Decision Support Tool for Prediction of Critical Da ta to the Satellite Integrity

The prospect of multiple launches by INPE's satelli te program, in the near future, has motivated the development of an application using t emporal planning techniques based on Artificial Intelligence (AI) concepts. This will be used for automatic generation of flight operation plans to control satellite activities. Ho wever, making a critical analysis of these plans before real world implementation is not possi ble. We propose a different approach as part of a strategy to validate these plans. This wi ll use a decision support tool AI-based data mining technique for data prediction, to assist exp erts in evaluating the performance of the plan, with the aim of maintaining satellite integri ty. I. Introduction here is general interest in automating satellite co ntrol operations related to the task of controlling multiple satellites in INPEs Space Program. In addition, i t is generally accepted that the automation of sate llite control activities represents a way of reducing in-orbit sa tellite maintenance costs. At INPE, autonomous syst ems to control satellite operations employing Artificial Intellige nce are being developed to automate ground segment operations. However, this increased autonomy in satellite contr ol operations can lead to distrust of the automatic control system behavior as compared to that of the well kno wn and routine manual control system. In such cases , these systems still require an improvement in reliability to become operational. In order to achieve this breakthrough in reliabilit y, predictability and safety, an AI-based strategy for automatic validation of a flight operations plan generated by a planner is presented. This is an architecture co mposed of software components, resulting from the combination of verification and validation techniques. As a re levant part of this strategy, a decision support tool is proposed in this article, to assist experts in evaluating th e actions of the plan, aiming at guaranteeing the integrity of the satelli te. This tool consists of software using Artificial Intelligence techniques aimed at predicting the behavior of crit ical platform subsystems, such as the power supply subsystem, directly affected by the actions contained in each flight operation plan. This paper presents in the following section some c oncepts related to the automation of the control ac tivities of the satellite in orbit. Section 3 describes the str ategy for validation of a flight operations plan, a n overview of the relevant architecture and general behavior. Section 4 discusses some Artificial Intelligence technique s for data prediction, such as Bayesian methods which make the required learning a form of probabilistic inferenc e, and data mining. Section 5 presents the design of the tool b eing developed for prediction. Conclusions are pres ented in Section 6.