was trained with human skills integrated.Simulation results verify that the LeNet-5 framework and the setting of the parameters made the CNN converge.An intelligent industrial robot platform for peg-in-hole assembly was built.Dual cameras swapping mode was used to provide visual feedback for the entire assembly motion planning process.In Eye-in-hand mode,removal of interference,workpiece recognition,centering and control of the robot were performed based on depth information sets.In Eye-to-hand mode,the real-time image was captured and processed.The trained CNN was used to obtain motion labels for robot motion planning.Comparing with the traditional mathod,the results of several assembly location experiments proved that the proposed method solved the perceptual aliasing avoiding problem effectively,had higher accuracy,success rate and efficiency of the assembly positioning.The location accuracy can satisfy the requirements of force-guidance assembly.