Modeling and Simulation of Payload System Design in Remote Sensing Satellites Based on Adaptive Neuro-Fuzzy Inference System

One of the main parts in modeling and simulation of space systems is to determine the nonlinear behavior of the system. In this article, it is shown how artificial intelligence systems could effectively be used in the early stages of a payload system design process in Remote Sensing Satellites (RSS). Modeling and simulation of space systems with nonlinearity and uncertainty in behavior recognition and decision making is vital. Time and cost of the conceptual design phase are decreased by using adaptive neuro-fuzzy approach, which enables the data that is stored in trained networks to be expressed in the form of a fuzzy rule base. In developing this methodology, a hybrid training algorithm was used to obtain system parameters achieving faster convergence and reduction in the size of the search space. This combined training used both least squares and descending gradient methods. The inference system was also based on the Takagi-Sugeno model with Gaussian membership functions. The present modeling system was implemented on the payload system of a satellite. The simulation results showed the effectiveness of the adaptive neuro-fuzzy inference system in the conceptual design of this system. The mean square error of the output variables for the four variables of mass, power and on-board memory of the payload and data compression in early stage of the design was acceptable.