Prediction-Based Particle Swarm Optimization Algorithm for Solving the Inverse Kinematics of Spatially Dexterous Finger

Abstract

Aiming at the problem of complex and computationally intensive inverse kinematics solution of the self-developed spatial humanoid dexterous hand, this paper proposes a prediction-based empirical shared Q-learning hybrid particle swarm algorithm (LQHPSO). Firstly, the fireworks algorithm is introduced into the particle swarm optimization algorithm (PSO) to update the individual particle extrema by using the fireworks explosion characteristic to improve the global search capability of the algorithm; secondly, a perturbation factor is introduced to the individual extrema after the explosion to improve the convergence speed of the algorithm; then a Q table is added to each particle and an experience sharing mechanism is designed to enhance the learning capability of the particles by sharing the behavioral experience of the optimal particles balance the global and local search ability of the algorithm; in addition, the LQHPSO algorithm combines the characteristics of spatial humanoid dexterity hand to predict the initial values of iterations, thus substantially improving the computational efficiency. The experimental results show that the LQHPSO algorithm has better real-time performance compared with other intelligent algorithms with the same convergence accuracy.