Abstract: In order to improve the stability and yield of apple picking robot， the method of inverse kinematics of picking manipulator was proposed. First， the DH model was used to establish the model of the apple picking robot. Kinematics problem of apple picking robot was thus translated into programming problem. The objective function was the Euclidean distance between the apple picking robot coordinate and unknown coordinate. In order to obtain the inverse kinematic parameters， the genetic algorithm was then used for global search， which included selection， crossover and mutation operators， and nonlinear programming was used for local search based on the result of genetic algorithm. Finally， the kinematic parameters were divided into 3 grades by random forest algorithm. Experiment showed that compared with the genetic algorithm， the algorithm proposed in the present study improved the accuracy by 8-25 mm in solution of inverse kinematics. In conclusion， the proposed method was valid for kinematics problem， and thus would help improve the yield of apple picking robot.

Key words: picking manipulator, genetic algorithm, inverse kinematics, DH model, random forest