Study Of Clustering-based Multi-objective Optimization Algorithms And Their Application To Manipulators

Redundant manipulator path planning is a typical multi-objective and multi-constraint optimization problem.Trajectory tracking control of manipulator is also an important field of robust control research for nonlinear systems in recent years.Through the in-depth study of the problems existing in these research fields,some difficult problems are solved.These sutdies have an important theoretical significance and application value to promote the progress of intelligent optimization algorithm and control technology of manipulator system.The process of solving multi-objective optimization problem is to find Pareto solution set,so that the corresponding target vector can approach and uniformly cover the whole Pareto frontier.If we can mine the distribution structure of the solution in the population and control the parental source with certain mating restriction probability,we should improve the quality of the progeny solution and the search efficiency of the algorithm.In addition,the trajectory tracking control problem of manipulator system with unknown external disturbance is studied in detail.The main research contents are summarized as follows:Due to the improved composite multi-objective particle swarm optimization algorithm still has the disadvantages of fast convergence and easy loss of population diversity when solving specific optimization problems,from the perspective of adaptive regulation of population diversity and convergence,a multi-objective evolutionary algorithm(Co DEMOPSO)combining compound difference algorithm and multi-objective particle swarm optimization algorithm is proposed.The algorithm uses adaptive mating restriction strategy to control the parent generation from particle swarm or differential evolution operator,and adaptively updates the mating restriction probability of the population in each generation according to the utility of different reproductive mechanisms in a certain algebra in the past,which makes the generation mechanism of new solutions more reasonable.The proposed algorithm is compared with the representative multi-objective evolutionary algorithm on the test set with complex PF and PS geometries.The experimental results show that the proposed algorithm has the advantages of fast search efficiency and high search quality.In order to effectively balance the exploration and production capabilities of the multiobjective evolutionary algorithm solving different optimization problems and solving the same optimization problem in different search stages,a self-organizing multi-objective evolutionary algorithm(ASMEA)based on adaptive mating restriction probability is proposed to minimize the computational overhead generated during the assisted computation of clustering algorithm.Compared with traditional algorithms,this algorithm has the characteristics of clustering and evolutionary fusion recombination operators.Compared with traditional algorithms,this algorithm has the characteristics of clustering and evolutionary fusion recombination operators.It is easy to produce high-quality solutions in the late stage of population evolution,and can be alternately trained and evolved(SOM).Mating restriction probabilities is used to control mating fathers from the SOM’s discovered neighbor population or the entire population,enhancing extraction and exploration capabilities.In addition,the mating restriction probability of the algorithm is adjusted adaptively according to the recombination utility of different parent sources in a certain algebra in the past.For the test questions with complex PF and PS geometries,ASMEA and representative multi-objective evolutionary algorithm will be used to carry out comparative experiments.The experimental results show that ASMEA is superior to the representative algorithm in search quality,search efficiency and visualization,which verifies that ASMEA algorithm has good performance in solving multi-objective optimization problems.In view of the path planning problems,Co DE-MOPSO and ASMEA multi-objective evolutionary algorithms are used to solve the path planning of the actual manipulator in two experimental scenarios with obstacles.The results show that the proposed two algorithms can effectively solve the path planning problem of the manipulator in scenario 1,but in scenario 2,both algorithms fail to solve the problem due to the narrowing of the feasible region.Therefore,considering that the algorithm uses the same mating restriction criteria for all waypoints and the physical significance of each waypoint,a more advanced multi-objective evolutionary algorithm(MRPE)based on evaluation function and clustering mating restriction strategy is proposed.The characteristics of MRPE are to establish a neighborhood relationship for each waypoint with the same serial number and set a separate probability of mating restriction for each waypoint,thus controlling the source of mating fathers to balance mining and exploration.In addition,the mating limitation probability of MRPE is updated adaptively by the waypoint quality measured by the evaluation function.Compared with the representative MOEA,the experimental results show that MRPE can solve the path planning problem of manipulator more effectively.On the basis of the above path planning,the trajectory tracking control problem of manipulator system with unknown external disturbance is deeply studied.Firstly,the barrierfree path sequence from the beginning point to the end point is obtained by using the path planning method in Chapter 4.Since the obtained path is a discrete path point,which does not contain motion velocity,acceleration and other information,the B-spline function is used to interpolate the discrete path in joint space to obtain a reference trajectory conforming to kinematic and dynamic constraints,and the reference trajectory is used as the reference input signal of the motion control system.Secondly,aiming at the trajectory tracking control problem of the manipulator system with unknown external disturbance,a sliding mode controller(control algorithm)with disturbance compensation function is designed by combining the nonlinear system sliding mode control method and nonlinear disturbance observer technology to realize the asymptotic tracking of the desired joint motion trajectory and give strict stability proof.In the design of gain matrix of nonlinear interference observer,a new design method based on adaptive technology is adopted,which solves a difficult problem in the design of traditional nonlinear observer.