Host: The Society of Instrument and Control Engineers, System Integration Division
Pages 37
In this paper, a designing method of the swing-up control for an Acrobot(acrobat robot) is proposed, which is used a real-coded genetic algorithm and a neural network. Firstly, the gene arrangement of one individual is assumed as one swing-up pattern, and one cell of the gene arrangement rates as the input signal of the actuator for every 1 step time. By using the genetic algorithm, the swing-up pattern for an Acrobot is generated. Next, the optimal swing-up pattern obtained by it is used as the teacher signal for the neural network. Then, the weights and structure of the neural network are optimized by using another genetic algorithm. The neural network is used as the swing-up controller.
