WO2019127947A1

WO2019127947A1 - Intelligent optimization and constraint reasoning-based single-satellite autonomous task planning method

Info

Publication number: WO2019127947A1
Application number: PCT/CN2018/080421
Authority: WO
Inventors: 何敏藩; 邢立宁; 白国庆; 石建迈; 王锐; 谭旭; 文翰; 熊彦; 甘文勇; 陈剑
Original assignee: 佛山科学技术学院; 佛山市有义家科技有限公司
Priority date: 2017-12-26
Filing date: 2018-03-26
Publication date: 2019-07-04
Also published as: CN108333922A; CN108333922B

Abstract

An intelligent optimization and constraint reasoning-based single-satellite autonomous task planning method, which uses a method of combining an intelligent optimization algorithm and a reasoning engine to solve a framework. The intelligent optimization algorithm comprises a task sorting step, a task deconstruction step and an activity chance search step. The sorted tasks are deconstructed into a structured or partially ordered activity map in the task deconstruction step, and imaging subtasks and imaging direction adjustment subtasks are scheduled in the activity chance search step, so as to obtain a quasi-optimal solution. The inference engine performs a conflict check on the quasi-optimal solution on the basis of logical relationships, time and satellite resource constraints in the activity map, optimizes, if there is a conflict, the quasi-optimal solution according to the conflict situation, performs local searching on the optimized quasi-optimal solution, so as to obtain a quasi-optimal solution again, and performs conflict check again until a verification result passes, or the predetermined maximum number of iterations is reached. The invention uses less computation, and produces a better task planning result.

Description

A Single Star Autonomous Task Planning Method Based on Intelligent Optimization and Constraint Reasoning

Technical field

The invention relates to the field of remote sensing satellite technology, in particular to a single-star autonomous task planning method based on intelligent optimization and constraint reasoning.

Background technique

As the level of imaging satellite hardware has increased, its application goals have also become higher requirements. Satellite mission planning has always been a difficult issue in the field of systems engineering because of complex constraints, unpredictable state information, and cumbersome types of requirements.

On-board autonomous mission planning is to rely little on external information injection and control. Based on the perceived state and external environment, intelligent satellites independently plan and schedule the information acquisition activities of satellite resources to develop satellites that meet the mission requirements of satellite applications. Earth observation plan.

The problem of imaging satellite autonomous management is essentially an artificial intelligence planning problem with resource constraints and time constraints, namely planning and scheduling integration. For imaging satellites, if there are 10 original propositions in the domain model, then there will be 2 ¹⁰ states; if there are a large number of actions, each step will be explored from an initial state to the target state. Many possible action branches. Assuming that the satellite has 10 actions, if it is known in advance that the number of actions required to reach the target state from the initial state is 5, then the possible search branches are 10 × 9 × 8 × 7 × 6 = 30240, if you do not know the need beforehand. The number of actions, then the possible search branch will be larger. In the satellite autonomous control problem, the action execution also has time constraints and resource constraints. As shown in Figure 1, a satellite plan has great time flexibility for imaging and backhaul. It is still a time after the adjustment start time. Feasible solution. If you consider these factors when searching, planning problems will be more complicated.

Therefore, there is a need for a new type of simple, computationally intensive autonomous mission planning method to meet the needs of satellite imaging technology.

Summary of the invention

In order to solve the problems in the prior art, the object of the present invention is to provide a single-star autonomous task planning method based on intelligent optimization and constraint reasoning, which modularizes and integrates domain model, time and resource constraint reasoning and problem model. Aerospace domain models such as attitude control model, battery model, solid model and antenna model, constructing an integrated planning and scheduling framework with scalability and versatility; constructing a single-star autonomous mission planning technique based on intelligent optimization and constraint reasoning: intelligent optimization module The task and activity opportunities are selected as the combined variables for local search, and the constraint reasoning module processes and conflicts the logical relationships, time and resource constraints in the task decomposition activity diagram. It uses the heuristic information and user preferences related to the satellite domain to guide constraint reasoning and plan generation, resulting in better mission planning results with less computational complexity.

In order to achieve the above object, the technical solution adopted by the present invention is:

A single-star autonomous task planning method based on intelligent optimization and constraint reasoning, which uses a combination of an intelligent optimization algorithm and an inference engine to solve a framework, the intelligent optimization algorithm including a task sequencing step, a task decomposition step, and an activity opportunity search step. The task sorting step includes sorting the tasks according to the selected sorting intelligence, and the task decomposition step decomposes the sorted tasks into structured or partially ordered activity graphs, and the active set in the activity graph includes only imaging subtasks and imaging Direction adjusting subtasks, the activity opportunity searching step scheduling the imaging subtasks and imaging direction adjustment subtasks to obtain a quasi-optimal solution, the inference engine is based on logical relationships, time, and satellite resource constraints in the activity graph The quasi-optimal solution performs conflict checking. If there is a conflict, the quasi-optimal solution is optimized according to the conflict condition, and the optimized quasi-optimal solution is locally searched to obtain the quasi-optimal solution again, and the conflict is again performed. Check until the verification result passes, or reach the set maximum number of iterations.

Preferably, the logical relationships in the activity diagram include logical requirements and pre-order relationships.

Preferably, the on-board activity is defined as a plurality of types of atomic tasks, including: recording, returning, real-time transmission, day orientation, ground orientation, payload erasure, whole star maneuver, camera boot, and camera shutdown. The imaging subtask includes starting the camera, recording the camera off, and the imaging direction adjustment subtask includes orienting the ground, and performing the conflict check includes calling a preset power model, a full star attitude maneuver model, a digital antenna maneuver model, and a day orientation. The maneuver model, the load code rate calculation model, and the imaging condition evaluation model perform collision detection, wherein each model for verification is based on an atomic task setting.

Preferably, the constraint reasoning of the inference engine comprises logical constraint inference, time inference and resource constraint inference, wherein the logical constraint inference adopts a conditional triggering manner to generate a new activity according to a condition and inserts; the time inference adopts a time constrained network The path consistency check and the constraint propagation technique satisfy the time value reduction and the time constraint; the resource constraint reasoning is based on the time network, the resource time network description problem, the calculation of the resource consumption level distribution, and the distribution according to the distribution Defects, and based on the defect management mechanism, adjust the constraints between activities.

Preferably, the intelligent optimization algorithm is triggered at a time point as follows:

(1) T-driven scheduling time point, the T-driven scheduling time point is to determine a specific scheduling time point lT according to a given time interval T, 0 ≤ l ≤ L, LT ≤ H < (L + 1) T, each time a scheduling time point lT is reached, the task plan for generating the latter scheduling interval [lT, (l+1)T] is calculated, where l is a positive integer, T is a given time interval, and L is the maximum T- The number of times the driver is scheduled, and H is the total scheduling interval.

(2) C ^* - driven rescheduling time point, when the satellite is operating in a given scheduling interval, if at some time t (0 < t < H), the cumulative observation task accumulation C _t on the star exceeds Given a threshold C ^* , then the time point is a C ^* -driven rescheduling time point, where the threshold C ^* is the critical cumulative number of emergency observation tasks,

In addition to the above two scheduling moments, the intelligent optimization algorithm is not triggered at any other point in time.

By adopting the above technical solution, the single-star autonomous task planning method based on intelligent optimization and constraint reasoning of the present invention has the beneficial effects compared with the prior art:

1. The present invention utilizes heuristic information and user preferences related to the satellite domain to guide constraint reasoning and plan generation, and produces better task planning results with less computational complexity.

2. Established a spacecraft knowledge representation method. Compared with the traditional satellite dispatching method, it is easier to continuously update according to the characteristics of future spacecraft and promote technological advancement.

DRAWINGS

Figure 1 is a schematic diagram showing an example of the status of each subsystem in the plan;

2 is a framework diagram of an autonomous task planning and solving according to the present invention;

3 is a schematic diagram of an intelligent optimization decision layer of the present invention;

4 is a schematic diagram of a constraint inference decision layer of the present invention;

FIG. 5 is a diagram showing an example of the first layer decomposition of the imaging task of the present invention; FIG.

Figure 6 is a diagram showing an example of decomposition of the second layer of the observation task of the present invention.

Detailed ways

In order to make the objects, technical solutions and advantages of the present invention more comprehensible, the present invention will be further described in detail below with reference to specific examples. It is to be understood that the description is not intended to limit the scope of the invention. In addition, descriptions of well-known structures and techniques are omitted in the following description in order to avoid unnecessarily obscuring the inventive concept.

The single-star autonomous task planning method based on intelligent optimization and constraint reasoning of the invention adopts a combination of intelligent optimization algorithm and inference engine to solve the framework. The intelligent optimization algorithm includes a task sorting step, a task decomposition step, and an activity opportunity searching step. The task sorting step includes sorting the tasks according to the selected sorting intelligence, and the task decomposition step decomposes the sorted tasks into structured or partial sequences. An activity map, the activity set in the activity map includes only an imaging subtask and an imaging direction adjustment subtask, and the activity opportunity searching step schedules the imaging subtask and the imaging direction adjustment subtask to obtain a quasi-optimal solution The inference engine performs a conflict check on the quasi-optimal solution based on logical relationships, time, and satellite resource constraints in the activity map, and if there is a conflict, optimizes the quasi-optimal solution according to the conflict condition, and optimizes the quasi-optimal solution The optimal solution is searched locally to obtain the quasi-optimal solution again, and the conflict check is performed again until the verification result passes, or the set maximum number of iterations is reached.

Preferably, the constraint reasoning of the inference engine comprises logical constraint inference, time inference and resource constraint inference, wherein the logical constraint inference adopts a conditional triggering manner to generate a new activity according to a condition and inserts; the time inference adopts a time constrained network The path consistency check and the constraint propagation technique satisfy the time value domain reduction and the time constraint; the resource constraint reasoning is based on the time network, the resource time network description problem, the calculation resource consumption level distribution, and the distribution according to the distribution Defects, and based on the defect management mechanism, adjust the constraints between activities.

By adopting the above technical solution, the single-star independent task planning method based on intelligent optimization and constraint reasoning of the present invention has the beneficial effects compared with the prior art:

The invention is based on the intelligent optimization and constraint reasoning single star autonomous task planning method, including the following contents:

(1) Establish a spacecraft knowledge representation method. In the traditional satellite dispatching research, the various operational rules of spacecraft are directly applied to the scheduling process, which is not easy to update according to the characteristics of future spacecraft. It is necessary to study the scalable knowledge representation method according to the characteristics of spacecraft for subsequent modeling. Used with solution tuning.

(2) Autonomous task planning and solving framework. The basic solution framework combining the inference engine and the intelligent optimization algorithm is shown in Figure 2. The intelligent optimization algorithm performs local search for the combined variables with the selection of tasks and activity opportunities. The underlying constraint reasoning method is used to process the logical relationships, time and resource constraints in the activity map of task decomposition and conflict resolution. The constraint check does not make a decision and only provides the current constraint satisfaction or conflict condition.

Intelligent optimization algorithms include task sequencing, task decomposition, and activity opportunity search, as shown in Figure 3. First, the user selects the sorting criteria in different sorting criteria, such as priority, demand, earliest start time, latest end time, free time, and resource contention; the task decomposition process breaks down the task into structured or partial order. Activity collection; there are sorting-based heuristics, knowledge-based intelligent selection, random selection, etc. in the activity opportunity search process. The optimization result is a structured or partially ordered activity set, ie the activity diagram in Figure 3. The event here selects the opportunity to select a time window for each activity.

Constraint reasoning includes logical constraint reasoning, time reasoning, and resource constraint reasoning, as shown in Figure 4. Logical reasoning mainly uses conditional triggering to generate new activities based on conditions and insert them. Time reasoning mainly uses time-constrained and constrained propagation techniques of time-constrained networks to achieve time-domain reduction and time-constrained satisfaction. Resource reasoning is based on the time network. The resource time network describes the problem. Because the activity changes the resource state in a relative way, it is necessary to calculate the distribution of resource consumption level, find the defect according to the distribution, and adjust the activity and activity based on the defect management mechanism. constraint.

(3) Task mode library. Future missions of the space system can be more refined, such as multi-point target imaging, stereo imaging, wide-slice imaging, dynamic scanning imaging and other modes. These imaging modes exhibit the following characteristics: one is to contain unknown, uncertain path selection or action sequences; the other is that there is an unclear and unpredictable connection between the task path and the result; the third is that the complex task may contain a series of Non-independent subtasks, so there may be complex internal connections between subtasks.

The invention defines five actions in the established satellite planning domain model, namely, an atomic task in the planning process; on the basis of the atomic task, two levels of complex tasks are defined, and the first layer is imaging, which refers to a camera imaging task. It consists entirely of atomic tasks. The corresponding decomposition method is d_imaging. The first layer decomposition example of the imaging task is shown in Figure 5.

The second layer is observing, which refers to the satellite observation task, which consists of imaging tasks and other atomic tasks. There are two decomposition methods according to the preconditions, namely d_observing_1 and d_observing_2, and the second layer decomposition example of the observation task is shown in Fig. 6.

(4) Scheduling resource knowledge model. Domain knowledge is the basis of problem modeling and solution. In particular, domain modeling and search guidance require domain knowledge support. Therefore, spacecraft system composition, on-board resources, functions and constraints are analyzed, and domain models are included. The domain knowledge base of formula information and preference knowledge is the basis of the research of the present invention. The present invention focuses on explicit knowledge, and therefore needs to focus on the explicit knowledge modeling of activities, constraints, heuristic information, and preference functions in spacecraft management. For the intelligent satellites studied in this paper, the object is used to represent the satellite subsystems and the target entities. The predicates are used to indicate the status attributes of each subsystem. The initial state defines the state values at the beginning of each subsystem. The status of each subsystem that the task requires is taken, and the state value conversion is implemented by action execution.

(5) Sub-system prediction model. In satellite mission planning, not only the task observation and data transmission schemes need to be planned, but also the satellite motion sequences corresponding to the mission observation and data transmission schemes need to be optimized, including recording, backhaul, real transmission, orientation to the sun, and Various satellite actions such as ground orientation, solid storage erasure, star motion, camera startup, camera shutdown, etc. These actions involve complex working principles and usage constraints of satellite power, load, storage, attitude control, and digital transmission systems. Failure to consider or simplify by assumptions will affect the viability of the satellite mission planning program. In the process of satellite mission planning modeling and solving, it is necessary to call the prediction models of the relevant satellite subsystems, including the electricity model, the star attitude maneuver model, the digital antenna maneuver model, the directional maneuvering model, the load code rate calculation model and Imaging condition evaluation model. These models need to be designed and developed according to the actual characteristics of each subsystem.

The invention modularizes the domain model, the time and resource constraint reasoning and the problem model, integrates the aerospace domain models such as the attitude control model, the battery model, the storage model and the antenna model, and constructs an integrated planning and scheduling with scalability and versatility. Framework; construct a single-star autonomous mission planning technology based on intelligent optimization and constraint reasoning: the intelligent optimization module selects the task and activity opportunity as the combined variable for local search, and the constraint reasoning module assigns the logical relationship, time and resource constraints in the task decomposition activity diagram. Processing and conflict resolution.

The specific embodiments described above are merely exemplary in order to enable those skilled in the art to understand the present invention and are not to be construed as limiting the scope of the invention; any equivalent changes made in accordance with the spirit of the disclosure. Or modifications, are included in the scope of this patent.

Claims

A single-star autonomous task planning method based on intelligent optimization and constraint reasoning, characterized in that a framework is solved by using an intelligent optimization algorithm combined with an inference engine, which includes a task sorting step, a task decomposition step, and an activity opportunity. a search step, the task sorting step includes sorting the tasks according to the selected sorting intelligence, and the task decomposition step decomposing the sorted tasks into a structured or partially ordered activity graph, wherein the active set in the activity graph includes only the imaging sub A task and an imaging direction adjustment subtask, the activity opportunity searching step scheduling the imaging subtask and the imaging direction adjustment subtask to obtain a quasi-optimal solution, the inference engine is based on a logical relationship, time, and satellite in the activity map The resource constraint performs a conflict check on the quasi-optimal solution, and if there is a conflict, optimizes the quasi-optimal solution according to the conflict condition, and performs a local search on the optimized quasi-optimal solution to obtain a quasi-optimal solution again, and Perform the conflict check again until the verification result is passed, or the set maximum is reached. Number.
The single-star autonomous task planning method based on intelligent optimization and constraint reasoning according to claim 1, wherein the logical relationship in the activity graph comprises a logical requirement and a pre-order relationship.
The single-star autonomous task planning method based on intelligent optimization and constraint reasoning according to claim 1, wherein the on-board activity is defined as a plurality of types of atomic tasks, and the atomic tasks include: recording, returning, real Transmission, day orientation, ground orientation, solid storage erasure, whole star maneuver, camera boot and camera shutdown, wherein the imaging subtask includes camera start, recording camera shutdown, imaging direction adjustment subtask including orientation to the ground, conflict The inspection includes calling a preset power model, a full-star attitude maneuver model, a digital antenna maneuver model, a day-to-day maneuvering model, a load code rate calculation model, and an imaging condition evaluation model for collision detection, wherein each is used for verification. The model is based on atomic task settings.
The single-star autonomous task planning method based on intelligent optimization and constraint reasoning according to claim 1, wherein the constraint reasoning of the inference engine comprises logical constraint reasoning, time reasoning and resource constraint reasoning, and the logical constraint The reasoning adopts the conditional triggering method, and generates new activities according to the conditions and inserts; the time reasoning adopts the path consistency checking and the constrained propagation technology of the time constrained network, so that the time domain reduction and the time constraint are satisfied; the resource constraint reasoning is established. On the basis of the time network, the problem is described by the resource time network, the distribution of resource consumption levels is calculated, defects are found according to the distribution, and constraints between activities are adjusted based on the defect management mechanism.
The single-star autonomous task planning method based on intelligent optimization and constraint reasoning according to any one of claims 1 to 4, characterized in that the intelligent optimization algorithm is triggered at the following time point:

(1) T-driven scheduling time point, the T-driven scheduling time point is to determine a specific scheduling time point lT according to a given time interval T, 0 ≤ l ≤ L, LT ≤ H < (L + 1) T, each time a scheduling time point lT is reached, the task plan for generating the latter scheduling interval [lT, (l+1)T] is calculated, where l is a positive integer, T is a given time interval, and L is the maximum T- The number of times the driver is scheduled, and H is the total scheduling interval.

(2) C * - driven rescheduling time point, when the satellite is operating in a given scheduling interval, if at some time t (0 < t < H), the cumulative observation task accumulation C t on the star exceeds Given a threshold C * , then the time point is a C * -driven rescheduling time point, where the threshold C * is the critical cumulative number of emergency observation tasks,

In addition to the above two scheduling moments, the intelligent optimization algorithm is not triggered at any other point in time.