Hybridisation of Neural Networks and Genetic Algorithms in an Application of Time-Optimal Control

This paper presents the use of neural network and genetic algorithms in the time-optimal control of a closed loop robotics system. Radial basis function networks are used in conjunction with PID controllers in an independent joint position control to reduce tracking error. Genetic algorithm is then used to solve a multi-objective optimisation problem where decision variables are torque limits on each joint and the objective variables are trajectory time and position tracking error. This represents a task hybridisation between neural network and genetic algorithm. Two approaches with genetic algorithms are used to solve this optimisation problem: Multi-objective Genetic Algorithm (MOGA) and genetic algorithm with weighted-sum approach.