Input selection for learning human control strategy

In this paper, we study the input selection in reducing the problem of the high dimension of input variables severely affecting the learning control performance of artificial neural networks. We first locally transform a nonlinear mapping problem into a nearly linear one by using the first-order derivatives of it. Then, we performed a local measure of the sensitivity of each of the model inputs (state variables) with respect to model outputs (human control inputs) under the least square error standard. Finally, based on voting, we defined a determination-rule to decide the importance order of the system state variables globally. By abstracting a human expert skill for controlling a dynamically stabilized robot: Gyrover, we validated the proposed approach.

Authors

• Yongsheng Ou
• Yangsheng Xu

Keywords

• Humans
• Input variables
• Control systems
• Automatic control
• Automation
• Computer networks
• Artificial neural networks
• Performance evaluation
• Least squares methods
• Error correction
• Artificial Neural Network
• System State
• State Variables
• Control Input
• Human Experts
• Learning Control
• System State Variables
• Changes In Parameters
• Experimental System
• Control Problem
• Linear Approximation
• Least Squares Estimation
• Rotation Axis
• Angular Momentum
• Moment Of Inertia
• Vector Data
• Gyroscope
• Curse Of Dimensionality
• Global Order
• Selection Of Input Variables
• Sensitivity Coefficient
• Unit Model
• Flywheel
• Local Sensitivity
• Wheel Speed
• Independent Normal Distributions
• Linear Unbiased Estimates
• Training Data