The Research On Multi-satellite Scheduling Method Oriented To Forest Resource Observation

Observation on forest resource could provide high resolution remote sensing images of forests.Calibration,processing and recognition based on the obtained remote sensing images could support the related department to adjust forest policy,make forest schedule and evaluate the business effect.Satellite observation has some unique advantages including large observation range and short observation period.As a result,it becomes an indispensable tool for forest remote sensing.Forests usually have large area and are distributed in different geographical locations.In addition,an increasing number of satellites are on orbit which have complex constraints.Therefore,how to effectively make use of the scarce satellite resources to observing forest areas of interests is an urgent problem to be resolved.The key to this problem is to design efficient satellite scheduling techniques,which help to generate high-quality satellite observation schedule to observe forest areas.This thesis investigates the multi-satellite schedule problem oriented to the forest observation.Research achievements and innovations are summarized as below.(1)Based on the working principle and mechanisms of satellites,the earth observation process and task organization process,the mathematical model for multi-satellite schedule problem is established.First,the factors involved in multi-satellite scheduling process are analyzed,including as observation tasks,satellite resources and optimization objectives.Second,the mathematical model of satellite observation scheduling problems is constructed based on the task requirements,capabilities of observation resources,constraints and optimization objectives.(2)Since a satellite cannot cover an area target with one observation activity,we propose a novel area target partition method.First,the forest observation area is transformed into a polygonal geometric target.Then,a decomposition method and standard flow for area targets is proposed on the basis of resolution of the related observation resource,observation type and time-window.In addition,methods for partitioning area targets into grid points and merging adjacent point targets are designed.After the decomposition process,meta-tasks are obtained,which provide inputs for subsequent scheduling process.(3)To solve the multi-satellite observation scheduling problem of big size,complex constraints and NP-Hard characteristic,a novel satellite scheduling framework based on the divide and conquer principle is proposed,an efficient task assignment and a single orbit scheduling method are presented.Under this framework,first an ant colony optimization algorithm is employed to distribute observation tasks to different satellite orbits.Second,an adaptive simulated annealing algorithm is designed to solve the satellite observation problem involved in each orbit.Then,according to the feedback on the scheduling results at each orbit,the task distribution schema will be adjusted.This process is repeated until the termination condition is met.To improve the efficiency of the algorithm,the domain knowledge of the satellite scheduling problem is considered into the heuristic information model of ant colony optimization algorithm.In addition,two neighborhood structures are designed in the simulated annealing algorithm.A dynamic selection strategy is used to choose the most appropriate neighborhood search structure.Extensive experiments show that,the proposed method could reduce the problem complexity effectively.Especially in solving the large-scale satellite observation scheduling problems,it exhibits extraordinary performance.(4)To response to deal with the disturbing situations,like cloud cover,emergent task insertion and resource malfunction,we propose an efficient multi-objective optimization algorithm to solve the multi-objective multi-satellite rescheduling problem.First,the multi-objective rescheduling model is constructed,with minimal deviation from the primary schedule,maximal observation profit and best load balance as the optimization objectives.Second,a hybrid evolutionary multi-objective optimization algorithm is proposed,which integrate scatter search and particle swarm optimization algorithms.Parameter adaptation and population diversity maintenance strategies are used in the hybrid optimization algorithm.At last,extensive simulated experiments demonstrate the effectiveness of the multi-objective rescheduling algorithm.