| Since the 21 st century,national aerospace industry has developed rapidly with an explosive increase in satellites.To coordinate the in-orbit satellites and their management resources for better social,economic,and military applications,the satellite task scheduling is required.As the satellite management and application manners change over the years,cross-agency and cross-model have become the new normal,and the integration and quick-response have become the new requirements.Under such circumstances,the“one satellite,one system” shortcoming in system development and application occurs,while those systems that differ in agencies,modes,and types cannot accommodate each other;hence,the generality of satellite scheduling models and algorithm must be strengthened.To address the “one satellite,one system” shortcoming,the cooperative,flexible,efficient applications,and the monopoly of the U.S.STK/Scheduler,a Satellite Task Scheduling Engine that contains general-purpose modeling and optimization methods is studied in this dissertation.The main contents include:1)A top-level framework of the satellite task scheduling engine is designed.Under the application background of the Satellite Task Scheduling Engine,four major satellite scheduling problems,including the remote-sensing satellite scheduling,relay satellite scheduling,navigation satellite scheduling,and satellite range scheduling,are defined.The scope of this dissertation and its reality-based and application-oriented principle are also determined.Then,the definition of the Engine and its functional requirements are given.On this basis,a top-level Engine framework that decouples the model,normal algorithm,and urgent algorithm is designed.With this framework,a new,general-purpose,and modularized modeling and optimization manner is presented for different satellite task scheduling problems,guiding the follow-up studies of models and algorithms in this dissertation.2)A general-purpose modeling method for satellite task scheduling is proposed.To address the model generality and “one satellite,one system” issues,the satellite task scheduling problem is systematically and hierarchically generalized to a task set,a resource set,a score set,and a decision matrix.Then,the decision-relation between the task set and the resources set,definitely between the satellite event and its executable opportunity,are creatively explained.With this relation,the general-purpose 0-1 mixed integer decision model is then built to play the key coupling point in the loosely coupled Engine framework.Based on this model and several examples,the constraints in satellite task scheduling problems are summarized and modeled via a general-purposed template,which is composed of an object,a threshold,and a relationship.The constraints can construct a network,and the constraint value is calculated incrementally and efficiently.Finally,the objective functions are formulated to complete the whole model.In summary,this general-purpose modeling method decouples the decision,constraints,and objectives,which can properly model the aforementioned four satellite task scheduling problems in a general-purpose manner.This manner opens up a new idea of general-purpose and refined modeling of satellite task scheduling problems,supporting the required generalpurpose models for the Satellite Task Scheduling Engine in this dissertation.3)An adaptive parallel memetic algorithm(APMA)for the normal satellite task scheduling is proposed.To address daily and weekly normal satellite task scheduling requirements,the APMA is proposed considering several algorithm factors,such as the initial solution,local optimization,global optimization,adaptivity,generality,and complexity.The APMA integrates four complementary strategies,including a heuristicbased initialization strategy,a parallel-search-based local optimization strategy,a competition-based algorithm and operator selection strategy,and an evolution-based local optimization strategy.Some Benchmark experiments,including the Orienteering Problem and the simplified remote-sensing satellite scheduling,show APMA’s generality and outperformance.As a result,the APMA provides a general-purpose and efficient way to address the normal satellite task scheduling requirements,supporting the required algorithm for the satellite task scheduling engine in this dissertation.4)A distributed dynamic rolling optimization algorithm(DDRO)for the urgent satellite task scheduling is designed.To address the urgent satellite task scheduling and time-consumption of normal scheduling algorithms in real-world situations,the DDRO is designed considering the urgent scheduling requirements.The DDRO also integrates four algorithm strategies,including a dynamic contract-net-based task negotiation and assignment strategy,a rolling-based single platform re-scheduling strategy,a schedulability-based task quick-insertion strategy,and a constraint-net-based real-time deconflicting strategy.With these strategies,the DDRO performs real-time and dynamic optimization on normal satellite task scheduling results,which offers a general-purpose and flexible way for urgent satellite task scheduling problems,supporting another required algorithm for the Satellite Task Scheduling Engine in this dissertation.5)The satellite task scheduling engine was applied to solve real-world problems.Based on the aforementioned contents,the Satellite Task Scheduling Engine was applied to solve the Super View-1 remote-sensing satellite scheduling,Tian Lian-1 relay satellite scheduling,Bei Dou-3 navigation satellite scheduling,and the satellite range scheduling from U.S.Air Force Institute of Technology.The experiments well examine the proposed general-purpose modeling method,the APMA,and the DDRO.In a word,the Satellite Task Scheduling Engine is proven feasible and promising in the dissertation,offering great potential in more real-world applications. |
