Mission Planning Method For Deep Space Probe Based On Constraint Satisfaction

With the increase number of deep space probe tasks and the increasing complexity of on-board scientific tasks,the autonomous task planning and scheduling of deep space probes has become a hot spot of research.It is of great significance to improve the robustness of remote detector,increase the scientific return and decrease the cost of transportation and control.However,compared to the classic task planning and scheduling,task planning of deep space detectors has the characteristics of unpredictability,suddenness,complexity and dynamics,which bring difficulties and challenges to the research of this problem.What's more,the complexity of detector system and the strong coupling of planning and scheduling under variety constraints make the traditional“planning-scheduling” linear processing mode unable to satisfy the need autonomous planning and scheduling of deep space detector.On the basis of the characteristics of the deep space detector and the analysis of the system constraints,this paper combines the intelligent planning theory with the constraint satisfaction technology,and studies the key technologies in the autonomous task planning and scheduling of the deep space detector.The main research contents and results are as follows:First,the relationship between the detector constraints and the activities in the deep space detection tasks is analyzed,and the multilevel constrained programmed planning based on multivalued variable(MCPP)is constructed.On the basis of the study of the semantic expression of planning actions and the transformation relationship between the planning problem and the constraint satisfiability problem,the domain knowledge in the planning model is extracted.The causal chains of the actions are transformed into the constraint form in the constraint satisfiability problem,and then the deep space exploration planning problem is transformed into a constraint satisfied problem.The process of planning and scheduling in deep space detection tasks is uniformly coded by extension constraints,and the task planning and scheduling model based on constraints can be established.This method helps to avoid the manual coding of the domain constraints in the planning,reduces the coupling between the model design and the code implementation,and improves the portability and independence of the system model.Secondly,based on the MCPP model of the deep space detector and the mutual exclusion of the actions in the planning problem,the dynamic characteristics of the constraints in the multilevel constrained programming model are studied.A table constraint fast filtering algorithm is designed.The algorithm classifies the newly added activities according to the conflict determination between the activities in the domain information.The consistency of variables in the constraints table is checked.The results show that the fast filtering algorithm based on the dynamic constraints set can effectively reduce the number of invalid constraints checking in the constraints processing and reduce the backtracking of the algorithm in the problem processing.Then,the quadratic constraints satisfaction algorithm framework of deep space exploration planning is designed,and the deep space exploration planning problem with complex constraints and highly coupled system state information is decomposed to two levels,"subsystem-action level" and "action instantiation level".In the subsystem-action level,we propose the concept of virtual action,and design a constrained planning algorithm based on virtual action as heuristic information.Combining the construction strategy and local search strategy in the constraints satisfaction technology,virtual action algorithm can effectively reduce the process of redundant variables processing in the planning process and get the feasible action sequence quickly.In the action instantiation level,combining the characteristics of the extension constraints,a heuristic backtracking algorithm based on MCPP is designed to instantiate the multi-value variables.By reducing the redundant branches of adjacent branches,the final planning solution is quickly generated.Finally,in order to verify the effectiveness of the constrained planning algorithm designed in this paper,we design and implement an intelligent platform for detector mission planning and scheduling.Taking the Mars Express task model as an example,the constraint coding technology proposed in this paper,the fast filtering algorithm based on dynamic constraint set,the influence of the virtual action algorithm and intelligent backtracking algorithm on the planning process under the framework of the quadratic CSP algorithm are verified respectively.The calculation results show that the proposed algorithm can fully excavate the domain information on the premise of ensuring the portability and independence of the system,thus reducing the number of backtracking,improving the planning efficiency and generating a better planning result.