Study on Automatic PID Gain Adjustment for a Four-rotor Flying Robot using Neural Network

A PID-gain auto-adjustment method using the neural network method with little computational complexity is proposed. The automatic PID gain adjustment technique based on the neural network can adapt to modeling errors and unknown disturbances by performing on-line learning during flight. When the robot becomes unstable due to overlearning, learning process is reset once. In addition, the object tracking, and obstacle avoidance systems are also developed to make the robot adapt to complex environment.