An Adaptive Particle Swarm Optimization Method for Solving the Grasp Planning Problem

This paper proposes an adaptive particle swarm optimization (APSO) approach to solve the grasp planning problem. Each particle represents a configuration set describing the posture of the robotic hand. The aim of this algorithm is to search for the optimum configuration that satisfies a good stability. The approach uses a Guided Random Generation (GRG) to guide the particles in the generating process. A shape-based object “parameter factor” is generated from the GRG process so that, it can be considered in the fitness function. According to the number of contacts between the fingertips and the object, the algorithm can take off the inactive particles. The kinematic of the modeled hand is described and incorporated in the fitness function in order to compute the contact positions. The APSO is tested in the HandGrasp simulator with four different objects and the experimental results demonstrate that this approach outperforms the compared simple PSO in terms of solution accuracy, convergence speed and algorithm reliability.