JPH01201770A - Mission scheduling device - Google Patents

Abstract

PURPOSE:To reduce the number of search times by grasping the scheduling problem of a mission as a packing problem which is to pack a consumption pattern into a supply pattern and using heuristic knowledge. CONSTITUTION:A scheduling order decision means 1 uniformly decides the scheduling trial order of the mission. A scheduling time setting means 3 decides a scheduling time in the order of precedence among discrete values. The order of trial precedence of the mission and the order of search precedence of the scheduling time are decided based on heuristic experienced knowledge which a human being has against the packing problem.

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] The present invention relates to a scheduling device for a mission carried out on a space station.

The aim is to use heuristic methods to overcome combinatorial explosion while finding suboptimal scheduling solutions within practical time and memory limits.

In a mission scheduling device for determining mission scheduling.

Scheduling order determination means for determining the scheduling order between missions, and scheduling that calculates a possible scheduling time by referencing and comparing the resource consumption of a scheduled mission and the amount of resources that can be supplied to that mission. A possible time slot calculation means, a scheduling time setting means for setting a scheduling time with a priority order during a schedulable time slot, and a backtrunk for backtracking a scheduled mission when a schedulable time slot cannot be determined. The scheduling order determining means includes an execution means and an evaluation value calculation means for calculating an evaluation value of the optimality of the schedule solution, and the scheduling order determining means selects the missions in descending order of resource consumption and time consumption length in the case of centralized scheduling, and in descending order of resource consumption and consumption time length of the missions in the case of centralized scheduling, and in the case of decentralized scheduling. In the case of , the scheduling order is determined in descending order of the resource consumption and the maximum consumption value, and the above-mentioned schedulable time slot calculation means and scheduling time setting means determine the scheduling order according to the order of the resource consumption and the maximum consumption value, and in the case of centralized scheduling, the scheduling order is determined in the order of time of a discretely determined time series. , and during distributed scheduling, the scheduling time is determined according to the order of the remaining resource supply amount among those times, and the optimality evaluation value of the schedule solution obtained by this scheduling method is calculated by the evaluation value calculation means. It is configured as follows.

[Industrial application field]

The present invention is for scheduling various experiments (hereinafter referred to as missions) such as materials experiments, life sciences, and scientific observations to be carried out on the Japan Experimental Module (hereinafter referred to as JEM), which will be constructed on the space station in the United States. The present invention relates to a mission scheduling device.

[Conventional technology]

” In the mid-1990s, JEM was built on the U.S. space station, and is scheduled to carry out nine different missions by effectively utilizing the zero gravity, high vacuum, and high-energy particle environment of outer space. The operational plan for implementing such a mission is a three-year plan after a long-term plan is created through international coordination over a time span of about five years.
Semi-annual (45 days) plan - Weekly plan - Daily plan 1
The structure is created in a step-by-step manner.

Among the operation plans structured in this way, ground operation control handles planning from annual plans to weekly plans, which are relatively long-term plans, and creates a weekly plan (base schedule) that corresponds to 7 days of the daily plan. is onboard JE
It will be uplinked to M integrated control. and,
The onboard crew executes missions based on this amplified and linked base schedule.

If the base schedule has to be reviewed due to the progress of the space station or JEM, they will be responsible for creating daily plans and changing the base schedule.

In this way, in JEM, the onboard crew is tasked with the daily task of creating a daily plan for mission operations as necessary.In creating this daily plan, the crew must The experiment start time for the multiple missions must be decided while taking into consideration that the resource consumption such as power consumption and fluid consumption required for implementation will be within the range of JEM's resource supply amount. Figure 11 shows a list of examples of restrictive conditions for implementing a mission.As shown in Figure 11, there are a considerable number of resource restricting conditions for implementing a single mission. Since the number of missions scheduled to be carried out in a day is quite large, about 10, the burden of creating a schedule for carrying out a mission is quite large for the crew, even if it is only for one day. For example, in the case of FMPT (First Materials Experiment), it takes about one week to create a similar experiment schedule for one day.

In the future, there is a need to develop a mission scheduling device that can automatically create a mission schedule through dialogue with the crew. However, participation in an experiment conducted on a space station such as JEM has never been seen before in our country, and the current situation is that there is no prior art that is compatible with the mission scheduling device of the present invention.

From now on, it is necessary to configure mission scheduling devices from a completely new perspective.

[Problems to be Solved by the Invention 3 As explained above, 9 In the present invention, as shown in FIG. Its purpose is to provide an operation scheduling system (abbreviated as "MISES" in FIG. 12). This mission operation scheduling system is implemented for the purpose of reducing the scheduling work of the onboard crew. If you try to solve the problem of combining multiple missions like this under predetermined limited conditions by blind trial and error using a computer,
As shown in the search tree in FIG. 13, [Search tree for the number of combinations of the number of missions:

And regarding the scheduling of one mission, [
There is a "possibility of the existence of infinitely valuable solutions within a schedulable time period," and search trees related to these will result in so-called "combinatorial randomization." In a scheduling system in which such combinatorial explosion occurs, it is virtually impossible to find an appropriate scheduling solution within a practical amount of time.

From now on, the present invention utilizes a heuristic method, that is, the heuristic knowledge possessed by humans, to overcome combinatorial explosion by eliminating techniques that are unlikely to be a solution at an early stage, while at the same time achieving practical results. The objective is to find a suboptimal mission schedule solution within the limits of time and memory.

[Means for solving problems]

FIG. 1 is a diagram explaining the principle of the present invention.

In the figure, l is a scheduling order determining means, which determines the scheduling order among missions to be scheduled. This decision.

During centralized scheduling, scheduling attempts are made in descending order of resource consumption and mission time length, and in decentralized scheduling, scheduling is attempted in descending order of resource consumption and maximum consumption value. Reference numeral 2 denotes a scheduling possible time slot calculation means, and storage means 6.2 stores the resource consumption of the scheduled mission whose mission trial order is determined by the scheduling order determining means 1 and the resource supply amount that can be supplied to the mission. By referring to and comparing 7, a schedulable time slot in the scheduling period is calculated, and a scheduling time with a search priority determined by the optimality evaluation criterion during the schedulable time slot is calculated. This search priority is also in the order of the best solution, in the order of the time of a discretely determined time series in the case of centralized scheduling, and in the order of the largest remaining resource supply amount within those times in the case of decentralized scheduling. The search is configured to find a solution according to 3 is a scheduling time setting means, which removes the one with the highest priority among the plurality of scheduling candidate time elements in the schedulable time zone determined by the schedulable time zone calculation means 2.

Meanwhile, set it as the scheduling time. If there are missions that have not been scheduled yet, the same process is repeated from the schedulable time slot calculation means 2 for those missions.

Reference numeral 4 denotes a backtrack executing means, which executes backtracking of a mission to be scheduled when a schedulable time slot cannot be determined by the schedulable time slot calculation means 2, that is, when scheduling of the mission becomes impossible. Reference numeral 5 denotes an evaluation value calculation means, which calculates an evaluation value of the optimality regarding "mission centralization or decentralization" of the schedule solution obtained by the means 1 to 4 (scheduling method) described above.

[Effect]

In the present invention, as shown in FIG. The system sets the scheduling time within the scheduled schedulable time period, and then generates a resource remaining pattern that will be the resource supply amount for the next scheduled mission (this means that if the resource supply becomes insufficient, If scheduling becomes impossible, backtracking is performed.

In this process, the scheduling order determining means 1 uniquely determines the mission scheduling trial order.

The scheduling time setting means 3 preferentially determines the scheduling time from among discrete values. The trial priority of this mission and the search priority of the scheduling time are determined based on the heuristic experiential knowledge that humans have regarding the backing problem.

These knowledge.

In mission scheduling, only the search branches that are judged to be likely to be successful in scheduling are generated, and techniques with low prospects are discarded.9 Regarding the search for the techniques that have been generated, the meaning of centralization/decentralization is determined at the scheduling stage of each mission. This has the effect of preferentially attempting scheduling based on what is locally determined to be the best option. Therefore.

Suboptimal mission scheduling can be achieved efficiently in terms of computation time and memory.

〔Example〕

Hereinafter, the present invention will be explained in detail according to examples.

A system configuration diagram for realizing the present invention is shown in FIG. 2, in which 20 is ESHELL, 24 is UDF, 2
5 is UTILISP, 26 is TSS, 27 is FOR
TRAN, 28 is a GPS, and 29 is a schedule file. E as an expert system construction tool
The SHELL 20 operates under the environment of the UTILISP 25 to construct an inference engine 21 and a frame file 22 and a KS file 23 that serve as a knowledge base.

tlDF24 is provided to make the functions of UTILISP25 usable, and calculates the remaining amount of resource supply, etc. FORTRAN 27 and GSP2
8 is provided to improve the man-machine interface.

Schedule file 29 contains the requested mission/
The schedule will be stored.

The frame file 22 includes a resource frame 2 that represents resource supply amount information of each resource source of JEM.
2a and a mission frame 22b representing resource consumption information of each mission to be executed in JEM are stored with a frame structure that is an expression of a semantic network. FIG. 3 shows an example of this resource frame 22a, and FIG. 4 shows an example of this mission frame 22b.As shown in FIG. 3, the resource supply amount of JEM resource sources such as electric power and crew changes over time. I will do it 1
In the present invention, this is approximated and expressed by a step-like function, and the resource supply amount patterned in this way is represented by a bare combination of time and value in the resource frame 22a.

The resource frame 22a is configured such that the area representing the resource supply capacity is also described. Similarly, as shown in Figure 4, the amount of resource consumption required by a mission to carry out its execution can also be approximated and expressed as a step-like function with respect to time, and the amount of resource consumption required for the mission to be carried out can be expressed as an approximate function of time. The resource consumption amount is represented by a combination of time and value pairs in the mission frame 22b. The mission frame 22b has an area (total resource consumption) representing the resource consumption characteristics of the mission.

It is configured so that the height (maximum resource consumption value) and length (consumed time) are described together.

Next, the blackboard that will be used as the work area of the inference engine 21 will be explained. A blackboard is an expression of a semantic network, in which the basic unit is a node that associates an attribute with a value of that attribute, and meaning is given by links between these nodes. In the blackboard of the present invention, as shown in FIG. 5, 'Management', 'Plan 1. "resource"
, "Solution" four types of nodes are prepared. This management node manages resource types, scheduling periods, mission types and scheduling methods, the planning node manages characteristic data for mission scheduling, and the resource node manages resource supply patterns. . Plan nodes are linked to each other in the relationship of previous plan and next plan, and plan nodes and resource nodes are linked to each other in the relationship of supply and consumption of resources.

Next, it will be explained how the mission scheduling device of the present invention realizes mission scheduling according to the knowledge source transition diagram shown in FIG. 6. here.

“Initial setting KS” and “possible time period calculation KS” shown in Figure 6
”, “allocation time determination KS”, “new pattern generation KS”,
The knowledge sources “scheduling success KS”, 6 scheduling failure KS” and “back trunk KS” are
It will be stored in the KS file 23 shown in FIG. These knowledge sources basically define the procedure for reasoning.

1F Condition THEN It is described by a production rule in the form of conclusion/action, and the inference engine 21 executes inference based on these knowledge sources using information stored in the frame file 22, thereby determining the mission. Scheduling will be realized.

(b) Initial setting KS The initial setting KS that will be executed first is:

It is a knowledge source for securing a work area, ordering missions, and setting initial values. In executing this initial setting KS, first, the resource type, scheduling period, scheduling mission name, and scheduling method, which are the limiting conditions for mission scheduling, are set in the management node of the blackboard.

Next, the resource frame 22a and the mission frame 2
The mission scheduling order will be determined using the data stored in 2b. This ordering is performed using the resource with the highest resource consumption rate (hereinafter referred to as D) according to the formula below.
(abbreviated as R) Once the area of the resource Ai supply pattern is determined, when the scheduling method is centralized scheduling (scheduling that concentrates missions in the first half of the scheduling period), calculate Si in the following formula for each mission.

Si=α・ki +β・1i ki: Serial number j in ascending order of DR consumption area! i:D
Serial numbers α and β in ascending order of consumption time of R: weighting coefficients for consumption area and consumption time.The scheduling order is determined in descending order of this Si, and 9 decentralized scheduling (
When performing scheduling in which missions are distributed over the entire scheduling period, Si of the following formula is calculated for each mission.

Si=α・ki+r−hi ki: Serial number in descending order of consumed area of DR hi: DR
Serial number α in descending order of maximum consumption value, T: Weighting coefficient of consumption area and maximum consumption value This is determined by determining the scheduling order in descending order of St.

The reason for determining the scheduling order according to this evaluation formula is to take advantage of the human neilistic knowledge that it is easier to find a solution by puffing from the largest one, and also to This is because the remaining amount of resources in a part decreases quickly, so if a mission with a horizontally long pattern comes later in order, scheduling becomes difficult. This is because scheduling tends to become difficult if a mission with a tall pattern comes later in the order.

Once the scheduling order is determined in this way,
As shown in Figure 5, the "mission name" of the management node on the blackboard is re-registered in this order (in the example in Figure 5, As
=A+ -Bz) At the same time, a plan node is generated, the mission names in this order are registered, and a link pointer is attached to the corresponding resource node.

Then, to start scheduling, register the mission name with the earliest scheduling order registered in the plan 1 node in the global variable area in Figure 5 (5th
In the example shown in the figure, A3) and resource frame 2, which represents the resource supply amount of each resource source in the "new pattern" of the resource node to which the plan 1 node links.
2a supply pattern will be set.

(El) Possible time period calculation KS Possible time period calculation KS, which will be executed following the initial setting KS, first calculates the time during which resources can be supplied to the mission of the plan node pointed to by the global variable area. Ask for an obi. An example of possible time slots is shown in FIG. The method for determining this possible time period is shown in FIG.

The resource consumption pattern of a mission is decomposed by the rectangle that is maximally included in the pattern (hereinafter referred to as the maximum included rectangle), and the resource supply pattern of the resource is similarly decomposed by the maximum included rectangle. This is achieved by comparing the size of each rectangle. Through such processing, the possible time slot is determined for each resource linked to the planning node pointed to by the global variable area, and by taking the common part, the possible scheduling time and slot for the mission can be determined. The schedulable time zone obtained in this way is set as the "schedulable time zone" of the planning node.

Furthermore, in this possible time slot calculation KS, scheduling time candidates are selected from among the determined available scheduling time slots. In the case of centralized scheduling, the scheduling candidate times are selected at intervals of a reasonable time width T within the schedulable time slot, and the times in the nine time sequences are selected as shown in FIG. 9(A). Priority is determined according to the order. Therefore, Fig. 9(A)
In the example above, the mission is first scheduled at time t, and if there is a backtrack described later, the mission is scheduled at tt, t3. ... will be retried in this order.

Also, when using decentralized scheduling, as shown in Fig. 9(B),
As shown in , if the schedulable time period is longer than a certain reasonable time width T, the resource remaining capacity pattern is searched and the two points with the largest remaining capacity value are found, and the schedulable time period is shorter than T. In this case, the times at both ends are considered as scheduling time candidates, and the priority is determined in descending order of the remaining amount of resources. Therefore, Fig. 9 (B
), the mission is first scheduled at time tt, and if there is a backtrack, retries will be made in the order of +tj+t4+t3.

The reason why scheduling candidate times were selected according to these selection criteria is that if scheduling fails at a certain time, there is a high possibility that scheduling will also fail in the vicinity, so it should be retried at a time a certain distance away. This is because, in addition to utilizing hierarchical knowledge, in the case of centralized scheduling, the purpose of optimization is to carry out as many missions as possible in the first half of the scheduling period, and in the case of decentralized scheduling, This is because the purpose of scheduling optimization is to consume the remaining resources on an average basis.

The scheduling time candidates obtained in this way are sorted in order of priority and set as the "candidate time" of the planning node.

(8) Allocation time determination KS The allocation time determination KS, which is executed subsequent to the available time slot calculation KS, selects the time with the highest priority among the scheduling time candidate times set in the planning node. This is a knowledge source for setting the time in the "assigned time" and deleting that time from the "candidate time". By executing this knowledge source, the scheduling time to try is set.

(d) New pattern generation KS When the scheduling time is set, the new pattern generation KS calculates the resource supply remaining amount according to the following formula.

Remaining resource supply amount = (Resource supply amount) - (Resource consumption amount) Set the remaining resource supply amount found by this calculation to the "new pattern" of the resource node linked to the next plan node of the plan node pointed to by the global variable area. . In the present invention, both the resource supply amount and the resource consumption amount for all resource sources are expressed in the form of a function that changes stepwise with respect to time, as shown in FIG. 9-1.

The remaining amount of resource supply, which is determined as the difference between these values, is also expressed in the form of a function that changes stepwise with respect to time. One new pattern generation KS is a knowledge source for updating and linking the planning node indicated by the global variable area to the next planning node and issuing a possible time slot calculation KS for scheduling the first mission. It has become. Therefore, the remaining resource supply amount set in the "new pattern" will be used as the resource supply amount for the 9th mission, and an attempt will be made to schedule the mission with the 2nd priority order. The knowledge source issues a scheduling success KS if all missions have been scheduled.

(N) Backtrack KS Backtrunk KS is used when a schedulable time slot cannot be determined by executing the available time slot calculation KS, and there is a "candidate time" for the mission of the plan node pointed to by the global variable area. is a candidate schedule solution for that mission that has not been tried.

In order to try them out, a 1 allocation time determination KS is issued. Also, if there is no "candidate time", it means that the mission could not be scheduled even after trying all the solution candidates for that mission, and therefore the previous mission must be rescheduled (re-scheduled). trial), so we set the global variable area anew to the plan node pointed to by the previous plan link of the plan node it points to, which is the knowledge source for entering the backtrunk KS again. Even the one mission before.

If the result is none, it means that all the search tree techniques generated heuristically have been searched, and a scheduling failure KS is issued.

In this way, by executing this knowledge source, the scheduling of a mission whose scheduling has failed midway through will be retried based on the principle of vertical search.

(to) Scheduling success KS When the scheduling of all missions is completed successfully, this knowledge source is executed and its schedule solution,
That is, the crew is informed of the experiment start times for all missions, and the evaluation value of the degree of concentration or dispersion of solutions is calculated and notified to the crew according to the evaluation value calculation formula shown in FIG. Furthermore, in response to a request, if a search for another solution is entered, a back trunk KS is issued.

(g) Scheduling failure KS This knowledge source is activated when no solution is found.
This is when scheduling fails, or when all alternative solutions have been explored and the process is terminated, and the crew is notified to that effect and the process is terminated.

In this way, in the present invention, the inference engine 21 determines the order of the missions to be scheduled according to the priority order determined by the initial setting KS, and schedules the target missions in the order of the scheduling time determined by the possible time slot calculation KS. will be carried out.

In the above description, centralized scheduling was shown using an example in which scheduling is concentrated in the first half of the scheduling period, but it is also possible to concentrate in the latter half of the scheduling period. In this case, higher priority is given to the later time in FIG. 9(A).

Furthermore, the blackboard used in Figure 5 can be experimented with a frame structure, while the resource frame and mission frame have been explained using the semantic network of the frame structure. It is also possible to express it by other oriented semantic networks. Furthermore, the present invention is not limited to JEM.

A similar effect can be achieved for experiment scheduling at ground-based experimental facilities where resource supply capacity is limited.

【Effect of the invention〕

As explained above, according to the present invention, the scheduling constraints are expressed as a two-dimensional step function (pattern) from the perspective of resource supply and consumption, thereby converting the mission scheduling problem into a supply pattern and a consumption pattern. In addition to understanding it as a puffing problem that involves backing, we also use the heuristic knowledge that humans have regarding backing problems in the scheduling method to greatly reduce the number of searches and find a quasi-optimal scheduling solution within a practical amount of time. It is expected that this will greatly reduce the burden on JEM's onboard crew.

[Brief explanation of the drawing]

FIG. 1 is a diagram explaining the principle of the present invention. FIG. 2 is a system configuration diagram of the present invention. FIG. 3 is an explanatory diagram of a resource frame. Figure 4 is an explanatory diagram of the Mitsushiran frame. FIG. 5 is a block diagram of the blackboard of the present invention. FIG. 6 is a knowledge source transition diagram of the present invention. FIG. 7 is an explanatory diagram of possible time slot calculation. FIG. 8 is an explanatory diagram of decomposition into maximum included rectangles. FIG. 9 is an explanatory diagram of scheduling time selection and resource supply amount. FIG. 1O is an explanatory diagram of scheduling evaluation values. FIG. 11 is an explanatory diagram of scheduling constraints. FIG. 12 is a conceptual diagram of the operation of the scheduling system. FIG. 13 is an explanatory diagram of combinatorial explosion. In the figure, 1 is a scheduling order determining means, 2 is a scheduling possible time slot calculating means, 3 is a scheduling time setting means, and 4 is a backtrack execution means.

Claims (1)

[Claims] In a mission scheduling device for determining mission scheduling under constraint conditions including resource supply capacity from a resource source and resource consumption by a mission, scheduling for determining a scheduling order among missions to be scheduled. The scheduling period is determined by comparing the order determining means (1) with constraint conditions including the resource consumption of the scheduled mission determined by the scheduling order determining means (1) and the amount of resources that can be supplied to the scheduled mission. A schedulable time slot calculation means (2) that calculates a schedulable time slot and further calculates the best solution from among the time slots, and a priority order determined by the schedulable time slot calculation means (2). Scheduling time setting means (3) selects and sets the time with the highest priority at that time from among a plurality of scheduling candidate times as the schedule solution for the mission, and the scheduling possible time slot calculation means (2) performs scheduling. All processing by the backtrack execution means (4) and the scheduling time setting means (3) for backtracking the scheduled mission in order to retry the scheduling of the previously scheduled mission when a possible time slot cannot be determined. and an evaluation value calculation means (5) for calculating an evaluation value of the optimality of the schedule solution obtained as a result of successfully executing the mission. The scheduling order is determined in descending order of resource consumption and consumption time length, and in decentralized scheduling in descending order of resource consumption and maximum consumption value. The setting means (3) determines scheduling times according to the time order of the discretely determined time series in the case of centralized scheduling, and according to the order of the largest remaining resource supply amount among those times in the case of decentralized scheduling, and A mission scheduling device characterized in that an evaluation value calculation means (5) calculates an evaluation value of optimality for every mission.