An Approach Of Autonomous Satellite Observation Task Onboard Continuous Planning For Dynamic Requests

Earth observation satellites obtain information of the Earth's surface with the advantage of spatial position and the obtained data has been widely used in the socio-economic and national security fields such as geographic mapping,land resource survey,disaster monitoring,and military intelligence application.With the continuous deepening and expansion of the application field,the timeliness requirements of the observation data are getting higher and higher.The traditional ground task planning mode cannot revise the onboard observation plan timely to users' observation requests due to the limited communication time window between satellite and ground,which makes it difficult to fully exploit the application advantages of satellites.Onboard task planning can reduce the dependence of satellites on the ground system and improve the responsiveness that attracts much attention from various space powers and scientific research institutions.At present,the existing onboard task planning methods mainly focuses on the responsiveness to dynamic requests(e.g.,new arrived observation requests),and there is still room for improvement in optimization.Because of the acute contradiction between limited satellite resources and increasing observation requests,how to improve the responsiveness of satellite to dynamic requests,while maximizing the satisfaction of users' observation needs has become an urgent problem to be solved.With regard to the above problem,the thesis studies the autonomous satellite observation task onboard continuous planning method for dynamic requests.The main contributions are as follows:(1)Based on the analysis of the related elements and basic assumptions of onboard task planning,as well as the uncertainty of the observation requirements and the limitation of the computing resources,a framework for solving the problem of the autonomous satellite observation task onboard continuous planning is proposed.Based on the framework,the problem is decomposed into three sub-problems: satellite autonomous planning method for deterministic tasks,satellite autonomous planning method for the situation of tasks changing,and satellite autonomous task planning strategy for dynamic requests.(2)For the problem of satellite autonomous planning method for deterministic tasks,a directed acyclic graph model with the objective of maximizing the total profit is established.And a path search algorithm based on graph partition strategy and path selection strategy is proposed and the completeness of the graph partition strategy is proved in theory.Based on this work,a path incremental search algorithm based on rolling optimization strategy is proposed.The experimental results show that the above two algorithms can obtain the approximate optimal solution to the problem in a reasonable time.(3)For the problem of satellite autonomous planning method for the situation of tasks changing,an observation task sequential decision model is established to improve the responsiveness to dynamic observation tasks.Then an observation task sequential decision algorithm based on ensemble learning and an observation task sequential decision algorithm based on deep neural network are proposed from the perspective of feature engineering and representation learning,respectively.These two algorithms can effectively overcome the weakness of the existing task re-planning methods which depends on the existing observation plan and the calculation time is easily affected by the number of observation tasks.Experiments show that the above two algorithms are superior to the existing heuristic search algorithm in computational efficiency and optimization.(4)For the problem of satellite autonomous task planning strategy for dynamic requests,the uncertainty of dynamic requests is taking into consideration and a task planning strategy decision algorithm based on convolutional neural network is proposed.The task planning strategy combines the advantages of satellite autonomous planning method for deterministic tasks and satellite autonomous planning method for the situation of tasks changing in optimization and computational efficiency.The simulation results show that the new planning algorithm after combination can respond to the dynamic requests in time,and is better than the single type of task planning algorithm in optimization,which verifies the feasibility and effectiveness of the task planning strategy decision algorithm.