CN104573856A - Spacecraft resource constraint processing method based on time topological sorting - Google Patents

Abstract

The invention relates to a spacecraft resource constraint processing method based on time topology sorting, and belongs to the technical field of autonomous control of spacecraft. This method first topologically sorts the actions in the planning results according to the execution time of the actions; then, according to the change of the number of resources in the action, the actions in the resource constraint network are layered, and each resource mutation is processed separately to increase the number of computing resources. Efficiency of the process; particularly applicable to resource management in autonomous planning of spacecraft missions in deep space exploration. This method uses the resource constraint network to describe the resource information in the plan. By analyzing the changes of vertices and edges in the network, the change of the number of resources when the spacecraft performs actions over time is obtained. The augmented path method of the maximum flow problem is combined with the preflow advance The advantage of the method is that the process of calculating the resource quantity is optimized by topologically sorting resource mutations according to the execution time; it can effectively deal with the resource constraints in the planning results and significantly improve the calculation efficiency.

Description

A Spacecraft Resource Constraint Processing Method Based on Time Topological Sorting

technical field

The invention relates to a spacecraft resource constraint processing method based on time topology sorting, and belongs to the technical field of autonomous control of spacecraft.

Background technique

Spacecraft mission autonomous planning is a key technology to realize spacecraft autonomous technology. When a spacecraft is performing a space mission, the spaceborne autonomous management system will automatically generate a sequence of mission actions after the current time based on the perceived state of the space environment, the state of the spacecraft itself, and the mission objectives that need to be performed, which can realize the task without human intervention. long-term autonomous operation. Autonomous mission planning of spacecraft can overcome the problems caused by large communication time delay, complex resource constraints, and dynamic changes in the operating environment in deep space exploration, and improve the autonomy and reliability of spacecraft long-term operation. Spacecraft's onboard resources (propellant, electric energy, memory, etc.) are very limited. When planning spacecraft actions, not only the amount of resources used, but also the constraint relationship between different resources must be considered. Multiple subsystems of a spacecraft can generate multiple actions that are performed simultaneously, and parallel actions lead to more complex resource changes. Designing effective resource constraint handling methods in planning is a key technology for spacecraft mission autonomous planning.

Among the resource constraint processing methods of spacecraft mission planning that have been developed, the prior technology [1] (joint scheduling of space-terrestrial measurement and control resources based on ant colony -90.), studied the resource scheduling problem in aerospace measurement and control, established a model for the constrained optimization problem, and used the method of combining ant colony optimization algorithm and simulated annealing algorithm to find the optimal solution. However, this method is researched on certain measurement and control events, which cannot meet the needs of dealing with dynamic uncertain events in deep space exploration.

Prior technology [2] (see Muscettola N. Computing the envelope for stepwise-constant resource allocations [J]. Berlin, Germany: Principles and Practice of Constraint Programming, 2002.), analyzed the impact of action sets on the number of resources over time, using network The flow model describes the resource constraints in planning, and obtains the amount of resources that change with time by calculating the maximum flow value of the network. However, the general network flow algorithm is used to solve the resource, which does not combine the structural characteristics of the resource-constrained network.

Contents of the invention

Aiming at the problem that the current resource constraint processing method does not utilize the structural characteristics of the resource constraint network and cannot meet the dynamic response requirements of the spacecraft, the present invention proposes a spacecraft resource constraint processing method based on time topological sorting, which is used in spacecraft Calculate onboard resource usage in autonomous mission planning.

This method first topologically sorts the actions in the planning results according to the execution time of the actions; then, according to the change of the number of resources in the action, the actions in the resource constraint network are layered, and each resource mutation is processed separately to increase the number of computing resources. Efficiency of the process; particularly applicable to resource management in autonomous planning of spacecraft missions in deep space exploration.

The spacecraft resource constraint processing method based on time topological sorting specifically includes the following steps:

Step 1: Spacecraft autonomous mission planning generates action sequences, multiple actions during task execution and their execution time, time constraints between actions, and resource constraints required for each action execution. Wherein, the action sequence is expressed as A={a ₁ ,a ₂ ,...,a _i ,...,a _n }. a _i is the action included in the planning result. The time constraint between actions is the time interval between every two actions. The resource constraint of an action is the change of the number of resources at the beginning and end of each action of the spacecraft.

When an action a _i needs to use a resource r, the quantity q _r of the resource r is only at the beginning of the action a _i or end time Change, each change of resource quantity is described as a resource mutation δ. According to the action sequence, a resource mutation set is generated for each required on-board resource.

Step 2: According to the resource mutation set described in step 1, using the execution time t of the resource mutation δ as the criterion, the resource mutations in each resource mutation set are topologically sorted and labeled in chronological order to generate resources with a hierarchical structure Constraint network G _R . G _R represents all resource constraints in an action sequence. The specific method is:

Step 2.1, according to the execution time of the resource mutation, record the sequence number of the point corresponding to the earliest resource mutation as 1, and then increase the sequence numbers of the subsequent resource mutation points sequentially according to the time sequence. While marking the serial number, mark all resource mutation points as production mutation point v _Pi and consumption mutation point v _Cj according to the change of resource quantity.

The production mutation point indicates that the resource mutation production resource corresponding to this point increases the number of system resources.

The consumption mutation point indicates that the resource mutation corresponding to this point consumes resources and reduces the amount of system resources.

In step 2.2, perform hierarchical processing on all resource mutation points to obtain G _R . G _R is divided into four layers: source point σ, production mutation layer V _P , consumption mutation layer V _C , and sink point τ. The edges between each point of production mutation layer _VP and source point σ are production edges E _P , the edges between each point of production mutation layer _VP and consumption mutation layer V _C are internal edges E _I , consumption mutation The edges between the vertices of the layer V _C and the sink point τ are consumption edges E _C .

Step 2.3, add a flow value for each edge after hierarchical processing. The initial value of the flow is set to 0, and it must not exceed the capacity of the edge when used.

The capacity is the maximum value of traffic on each edge. The capacity E _P of the production side is set to the value of the production mutation increase resource corresponding to the production mutation point v _Pi connected to the production side, and the capacity E _C of the consumption side is set to the consumption corresponding to the consumption mutation point v _Cj connected to the consumption side Mutations reduce the value of resources,

Step 3: Carry out flow promotion from the source point σ to the production mutation layer _VP . According to the order of labels from small to large, select each production mutation point v _Pi in turn, and carry out saturation advancement from source point σ to production mutation point v _Pi on the production edge E _Pi = (σ, v _Pi ). When performing saturated propulsion, set the flow value of E _Pi to the capacity of E _Pi . For v _Pi , when the resource mutations with sequence numbers greater than i are all production mutations, end step three and execute step four.

Step 4: Carry out traffic advancement from the production mutation layer _VP to the consumption mutation layer _VC , and then carry out traffic promotion from the consumption mutation layer _VC to the sink point τ. The specific method is:

Step 4.1, according to the label sequence from small to large, sequentially select the production mutation point v _Pi ;

Step 4.2, for a production mutation point v _Pi , select all consumption mutation points v _Cj (j>i) whose labels are greater than v _Pi according to the order of labels from small to large, and on the internal edge E _I =(v _Pi ,v _Cj ) Carry out saturation advancement from the production mutation point v _Pi to the consumption mutation point v _Cj ;

Carry out saturation advancement from the production mutation point v _Pi to the consumption mutation point v _Cj specifically: use the flow rate of the production edge E _Pi = (σ, v _Pi ) of the selected production mutation point v _Pi to subtract the current selected internal edge The flow of E _I is obtained to obtain the available flow of E _Pi ; then the flow value of the internal edge E _I = (v _Pi , v _Cj ) is set as the available flow of E _Pi , and the flow value of the production edge E _Pi is subtracted from E _Pi available traffic.

Step 4.3, for the consumption mutation point v _Cj that has completed saturation advancement in step 4.2, perform saturation advancement on the consumption edge E _C =(v _Cj ,τ);

The saturation propulsion on the consumption edge E _C = (v _Cj , τ) is specifically: subtract the current flow of _EC from the capacity of the consumption edge E _Cj = (v _Cj , τ), to obtain the available capacity of _EC flow; if the available flow of E _C is greater than the flow of the interior edge E _I = (v _Pi , v _Cj ) selected in step 4.2, then add the flow of E _C to the flow of E _I , and set the flow of E _I to is 0; if the available flow of E _C is less than or equal to the flow of the internal edge E _ij = (v _Pi , v _Cj ) selected in step 4.2, then set the flow of E _C to the capacity of E _C and set the flow of E _I flow minus the available flow of the _EC .

Step 4.4, select the next consumption mutation point whose serial number is greater than v _Cj , and carry out saturation advancement on the consumption side; if all consumption mutation points with a serial number greater than v _Cj have been processed, select the next production mutation point with a serial number greater than v _Pi , return to step 4.2, until all the production mutation points complete the saturation advancement to the sink point τ through the consumption mutation layer V _C .

Step 5, the above steps complete the saturated advancement of the source point σ to the sink point τ through the middle two layers. According to the sum of all flows of E _C =(v _C ,τ) connected to the sink τ in _GR , the maximum flow value F( _GR ) of the network is obtained. The change value of the resource quantity is calculated according to the following formula:

${\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; C</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; P</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; u</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}^{\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; C</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; \&Delta;</span>}{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}^{\mathrm{<span\; class="notranslate">\; C</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; R</span>}}<span\; class="notranslate">\; )</span>$

Among them, q _r (t) represents the change value of the resource quantity at time t. According to the relationship between the execution time of the resource mutation and the current time t, the resource mutation caused by the action sequence A is divided into three parts: the mutation has occurred to be mutated and unmutated have mutated Including all resource mutations performed at time t, the total change to the resource quantity is to be mutated Including all resource mutations performed at time t, the total change to the resource quantity is Not mutated Including all resource mutations that have not yet been executed at time t, the total change to the resource quantity is The value of is calculated by calculating the sum of all mutated resource change values at time t get, The value of is obtained by calculating the maximum flow value F( _GR ) of the resource-constrained network at time t.

Thus, the change value of the number of spacecraft resources in the planning result is obtained.

Beneficial effect

The spacecraft resource constraint processing method based on time topology sorting given by the present invention has the advantages of simple algorithm and high calculation efficiency, can process complex planning results, and effectively utilizes the time factor in planning, which is convenient for actual execution Perform quick calculations.

This method uses a resource-constrained network to describe the resource information in the plan. By analyzing the changes of vertices and edges in the network, the change of the number of resources when the spacecraft performs actions over time is obtained. Based on the hierarchical characteristics of the resource-constrained network, the maximum flow problem is combined. The advantages of the augmented path method and the pre-flow push method optimize the process of calculating the number of resources by topologically sorting resource mutations according to execution time. This method can effectively deal with resource constraints in planning results, and significantly improves computational efficiency when calculating complex planning results.

Description of drawings

Fig. 1 is the resource constraint network in the background technology;

Fig. 2 is the layered structure of the resource constrained network adopting the method of the present invention in the specific embodiment; wherein (a) is the schematic diagram of the flow direction of the traffic in the resource constrained network between the layers of the network, and (b) is the topological sorting of the traffic according to the time at the apex of the network transfer process between

Fig. 3 is the result of spacecraft resource constraint processing using the method of the present invention in a specific embodiment.

Detailed ways

In order to better illustrate the purpose and advantages of the present invention, the content of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

First analyze the feasibility of the inventive method:

In the existing resource-constrained network as shown in Figure 1, when solving F( _GR ), the vertices and edges of _GR can be classified and processed, and the resource-constrained network can be transformed into a hierarchical structure, as shown in Figure 2. The vertices in the network are divided into four layers horizontally: source point σ, production mutation layer V _P , consumption mutation layer V _C , and sink point τ. Arranged vertically according to execution time. Edges are divided into three layers horizontally: production edge E _P , internal edge E _I , and consumption edge E _C . Figure 2(a) shows the flow of network traffic.

It can be seen from Fig. 2 that the following steps are required to advance the flow from σ to τ: 1) From the source point σ along the production edge E _P to the production point V _P , the advancing flow cannot exceed the capacity of E _P ; 2) From the production point From the point V _P along the internal edge E _I to the consumption point V _C , since the capacity of E _I is +∞, the advancing flow is unlimited; 3) From the consumption point V _C along the consumption edge E _C to the sink point τ, the advancing flow Do not exceed the capacity of the _EC .

In G _R in Figure 1, the direction of the production side E _P is σ→ _VP , and the direction of the consumption side E _C is V _C →τ, so there is only one case when the first and third steps of the above process are executed. When executing the second step of the process, since the direction of the internal edge E _I can be from _VP to _VC , or from _VC to _VP , it can also be within _VP or _VC . If the flow is only advanced along the E _I from V _P to V _C , such as the path Then the generated augmenting path is the shortest, and the number of passing edges is 3. If traffic is pushed from within the vertex layer or from E _I between _VC and _VP , such as the path Then the length of the augmenting path will be greater than 3.

The following analyzes the influence of the path from _VC to _VP on the flow advancing process when advancing flow in the vertex layer. Considering the time order, for any two resource mutations δ _i and δ _j in _GR , the execution time , that is, E _I =(v _i ,v _j ). Therefore, there is a direct connection edge (v _P , v _C ) between the production mutation v _P and all consumption mutations v _C whose execution time is after v _P. Since the capacities of E _{and I} sides are +∞, when the augmented path includes the edges in the vertex layer Or from V _C to the edge (v _C , v _P ) of VP, _it can be transformed into an equivalent path that only includes the edge (v _P , v _C ). Therefore, when selecting E _I , we only need to consider the edge (v _P , v _C ) from V _P to V _C according to the time relationship between resource mutations.

When the flow is promoted from V _P to V _C , the determination of the promotion flow and the order of selecting E _I will affect the efficiency of the operation. When pushing traffic from v _P to v _C , the traffic stored on v _P is _{equal to the capacity of EP connected to v P. Since} the capacity of E _I side is +∞, all the traffic stored on v _P will be pushed to v _C superior. However, the flow that can be pushed by v _C is equal to the capacity of the _EC connected to v _C. When the flow that v _P pushes to v _C exceeds the flow that can be pushed by v _{C, the excess flow on v C needs} to be returned, adding additional operations. When advancing flow, use the capacity of E _P as a reference to reduce unnecessary operations.

Consider the order of choosing E _I below. If the resource mutation execution time in _GR is different, that is , G _R is a directed acyclic graph. Topologically sort all vertices in _GR according to the execution time to generate a linear sequence of resource mutations. Selecting v _P and v _C boost flow according to the order of generating linear sequences can ensure saturation boost on all production mutations and reduce the number of boosts as much as possible. The connection relationship between v _P and v _C is only related to the time relationship of resource mutation in δ _A , so the linear sequence generated by topological sorting reflects the time order of resource mutation execution.

If there are resource mutations with the same execution time in _GR , that is When , δ _i and δ _j have the same time relationship with other resource mutations, sorting them according to the size of |Δq _r,δ | can effectively reduce the number of pushes.

A specific example is given below to illustrate the specific implementation steps of the method of the present invention. In this example, consider a spacecraft with 8 subsystems, the action sets of the planning results are distributed on the timelines of each subsystem, and the planning results to be processed are generated according to the following parameters: the execution time T∈ of the action sets in the planning results [0,200], the total number of actions included is n=50, the resource used is memory, and the number of resources needs to satisfy q _r ∈ [0,2000]. At the initial time t=0, the number of memory resources q _r (0)=2000. The connectivity probability cp=0.1 of the resource-constrained network generated by the planning result, this parameter represents the parallelism of actions in the spacecraft system. The specific process is as follows:

Step 1: plan the generated action sequence, multiple actions in the task execution process and their execution time, the time constraints between each action, and the resource constraints required for each action execution. The action sequence is expressed as A={a ₁ ,a ₂ ,...,a _i ,...,a _n }. a _i is the action contained in the planning result, such as attitude rotation, data transfer, etc. The time constraint between actions is the time interval between every two actions. The resource constraint of an action is the change of the number of resources at the beginning and end of each action of the spacecraft.

The action a in the planning result can be described as a set of action-related variables The time constraint of action a _i is represented by the execution time t, and its value range is t∈[t ^s , t ^e ], where t ^s and t ^e are the start time and end time of action a execution. The resource constraints for action a are defined by the resource variable express, The number of represented resources is denoted as q _r , and the maximum value it can take is , the amount of resources must satisfy An action can be affected by multiple resource constraints at the same time.

When an action a needs to use a resource r, set the number of resources q _r only at the beginning of any action or end time Each change in resource quantity can be described as a resource mutation δ. According to the action sequence generated by the autonomous mission planning of the spacecraft, a resource mutation set can be generated for each required on-board resource. When the time is t, the change value of resource quantity is $q_r\left(t\right)=q_{r,{\mathrm{\&delta;}}_A^C}\left(t\right)+q_{r,{\mathrm{\&delta;}}_A^P}\left(t\right)+q_{r,{\mathrm{\&delta;}}_A^u}\left(t\right).$ In the above formula, the changed value without mutation can be ignored. The change value of the completed mutation That is, the upper limit of the range of execution time is the sum of δ of t to the resource change. Change value to be mutated It is more complicated and needs to be calculated through the following steps.

In step two, according to the resource mutation set generated in step one, a resource constraint network _GR representing all resource constraints in the action sequence can be generated. Each resource mutation δ in δ _A is represented as a vertex v of G _R . In order to represent the production and consumption of resources, it is also necessary to add source point σ and sink point τ respectively representing resource production and consumption in _GR . The time relationship between resource mutations is expressed as an internal edge E _I =(v _p , v _c ), the direction of the edge is from the resource mutation executed later to the resource mutation executed earlier, and the capacity of the edge is infinite. According to the increase or decrease of the resource amount according to δ, the vertexes representing the increase in resource value are collectively called the production mutation point v _P , and the vertices that decrease the resource value are collectively called the consumption mutation point v _C . The production mutation point v _P and the source point σ are connected by the production edge E _P = (σ, v _p ), the direction of the edge is from σ to v _P , and the capacity of the edge is the resource change value Δq _{r, δ} ; the consumption mutation point It is connected with the sink τ by the consumption edge E _C =(v _c ,τ), the direction of the edge is from v _C to τ, and the capacity of the edge is the value of resource change Δq _r,δ .

Taking the execution time t of the resource mutation δ as the criterion, the resource mutations in each resource mutation set are topologically sorted and labeled in chronological order to generate a resource constraint network with a hierarchical structure. When sorting, according to the execution time of the resource mutation, the sequence number of the point corresponding to the earliest resource mutation is recorded as 1, and then the sequence numbers of the subsequent vertices are sequentially increased according to the time sequence. When labeling the serial number, the production mutation point and the consumption mutation point are marked together, and the final form of each vertex is v _Pi or v _Cj .

In order to facilitate subsequent processing, it is also necessary to perform hierarchical processing on _GR . The vertices of _GR are divided into four layers: source point σ, production mutation V _P , consumption mutation V _C , and sink point τ. The edge between the production mutation _VP and the source point σ is the production edge E _P , the edge between the production mutation _VP and the consumption mutation _VC is the internal edge E _I , and the edge between the consumption mutation _VC and the sink point τ All edges are consumption edges E _C .

For subsequent use, a flow value is added to each edge after the layering process. The flow values are all set to 0 and must not exceed the capacity of the edge when used.

Step 3: Carry out flow promotion from the source point σ to the production mutation layer _VP . According to the label sequence from small to large, select each production mutation point v _Pi in turn, and carry out saturation advancement from source point σ to production mutation point v _Pi on the production edge e _Pi = (σ, v _Pi ).

Set e _Pi 's flow value to e _Pi 's capacity when doing saturated propulsion. For v _Pi , when the resource mutations with sequence numbers greater than i are production mutations, end step three.

Step 4: Carry out traffic advancement from the production mutation layer _VP to the consumption mutation layer _VC , and then carry out traffic promotion from the consumption mutation layer _VC to the sink point τ.

Step 4.1, among the untreated production mutation points in the production mutation layer V _P , sort according to the labels of the production mutation points, and select the production mutation point v _Pi with the smallest label;

Step 4.2, for the selected production mutation point v _Pi , among all the consumption mutation points whose labels are greater than v _Pi , select the consumption mutation point v Cj (j>i) according to the label order from small to large, and select the consumption mutation point v _Cj (j>i) from the production mutation point v _Pi points to the internal edge E _ij = (v _Pi , v _Cj ) that consumes the mutation point v _Cj , and performs saturation advancement;

The available flow of E Pi can be obtained by subtracting the flow of the currently selected internal edge E _ij from the flow of the production edge E _Pi = ₍ σ,v _Pi ) pointing to the selected production mutation point v _Pi . Then the flow value of the internal edge E _ij =(v _Pi ,v _Cj ) is set as the available flow of E _Pi , and the flow value of the production edge E _Pi is subtracted from the available flow of E _Pi .

Step 4.3, starting from the consumption mutation point v _Cj selected in step 4.2, carry out saturation advancement on the consumption edge E _Cj =(v _Cj ,τ). The available flow of E Cj can be obtained by subtracting the current flow of E _Cj from the capacity of consumption edge E _Cj = ₍ v _Cj ,τ). If the available flow of E _Cj is greater than the flow of the internal edge E _ij = (v _Pi ,v _Cj ) selected in step 4.2, add the flow of E _Cj to the flow of E _ij and set the flow of E _ij to 0 ; If the available flow of E _Cj is less than or equal to the flow of the internal edge E _ij = (v _Pi , v _Cj ) selected in step 4.2, then set the flow of E _Cj to the capacity of E _Cj and reduce the flow of E _ij Available traffic to E _Cj .

In step 4.4, after the saturation advancement is completed on v _Cj , return to step 4.2 and select the next consumption mutation point whose label is greater than v _Pi ; if the consumption mutation point with the largest sequence number has been reached, return to step 4.1 and select the next Production mutation points with labels greater than v _Pi . Until all the production mutation points complete the saturated advancement to the sink point τ through the consumption mutation layer _VC .

Step 6: After the above steps are completed, the source point σ is saturately advanced to the sink point τ through the middle two layers of vertices. According to the sum of all E _C =(v _C ,τ) flows connected to the sink τ in _GR , the maximum flow value F( _GR ) of the network can be obtained. According to the following formula

${\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; C</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; P</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; u</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}^{\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; C</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; \&Delta;</span>}{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}^{\mathrm{<span\; class="notranslate">\; C</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; R</span>}}<span\; class="notranslate">\; )</span>$

The change value of the number of spacecraft resources in the planning result can be obtained.

The results of spacecraft resource constraint processing in this example are shown in Figure 3. It can be seen from the figure that when n=50 and cp=0.1, when the planning result is executed, the memory resource margin is always kept within the resource value range, and the resource constraint of the planning result is satisfied.

Claims (3)

1., based on a spacecraft resource constraint disposal route for time topological sorting, it is characterized in that: comprise the steps:

Step one, the autonomous mission planning of spacecraft generates time-constrain, each resource constraint needed for action executing between action sequence, multiple action in tasks carrying process and execution time thereof, each action; Wherein, action sequence is expressed as A={a ₁, a ₂..., a _i..., a _n; a _ifor the action comprised in program results; Time-constrain between action is the time interval that every two actions occur; The resource constraint of action be each action executing of spacecraft start with at the end of change to resource quantity;

As action a _iwhen needing to use a kind of resource r, if the quantity q of resource r _ronly at action a _istart time or finish time change, the change of each resource quantity is described as first resource sudden change δ; According to action sequence, for resource on the star used needed for each, generate a resource catastrophe set respectively;

Step 2, according to the resource catastrophe set described in step one, with the execution time t of resource sudden change δ for criterion, carries out topological sorting to the resource sudden change in each resource catastrophe set and in chronological sequence order label, generates and have the resource constraint network G of hierarchy _r; G _rrepresent all resource constraints in action sequence; Concrete grammar is:

Step 2.1, according to the execution time of resource sudden change, is designated as 1 by the sequence number of point corresponding for the resource performed the earliest sudden change, then increases the sequence number of following resource catastrophe point according to time sequencing successively; While marking serial numbers, all resource catastrophe points are labeled as production catastrophe point v according to the change of resource quantity _piwith consumption catastrophe point v _cj;

Step 2.2, carries out layered shaping to all resource catastrophe points, obtains G _r; G _rbe divided into four layers: source point σ, production sudden change layer V _p, consume sudden change layer V _c, meeting point τ; Produce sudden change layer V _peach point and source point σ between limit be production limit E _p, produce sudden change layer V _pwith consumption sudden change layer V _ceach point between limit be internal edges E _i, consume sudden change layer V _cpoint and meeting point τ between limit be and consume limit E _c;

Step 2.3, increases a flow value for every bar limit after layered shaping; Flow initial value is set to 0, and must not exceed the capacity on this limit in use;

Described capacity is the maximal value of flow on every bar limit; The capacity E on production limit _pbe set to the production catastrophe point v be connected with production limit _picorresponding production sudden change increases the numerical value of resource, consumes the capacity E on limit _cbe set to and the consumption catastrophe point v consuming limit and be connected _cjcorresponding consumption sudden change reduces the numerical value of resource,

Step 3, from source point σ to production sudden change layer V _pcarry out flow propelling; According to label order from small to large, select each production catastrophe point v successively _pi, and at production limit E _pi=(σ, v _pi) on from source point σ to production catastrophe point v _picarry out saturated propelling; When carrying out saturated propelling, by E _piflow value be set to E _picapacity; For v _pi, when the resource sudden change that sequence number is greater than i is production sudden change, end step three, performs step 4;

Step 4, from production sudden change layer V _pto consumption sudden change layer V _ccarry out flow propelling, then from consumption sudden change layer V _cflow propelling is carried out to meeting point τ; Concrete grammar is:

Step 4.1, according to label order from small to large, selects production catastrophe point v successively _pi;

Step 4.2, for a production catastrophe point v _pi, according to label order from small to large, select all labels to be greater than v _piconsumption catastrophe point v _cj(j > i), at internal edges E _i=(v _pi, v _cj) on from production catastrophe point v _pito consumption catastrophe point v _cjcarry out saturated propelling;

Step 4.3, to the consumption catastrophe point v completing saturated propelling in step 4.2 _cj, at consumption limit E _c=(v _cj, τ) on carry out saturated propelling;

Step 4.4, selects sequence number to be greater than v _cjthe next one consume catastrophe point, consumption limit on carry out saturated propelling; If oneself is greater than v at treated complete all sequence numbers _cjconsumption catastrophe point, then select sequence number be greater than v _pinext production catastrophe point, return step 4.2, until all production catastrophe points all complete through consuming sudden change layer V _cto the saturated propelling of meeting point τ;

Step 5, above-mentioned steps completes source point σ through the middle two-layer saturated propelling to meeting point τ; According to G _rin to be connected with meeting point τ all E _c=(v _c, τ) flow sum, obtain the maximum flow valuve F (G of network _r); Changes values according to following formula computational resource quantity:

$q_r\left(t\right)=q_{r,{\mathrm{\&delta;}}_A^C}\left(t\right)+q_{r,{\mathrm{\&delta;}}_A^P}\left(t\right)+q_{r,{\mathrm{\&delta;}}_A^U}\left(t\right)=\underset{{\mathrm{\&delta;}}^C\&Element;{\mathrm{\&delta;}}_A^C}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}{q}_{r,{\mathrm{\&delta;}}^C}\left(t\right)+F\left(G_R\right)$

Wherein, q _rt () represents the changes values in t resource quantity; According to the execution time of resource sudden change and the relation of current time t, the resource caused by action sequence A sudden change is divided into three parts: oneself undergos mutation wait to undergo mutation do not undergo mutation oneself undergos mutation comprise all resource sudden changes that oneself performs at moment t, to total changes values of resource quantity be wait to undergo mutation comprise all resource sudden changes performed at moment t, to total changes values of resource quantity be do not undergo mutation comprise all in the still unenforced resource sudden change of moment t, to total changes values of resource quantity be value by calculate moment t time all resource changing value summations that oneself undergos mutation obtain, value by calculating moment t time resource constraint network maximum flow valuve F (G _r) obtain;

Thus obtain the changes values of spacecraft resource quantity in program results.

2. a kind of spacecraft resource constraint disposal route based on time topological sorting according to claim 1, is characterized in that: described in step 4.2 from production catastrophe point v _pito consumption catastrophe point v _cjcarry out saturated propelling to be specially: with selected production catastrophe point v _piproduction limit E _pi=(σ, v _pi) flow deduct the internal edges E with current selected _iflow, obtain E _piutilizable flow; Then by internal edges E _i=(v _pi, v _cj) flow value be set to E _piutilizable flow, and by production limit E _piflow value deduct E _piutilizable flow.

3. a kind of spacecraft resource constraint disposal route based on time topological sorting according to claim 1, is characterized in that: described in step 4.3 consumption limit E _c=(v _cj, τ) on carry out saturated propelling and be specially: with consuming limit E _cj=(v _cj, τ) capacity deduct E _cpresent flow rate, can E be obtained _cutilizable flow; If E _cutilizable flow to be greater than in step 4.2 selected internal edges E _i=(v _pi, v _cj) flow, then by E _cflow add E _iflow, and by E _iflow set be 0; If E _cutilizable flow to be less than or equal in step 4.2 selected internal edges E _ij=(v _pi, v _cj) flow, then by E _cflow set be E _ccapacity, and by E _iflow deduct E _cutilizable flow.