Preprocessing Strategy Analysis of Space Station Flight Mission Planning based on Multi-agent

Abstract: In order to simplify the space station mission planning model and reduce the complexity of the planning algorithm, The main planning preprocessing factors is analyzed according to the diverse event requirements and on-orbit constraints of the planning objects. The preprocessing strategies of different planning factors are researched, and the preprocessing agent models are constructed; the new planning event set is formed through multi-agent cooperation, which is convenient for unified planning design variables. Relevant examples are verified based on the mission planning system, and the results show that the strategy is more reasonable for time-planning and resource allocation, and the proportionality and robustness of the planning is improved.


INTRODUCTION
In order to ensure the safety of the platform, personnel and application efficiency during the long-term operation of the Space Station, it's need continuous and real-time planning and management [1][2][3].Among them, mission planning is mainly to make overall planning for space application test (trial), maintenance and repair, astronauts' in-orbit work and other events within the mission cycle, so as to obtain the arrangement scheme based on the time line (month and week) of a single mission cycle, which serves as the input of the in-orbit daily work plan of the space station [4].In view of various "Event" requirements and on-orbit constraints of planning objects, how to automatically process and transform numerous requirements and constraints, simplify the planning model, and improve the balance and robustness of planning schemes is a problem that needs to be solved.
Therefore, this paper proposes a multi-agent planning preprocessing strategy for planning optimization.On the basis of the analysis of preprocessing factors, the preprocessing methods of different factors are studied, and a multi-agent planning preprocessing model is constructed to improve the automation level and efficiency of planning preprocessing and further reduce the complexity of planning algorithms.

MISSION PLANNING PREPROCESSING OBJECT ANALYSIS
The object of space station mission planning is "Event", such as platform maintenance and repair, rendezvous and docking, orbital propellant replenishment, scientific experiments, and extravehicular activities.These events involve a variety of complex types, and each event includes multiple complex activities.There are also significant differences in the resource requirements and constraint satisfaction for on-orbit execution.The main planning factors included in the event include priority, frequency of occurrence, astronaut man-hours, information transmission, measurement and TT&C support, equipment, self constraints, flight attitude, etc [4][5].
(1) Priority needs to be set for events before planning, assuming that they can be distinguish into five categories: Level 1, Level 2,…, and Level 5 events (from high to low).Among them, high-priority events should be evenly distributed throughout the mission cycle, reducing the conflict of on-orbit resource requirements during the implementation of important events, to enhance the enforceability of high-priority events in the planning scheme, and improve the security of the planning scheme; other events also need to be distributed evenly over the mission cycle based on priority to improve the robustness of the planning scheme.Therefore, high-priority and same-priority events require planning and preprocessing (2) According to the frequency classification of events, the preliminarily identified event types mainly include five categories: single occurrence event, continuous occurrence event, periodic occurrence event, non periodic multiple occurrence event, and random event.Flight mission planning mainly involves arranging activities for a single event, so it is necessary to preprocess multiple events into multiple single events that can be recognized by the planning system.
(3) During the in-orbit operation of the space station, there are some long-term continuous events, spanning several months, half a year or more, which are not suitable for unified planning within the mission cycle and need to be handled in advance according to its resource requirements.
(4) Currently, for events where the number of astronauts has been designated but a specific astronaut has not yet been designated, it is necessary to assign a corresponding astronaut to execute the event according to a certain strategy before planning.When reasonably arranging the man hour, it is necessary to give full play to the work efficiency of the astronauts.
(5) Some events have complex constraints, especially applied experiment events, such as scientific load experiment and extravehicular load test.Before planning, it needs to further specify the time range for the event to be executed based on the constraint requirements to improve the reliability of the planning result.
(6) After analysis, the following types of planning preprocessing objects need to be further studied for preprocessing strategies related to mission planning requirements, so as to facilitate the unification of planning design variables and reduce the complexity of planning algorithms.
(5) Continuous occurrence of events.(6) Events in which astronauts are not assigned but are required.(7) Application load events with complex constraints.

PLANNING PREPROCESSING AGENT DESIGN
Mission planning of space station is a complex planning problem with multiple constraints, which aims to determine the specific execution of events in the time line.Its essence is to plan the start time of events [6][7].Therefore, the event occurrence time is taken as the basic design variable (v) of mission planning, and the set of planning variables V can be obtained.
"m" means the amount of planned events.Range: The value of each design variable is within the range of the earliest start time ( ) and the latest end time ( ) of the corresponding event.
Therefore, in order to unify the design variables in the planning process and reduce the complexity of the planning algorithm, it is necessary to perform consistency processing on the previously analyzed planning preprocessing objects, build an automated preprocessing process, design preprocessing agents for different objects, and achieve automatic conversion of planning objects to design variables.

Construction of preprocessing Agent
3.1.1Preprocessing agent of high priority events.Based on the above analysis, in order to ensure that all high-priority events within the mission cycle can be executed, it is necessary to perform balanced allocation preprocessing and evenly distribute them within the planning cycle.Build a high priority event preprocessing agent, and the preprocessing steps are as follows: (1) Obtain Level 1 events and attributes from the planning event library ( ), and count the number "Num".(2) Sort the total duration of events ( ) from largest to smallest to form an event list.
(3) Divide the cycle time into Num time segments (the duration is T ), and take the start time ( ) of the time segment to form a time list ( ).
(4) Determine if there are Level 1 events with specific time requirements?If yes, mark the corresponding time segment as occupied.
(5) Assign the unoccupied values in the time list to the start time attribute item of the remaining events in the event list in turn . .
(6) Judge whether the new start time meets the time requirement ( . ) for completing the event.If not, the start time of the event is reassigned to the start time of the previous time segment, and cycle through the judgment until a time value that meets the time requirement is found.
3.1.2Preprocessing agent of same priority events.In order to avoid arranging most events in the planning scheme in the early stage of the mission cycle and a small number of events in the later stage, it is necessary to spread the on-orbit events in each month to improve the balance of the planning scheme; at the same time, in the event of resource conflicts during the execution process, it is possible to ensure the execution of high-priority events by abandoning lowpriority events, thereby improving the robustness of the planning scheme.Build the event preprocessing agent with the same priority, and the main preprocessing steps are as follows: (1) Sort out the number of events with the same priority within the mission cycle, and acount the total number ( , with priority level 2 as an example).
(2) Determine whether there is a correlation between events?If so, determine whether .
is a "Concurrent" relationship?If there are clear timing and time interval requirements, package the related events into a large event, count the number of related events, and record them as P.
(3) Assign events with specific time requirements to corresponding months.
(4) Count the number of months in the mission cycle NumMonth, and calculate the first and last months based on the actual number of days/30.
(5) Divide the remaining events into months based on the principle of equal sharing, and set the corresponding event start time to the monthly start time.
(6) Judge whether the working time ℎ . of the last month meets the requirement for the completion time . of the event.If not, reassign the start time of the event to the start time of the previous month; if the execution duration is satisfied, an event with a duration shorter than ℎ . in the previous month will be transferred to the last month.
3.1.3Preprocessing agent of non periodic events.The planning list contains events that occur multiple times in an aperiodic manner, and provides information about how many times the event occurs.However, the ID attribute of the event is the same.In order to facilitate the identification of the planning algorithm, it is necessary to preprocess such events.The main preprocessing steps for building a non periodic multi occurrence event preprocessing agent are as follows: (1) Acquire non periodic multiple occurrence events and occurrence times to form event sequences.
(2) Based on the number of occurrences, the recurring event is decomposed into multiple single events, and a new event ID is generated according to certain rules.The source event ID can be traced, while other attribute items remain unchanged.Acount the events .
• == (4) Repeat this step to process the activity ID and association relationship of all non periodic events that occur multiple times.
(5) Add the decomposed events back to the planning event list.
3.1.4Preprocessing agent of periodic occurrence events.The planning list contains events that occur multiple times in a cycle, but only provides information about the occurrence of the event once.
In order to facilitate the identification of the planning algorithm, it is necessary to preprocess such events into multiple single events.The main preprocessing steps for the event preprocessing agent that occurs multiple times during the construction cycle are as follows: (1) Obtain periodic events and attributes, mainly occurrence frequency and occurrence time range, from the planning event library.
(2) According to the frequency of occurrence, the periodic occurrence events are decomposed into multiple occurrence events of corresponding frequency to form an event list.
(3) Based on the number of occurrences, the periodic event is decomposed into multiple single events, and a new event .number is generated according to certain rules.The source event . number can be traced, while other attribute items remain unchanged.Acount the events .
(4) Divide different time segments according to the number and time range of events.
(5) Assign the values in the time list to the start time attribute of the event in the event list in turn.
(6) Repeat the above steps to preprocess all periodic events and form a periodic event list.
(7) Add the events back to the planning event list.

Preprocessing agent of continuous occurrence events.
There are some long-term tests/experiments in space application systems, which are reflected in the requirements as long-term continuous execution of the event.In order to ensure that these long-term events can be scheduled, preprocessing is required before planning.The main preprocessing steps for building a continuous event preprocessing agent are as follows: (1) Count the number of events with a long duration, sort them by priority, and sort them by execution time for the same priority to form an event list.
(2) Determine whether there are events with clear start time requirements?If any, first arrange the event according The device provides the available load at any time during the mission cycle.
3.1.6Preprocessing agent of astronauts assigned.In the event of mission planning, there is a situation where there is no designated astronaut, but there is a need for 1 to 3 astronauts.To ensure that the correct astronaut information is obtained during the planning process, it is necessary to assign an astronaut to this event.The main preprocessing steps for building an astronaut allocation preprocessing agent are as follows: (1) If it is judged that there is a man-hour demand for the event ( . ) but no astronaut is specified, then record the number of astronaut .(2) Given that the time required for astronaut participation is not long, a rotation mechanism is adopted to assign astronauts ( ) to the corresponding events.
Where "k" is a global variable, "P" is the number of astronauts, ].
3.1.7Preprocessing agent of complex constraints events.In mission planning objects, the diversity and complexity of application load events bring more complex constraints.According to the general load constraint model, constraints cover many factors such as field of view, ground targets, TT&C, SAA, and space environment, resulting in a large constraint model for on-orbit events and complex solution algorithms and processes.Therefore, in order to reduce the variable parameters in the planning process, it is necessary to simplify some constraints before planning and solve them as time constraints for the event.The main preprocessing steps are as follows: (1) Obtain the constraint model of the event . .
(2) Obtain nominal orbit, sun position, measurement and control station position and other data as required.
(4) Take the solving time interval of each constraint of the event, calculate the intersection ([ , ]), eliminate the time interval smaller than the event duration, and determine the final executable time interval of the event .
(5) Repeat the above steps to preprocess events with related constraints and update the mission planning event database.

Multi Agent Preprocessing Process
Currently, there are seven types of preprocessed objects that have been sorted out.Some preprocessed objects have an impact on other objects, such as periodic or non periodic events that occur multiple times, which can affect events of the same priority.Therefore, it is necessary to clarify the preprocessing order of different objects [8][9].To ensure the data consistency of the planning event set, a master agent is constructed, scheduling different preprocessing models at three levels, preprocessing the planning objects, and forming a new planning event set.The main process is shown in Figure 1.
1) The first procedure is to initialize the planning object, combine the periodic multiple occurrence event and non periodic multiple occurrence event preprocessing agent model, decompose the events, and form the latest event set as the planning input.2) The second procedure is to preprocess events with resource requirements.The master agent counts events with man-hour requirements and forms a new event list.The astronaut assigns the preprocess agent to assign astronauts to all events according to the strategy.
3) Then, according to the preprocessing strategy, simplify the planning factors for persistent events and application load events, initialize some attributes, and count the resource requirements for some events.
4) The third procedure is to initialize all event planning start times based on different event priorities, forming the final planned event set.
After this preprocessing is completed, an event list with initial time attributes is formed.Planning based on this event list can reduce the complexity of the planning algorithm on the one hand, it can improve the balance and robustness of the planning scheme on the other hand.

EXPERIMENTAL ANALYSIS
In order to verify the preprocessing strategy proposed in this paper, the flight event group in the actual mission is designed as an example, both use the multi-agent preprocessing method and nonpreprocessing method to program solution, Compared to verify the validity and rationality of this method.The example is derived from the mission requirements of each system in the actual mission, and the simulation data is formed after processing, including 400 original events to be planned.Table 1 shows for other setting input constraints.
400 examples are sorted and classified according to the preprocessing Agent.The analysis is shown in Table 2, where the same event belongs to multiple categories in different dimensions.
Level 3 events contain 13 non-periodic multiple occurrences and 40 periodic multiple occurrences.According to the processing process, the 40 periodic multiple occurrences are disassembled into 129 missions according to the frequency and time range requirements, each mission is assigned its attribute information and constraints in turn.The 13 non periodic events are disassembled into 94 missions and are assigned corresponding attribute information.A new set is formed after the above preprocessing is completed.Events with astronauts needs are selected, and events that are not clearly designated astronauts are rotated and distributed.Currently, 3 astronauts are set in orbit, and designated when the working time constraints of astronauts are met.The continuous events are sorted to form an event list, and the complex constraints in the events are processed into executable time intervals.Five level events are sorted according to their priority, and each level of events is further sorted to form an event chain.Judging the events with time constraints in the event requirements, there are 15 level 1 events with specific time requirements.
The method based on multi-agent preprocessing and nonpreprocessing is used for programming and solving.The results are shown in Table 3. (1) As can be seen from the comparison of different planning methods in Table 3, the calculation speed difference between the multi-agent preprocessing method and the non-pre-processing algorithm is basically consistent.The pre-processing multi-agent preprocessing method not only ensures the complexity of the preprocessing logic, but also reduces the complexity of the planning algorithm and retains the algorithm performance.
(2) Pruning events in Table 3 represent events that cannot be scheduled due to time or resource conflicts.In general planning, events such as repeated events and continuous events cannot be arranged due to conflicts with other events or resource conflicts during the continuous process.The multi-agent-based preprocessing strategy dismantles the projects accordingly, releases part of planning resources, arranges more unplanned events into the planning, and reduces the reduction events.
(3) Figure 2 and Figure 4 are the results of the strategy in this paper, Figure 3 and Figure 5 are the non-pre-processed method    results, Figure 2 and Figure 3 are the Gantt display charts of event arrangement in the whole 6-month mission cycle, with color discriminating the priority of events.By comparing, it was found that a larger number of event arrangements and a more uniform distribution of priority events through the pre processed algorithm.Figures 4 and Figures 5 show the arrangement of events within a certain locality time period.By comparison, it can be found that the final result generated by non-pre-processed method will arrange the events without time constraints tightly, resulting in parallel mission stacking.Table 4 Statistical results show that after Agent method preprocessing, except for events with specific time constraints, other events without specific time constraints or events with a wide range of time constraints are distributed evenly in the planning.High priority events and events with the same priority are distributed evenly on the basis of ensuring feasible constraints.The number of comparison missions in the time interval is basically the same.(4) Mission planning in the case of does not specify an astronaut, astronauts take rotation mechanism distribution.The results in Figure 6 show the distribution of astronauts in each week according to the natural week.It shows that the working hours of astronauts in each week are basically the same.

CONCLUSION
In this paper's method for space station mission planning is designed, and the preprocessing objects are executed, and unified design variables are constructed to ensure the computational efficiency of the algorithm.The experimental results show that by constructing preprocessing agents for different objects, the rationality of time and resource allocation in the planning cycle can be improved, and the robustness and balance of the planning scheme can be enhanced.

Figure 2 :
Figure 2: Planning result of multi-agent preprocessing method.

Figure 3 :
Figure 3: Planning result of non-preprocessing method.

Figure 4 :
Figure 4: Locality planning result of multi-agent preprocessing method.

Figure 5 :
Figure 5: Locality planning result of non-preprocessing method.

Table 2 :
Input classification statistics

Table 3 :
Comparison of planning results of different methods Multi-agent Preprocessing method

Table 4 :
Planning result of multi-agent preprocessing method statistics