Optimising Robot Manipulator Inverse Kinematics with Genetic Algorithm

Abstract

As traditional, robot manipulator inverse kinematics solutions are computationally intensive, numerous algorithms has been introduced to minimize the time required to compute inverse kinematics. In this paper an Artificial Neural Network (ANN) was utilized to solve the inverse kinematics of a robot manipulator, and a Genetic Algorithm (GA) was used to optimize the ANN weights and biases. This research primarily examined how crossover and mutation variation affect inverse kinematics problems solved with a GA optimised ANN model. In GA optimisation, single point and two-point crossover were performed to test the crossover effects while Gaussian and uniform mutations were used in the mutation. The ANN GA optimised inverse kinematics model was trained and tested using the Forward Kinematics data set. The single point crossover and Gaussian mutation produced the best results based on the results for the generated data. In conclusion GA can be used with variety of parameters to achieve optimal results which may differ depending on the parameters chosen.