With the development of rapid-response Earth-observing techniques, the demand for reducing a requirements-tasking-effects cycle from 1 day to hours grows rapidly. For instance, a satellite user always wants to receive requested data in near real-time to support their urgent missions, such as dealing with wildfires, volcanoes, flooding events, etc. In this paper, we try to reduce data transmission time for achieving this goal. The new feature of a responsive satellite is that users can receive signals from it directly. Therefore, the traditional satellite control and operational techniques need to be improved to accommodate these changes in user needs and technical upgrading. With that in mind, a data transmission topological model is constructed. Based on this model, we can deal with the satellite data transmission problem as a multi-constraint and multi-objective pathscheduling problem. However, there are many optional data transmission paths for each target based on this model, and the shortest path is preferred. In addition, satellites represent scarce resources that must be carefully scheduled in order to satisfy as many consumer requests as possible. To efficiently balance response time and resource utilization, a K-shortest path genetic algorithm is proposed for solving the data transmission problem. Simulations and analysis show the feasibility and the adaptability of the proposed approach.
With notably few exceptions, the existing satellite mission operations cannot provide the ability of schedulability prediction, including the latest satellite planning service (SPS) standard-Sensor Planning Service Interface Standard 2.0 Earth Observation Satellite Tasking Extension (EO SPS) approved by Open Geospatial Consortium (OGC). The requestor can do nothing but waiting for the results of time consuming batch scheduling. It is often too late to adjust the request when receiving scheduling failures. A supervised learning algorithm based on robust decision tree and bagging support vector machine (Bagging SVM) is proposed to solve the problem above. The Bagging SVM is applied to improve the accuracy of classification and robust decision tree is utilized to reduce the error mean and error variation. The simulations and analysis show that a prediction action can be accomplished in near real-time with high accuracy. This means the decision makers can maximize the probability of successful scheduling through changing request parameters or take action to accommodate the scheduling failures in time.
