Autonomous Planning For Agile Earth-Observing Satellite Integrating Task And Action

In the future,the earth observation system can be able to provide services according to dynamic observation demand normalization,accept observation demands proposed by users at any time,and answer what changes have taken place in what objects at what time and what place in a rapid and accurate manner.However,such efficient and convenient user experience brings a great challenge to organizational processes of earth observation,especially that,it puts forward a higher requirement for observation mission planning modes.At present,planning schemes implemented by a vast majority of earth observation satellites are produced by a ground mission planning system,after which,they can be uploaded to the satellite by virtue of a satellite-to-earth link.Such a supervision mode of Ground-Heaven Loop ranges from demand collection,data acquisition to product generation.Without doubt,the entire process requires a very long time so that it has no chance to satisfy a timeliness requirement for earth observation system in the future.Comparing with the Ground-Heaven Loop,onboard autonomous planning technology has a giant advantage in terms of improving quick response ability of satellites to dynamic observation demands,enhancing cooperation ability among satellites,reducing the dependency on ground test and control systems,etc..Therefore,it is an extremely appropriate technique that can be adopted to resolve planning problems related to dynamic observation demand normalization.Main research contents of this paper are as follows.(1)Issues about autonomous satellite planning were analyzed to define the relevant problems and research category,summarize features and difficulties of these problems,and propose fundamental assumptions for them as well.In addition to the completion of onboard autonomous planning module architecture design,relationship among submodules within the module was clearly pinpointed together with their functions.According to functional characteristics of each sub-module,autonomous planning problems of agile earth observation satellites were broken up into several sub-problems of coordination control strategy design,onboard task planning and onboard motion planning.On this basis,a solution framework was further presented,that is,autonomous planning problems could be tackled by resolving the above three sub-problems on by one.(2)A coordination control strategy based on finite state machine(FSM)was put forward.By means of shortening planning cycles and the planning range,planning frequency was dramatically raised to repeatedly solve small-scale planning problems,which took the place of one-time large-scale planning problem solving.Due to considerable increases in the planning frequency,every planning can be carried out based on the latest mission information.As a result,dynamic changes in observation demands are timely reflected in optimization.(3)Mathematical abstractions were performed from three perspectives of constraint conditions,decision variables and optimization objectives to establish a constraint satisfaction model for onboard task planning sub-problem.On the basis of such a problem model,a variable neighborhood search algorithm was designed.In addition,a 2-dimensional coding method,three neighborhood structures,multiple heuristic rules for observation mission sequence generation,and a computing method for decision variables were proposed.As demonstrated by experiments,the variable neighborhood search algorithm has the capability to effectively solve sub-problem of onboard task planning.(4)A constraint satisfaction model that took 9 satellite actions into consideration(e.g.,earth observation activity,downloading activity,sun-pointing activity,geo-pointing activity and attitude transformation activity,etc.)was set up for onboard action planning sub-problem.By analyzing the composition relationship among satellite actions,not only were 9 action sequences needed to accomplish various satellite actions summed up,but a three-step variable neighborhood search(TVNS)algorithm available for action sequence oriented searches was raised.Relevant experiments have indicated that the onboard action planning sub-problem can be preferably resolved by such an algorithm.(5)To validate diverse key technologies presented in this paper,an experimental autonomous satellite planning system was also designed on the basis of above research achievements.