Evaluation of genetic algorithm on grasp planning optimization for 3D object: A comparison with simulated annealing algorithm

Grasp planning based on geometrical information of objects can be approached as an optimization problem where a hand configuration that indicates a stable grasp needs to be located in a large search space. In this paper, we study the applicability of genetic algorithm (GA) on grasp planning optimization of 3D objects. The details are given on the selection of operators and parameters. Different sampling methods in the implementation of crossover and mutation operators are tested. A quantitative analysis including the comparison with random planner and simulated annealing (SA) method is performed to evaluate the performance of the GA based planner. GraspIt! simulator [1] is used for implementing the proposed algorithm and as the test environment. Two different quality metrics are considered. The result shows that GA is a robust method in the field of grasp planning. And the GA planner outperforms the SA planner in both pre-grasp quality and stability of the final grasp.