The Research On Robotic Multi-Object Grasp Planning In Cluttered Environment

With the transformation and upgrading of the traditional manufacturing industry,robots have been widely used in industrial production.In order to improve the flexibility of robot in the actual grasping task,based on grasp planning as the research target,a universal grasp framework has been constructed to generate high-quality,collision-free and reachable ranked list of grasping set.This framework makes the robot achieve high efficiency,stable grasping operation in cluttered environment.The main research work of this thesis is as follows:(1)A grasp planning algorithm based on the combination of particle swarm optimization and simulated annealing is proposed.This algorithm selects force-closure as grasp strategy and use eigengrasp to reduce the dimension of grasp search space.This thesis proposes a combining simulated annealing with particle swarm optimization algorithm for global optimization of pregrasp sets,then use task-oriented ellipsoid quality measure algorithm to evaluate grasp stability,build error model to evaluate grasp robustness.Finally,a ranked list of high-quality grasp sets is obtained to meet the requirements of grasp stability,robustness and task compatibility.(2)A grasp set filtering algorithm in cluttered environment is proposed.In this algorithm,the grasp priority decision is made for the multi-objective grasping task.The reachability and collision probability of grasp set are analyzed by using the reachability space and collision-free probability model.A grasp comprehensive evaluation formulation is proposed for evaluating optimal grasp in complex environment.Finally,the inverse kinematics and collision detection are used for verification.The algorithm can be applied to a variety of grasping tasks in cluttered environment and a ranked list of comprehensive grasp sets satisfying quality indicators,robot and environmental constraints can be obtained.(3)In order to evaluate the universality,feasibility and effectiveness of the automatic grasping framework,a simulation platform and an experimental platform for disordered grasp are built respectively.The platform consists of a Franka 7-dof robot,YCB library containing a large number of objects,and three robot hands.Among them,the optimal grasp planning algorithm is evaluated by generating a large number of grasp sets,then analyzing their efficiency,stability and robustness.The grasp set filtering algorithm is evaluated by randomly generating complex grasp scenes,and analyzing efficiency and success rate of the algorithm.Simulation and experimental results show that the proposed framework has good performance.By pre-computing the object grasp set offline and effectively calculating a grasp score function online to comprehensively sort the grasps,we can quickly find highquality,collision-free and robot-reachable grasp in cluttered environment.The algorithm has been applied to the simulation and experimental platform of disordered grasping,which can satisfy the robot to quickly perform complex grasping tasks in the dynamic environment.