Application Research On Control Algorithm Of An Airborne Three-axis Stabilized Platform Based On Neural Networks

The airborne camera mapping and range detection technology has progressed rapidly and has been widely used in military application, public security, forest fire prevention and environment monitoring. In order to capture stabilized and clear image, the airborne platform to carry the camera payload needs to sustain the wind disturbance torque and aircraft vibration while in the CCD integration and ensure that the camera is stabilized in the imaging visual axes. The accuracy of control of the stabilized platform is critical to the performance of airborne camera imaging system. The objectives of this paper is to study the three axis airborne stabilized platform, conducting lab experiments and research in depth of the platform servo control techniques.This paper summarizes current global and domestic research status of the airborne camera imaging system and the stabilized platform, and introduces some well known control algorithms and describes the characteristics of the electric platform and mathematics models of the stabilized platform servo control system and disturbance torques. Through the ADLINE neuron experiments to derive the parameters in the platform servo system. Based on these parameters to do lead and lag compensation to the servo system. And analyzes the compensation results of in the frequency domain.The paper also discusses the parameter auto adaption solution base on the neuron network to solve the wind disturbance torques and load changes during the operation. based on the analysis of the PID neural network theory, topology structure, methods and initial values selection, describes the advantages and disadvantages of using the PID neural network as an auto adaption compensator for the servo system. With a reference to principle of selecting the initial weight value, this paper discusses the an optimization selection based on the generic algorithm. And further discuss an auto adaption control algorithm based on PID neural network, which combines the normal frequency domain compensation and neural network auto adaptive compensation. This algorithm can not only compensate the control system, but also effectively compensate the disturbance and load changes. The paper discusses in depth the application of using the PID neural network reference model auto adaption, analyzes the weight value stability of the servo system during the convergence time, defines the step range, to assure the conditions of the Lyapunov stability theory are still satisfies when the weight values of the neural network change. Therefore ensures the stability requirement are met when servo system automatically adapting. The paper discusses a method to select the initial weight value for the neural network based on the generic algorithm, to ensure the initial system stability and make it very close to the global optimal solution. The results is approved and verified by the experiment.