A planning method for autonomous avoidance of spacecraft orbital threats with uncertainty

Abstract

The invention belongs to the field of aerospace technology, and specifically relates to a method for autonomously avoiding a spacecraft orbital threat with uncertainty. The method comprises the following steps: step 1: analyzing some uncertainty problems in the autonomously avoiding a spacecraft orbital threat mission planning; step 2: describing the resources, subsystem functions, composition and constraints of the spacecraft when planning the mission; step 3: establishing a model for autonomously avoiding a spacecraft orbital threat mission planning; step 4: based on the mission requirements of autonomously avoiding orbital threats, adopting a threat avoidance process, performing on-board autonomous mission planning, and realizing threat avoidance actions; step 5: adopting LP to solve the uncertain constraint boundaries of the duration of the threat avoidance action and the resource constraints, and checking the time and resource numerical consistency in the state. The invention can complete the autonomously avoiding a spacecraft orbital threat mission planning with uncertain duration and resource requirements.

Description

Technical Field

[0001] The present invention belongs to the field of aerospace technology, and in particular relates to a method for planning an autonomous avoidance mission for spacecraft orbital threats with uncertainty. Background Art

[0002] During the threat avoidance process, the spacecraft needs to store some important data and wait for it to be transmitted to the ground control center after the threat avoidance is completed. In the initial planning stage, the spacecraft needs to allocate an appropriate amount of available storage resources;

[0003] However, at the initial moment of planning, it is not known exactly how much resources need to be allocated to threat avoidance. This can only be determined after the entire threat avoidance plan is established. The resource allocation action should have flexible uncertainty about the resource capacity that can be allocated, so that when generating the rest of the plan under a series of accumulated numerical constraints, the planner can freely determine the resource capacity allocation value;

[0004] Uncertainty also appears in the duration constraints of actions, allowing the duration of actions to change; the duration of the change of trajectory cannot be determined at the beginning of planning, and needs to be determined specifically during the avoidance process by the change of trajectory path and maximum thrust constraints given by the decision;

[0005] To sum up, the uncertain duration and resource requirements in threat avoidance greatly increase the difficulty of completing the planning of orbital threat autonomous avoidance missions. In severe cases, it may even lead to the inability to complete the orbital threat autonomous avoidance mission. Summary of the invention

[0006] The purpose of the present invention is to provide a method for planning an autonomous orbital threat avoidance mission for a spacecraft with uncertainty, which can complete the autonomous orbital threat avoidance mission planning for a spacecraft with uncertain duration and resource requirements.

[0007] The technical solution adopted by the present invention is as follows:

[0008] The autonomous avoidance mission planning method for spacecraft orbital threats with uncertainty includes the following steps:

[0009] Step 1: Analyze some uncertainties in the planning of autonomous spacecraft orbital threat avoidance missions;

[0010] Step 2: When planning the mission, describe the spacecraft's resources, subsystem functions, composition, and constraints that need to be met;

[0011] Step 3: Establish a mission planning model for autonomous avoidance of spacecraft orbital threats;

[0012] Step 4: Based on the mission requirement of autonomously avoiding orbital threats, adopt the threat avoidance process to carry out autonomous mission planning on board and implement threat avoidance actions;

[0013] Step 5: Use LP to solve the uncertain constraint bounds of the threat avoidance action duration and resource constraints, and check the consistency of the time and resource values ​​in the state;

[0014] Step 6: Obtain the estimated value of the uncertainty within the constraint interval through controlled planning management;

[0015] Step 7: Planning uses a heuristic algorithm based on a time-relaxed planning graph to guide the spacecraft system to search towards the goal in the planning space. The time-relaxed planning graph associates a timestamp with each action and fact layer.

[0016] Furthermore, the step 3 comprises the following steps:

[0017] Step 301: Determine state variables in the spacecraft;

[0018] Step 302: define a task planning problem with uncertainty;

[0019] Step 303: Create a persistent action tuple;

[0020] Step 304: Define a spacecraft orbit threat autonomous avoidance mission planning solution with uncertainties.

[0021] Furthermore, the state variable of step 301 is the state S of the spacecraft, which is a tuple:

[0022] S=<F,V,B,Q,P,C,T,D>

[0023] Among them: F is a fact that holds in the state, represented as a Boolean variable; V is a resource variable in the spacecraft; B maps the numerical value of the resource variable to an upper and lower limit, and the upper and lower limits of the variable value defined in V are equal to this value; Q is a set of actions that have been started but not yet completed; P is the set of steps in the plan to reach state S; C is the set of time constraints and resource constraints accumulated about the steps in P; T is the upper and lower limits of the time when the state must end; D records the upper and lower limits of all uncertain quantities associated with the actions in P in S.

[0024] Furthermore, the task planning problem with uncertainty in step 302 is represented as a tuple:

[0025] Pro=<S,A,I,G,M>

[0026] Among them: A is the duration action; the resulting action is called a continuous action, I is the initial state of the spacecraft under orbital threat; G is to achieve the target state in order to avoid orbital threats; M is the metric objective function, which is used to select objective functions such as the minimum total planning time or resource variable consumption in planning.

[0027] Furthermore, the continuous action tuple in step 303 is a∈A and is specifically as follows:

[0028]

[0029] Among them, dur represents the duration constraint of the continuous action tuple a, constraining the time length between the start and end of the continuous action tuple a, each of which can be expressed as <g(v,d a ,dur),comp,c>, where c∈R, g(v,d a ,dur) is applied to v, d a Any function of and an uncertainty; comp∈{≥,>,≤,<} is a set of comparison operators; cont is a set of bounded uncertainty constraints on the continuous action tuple a, in the form of is applied to d a Any function of ;

[0030] pre and pre ⊥ They represent the start and end conditions of the continuous action tuple a, respectively, where each condition consists of a proposition pre that must be true and a resource value pre n Condition composition; the form of a numeric condition is as follows: n = <f(v,d a ),comp,c>, where f(v,d a ) is applied to v and d a Any function of ;

[0031] represents the propositional and numerical invariant conditions of a persistent action tuple a that must hold between the start and end of the action; eff and eff ⊥ Represents the start and end effects of a continuous action tuple a; assuming eff xx represents each given set of effects, where x∈{,⊥}; each eff xx include: represents the proposition to be added to the state, represents the proposition to be removed from the state, Represents a numerical effect acting on a state variable.

[0032] Furthermore, the autonomous spacecraft orbit threat avoidance mission with uncertainty in step 304 is planned as planning π, which is an action sequence with a timestamp and a duration of π, as shown below:

[0033]

[0034] Where: n is the planned length, ts i ∈R is the timestamp, i.e., application action a i time, Δt i ∈R is an action a i Duration of is a i Lower and upper limits of uncertainty Rational number matrix, that is The plan is to calculate the time points (t = ts i ) and (t=ts i +Δt i ) Apply action a i The start and end of , generating a sequence of up to 2n states (each state is represented by s(t)).

[0035] Furthermore, the step 5 comprises the following steps:

[0036] Step 501: Use LP to check the temporal and numerical consistency of the spacecraft state. The variables that capture the evolution of resource numerical variables along the planned trajectory are defined as follows:

[0037] At each step i in the planning, construct a variable vector pair V i and V i '; each v i ∈V i Record the value of the state variable v just before step i; similarly, each v i '∈V i 'Record the value of state variable v immediately after step i;

[0038] Any discrete effect on a variable v at step i will change its value from v i Update to v i '; in step i and step i+1 Any subsequent influence on v will cause its value to change from v i 'Continuously update to v i+1 ; Assume Δv now Indicates time step i and step now The cumulative rate of successive changes in v between now Given the value of v between steps, it is calculated as follows:

[0039] v now =Δv now ×(step now -step i )

[0040] Among them, the duration of the action is a free variable, while the rate of change of the resource value remains constant;

[0041] Step 502: Reasoning about uncertainties, which are variables that appear only as part of the action structure in the plan and not as part of the state; when a new action is applied, variables can be added to the constraint set in the state, representing not only the value of the state variable, but also the newly introduced uncertainty;

[0042] In LP, the linear constraint structure of uncertain quantities is as follows:

[0043] Let a be the index of the action that occurs in step s, with m a uncertainties, each of which has an index of j, where j = {0,…,m a -1}; Assume v is the vector of state variables, w v is the vector of constant coefficients of the LP state variable v, d a A set of uncertain LP variables in the continuous action tuple a is defined, and each variable can be expressed as d(j) (a,s) , w d Yes a constant coefficient vector, c is a constant, f(v,d a )and is applied to r resource numeric variables (v) and m a The uncertainty (d a ) is a linear function; the function can be expressed as:

[0044] f(v,d a )=w v ·v+w d ·d a +c

[0045]

[0046] In addition to the time and value constraints that may appear in step 501, the following constraints are inserted into LP during the state process:

[0047] Any numeric precondition given in the following form:

[0048] <f(v,d a ),comp,c>(very bounded control parameter constraints)

[0049] (often bounded controller parameter constraints)

[0050] Any numerical effect of the form:

[0051] <v,op,f(v,d a )>

[0052]

[0053] The main difference between the constraint formulations encoded in the LP and the planning problem given in step 302 is that the uncertainties are defined as global variables in the LP using unique identifiers; in contrast, in the planning problem definition, they are defined as local action variables; this plays a key role when propagating constraints with such local parameters in the state process;

[0054] After expanding the constraint set C as part of the state progression, C is tested for feasibility; if not feasible, then the state corresponds to an empty set of temporal states, so the state needs to be pruned and the search is restored to the previous state; the set C is solved for minimum and maximum bounds T, B, and D.

[0055] Furthermore, the controlled planning management is solved by Algorithm 1, which uses C FINAL , using it as the initial state, and returning at the end a list of execution value decisions for each flexible uncertainty quantity, it propagates all temporal numerical constraints through the model states until it is in the model state E, where each uncertainty quantity is assigned a fixed value; the algorithm then reports the list of uncertainty quantity assignments as the planning solution;

[0056] Among them, C FINAL Is the cumulative state S FINAL The final set of time numeric constraints until

[0057] The technical effects achieved by the present invention are:

[0058] The method for planning an autonomous orbital threat avoidance mission for a spacecraft with uncertainty of the present invention can complete the planning of an autonomous orbital threat avoidance mission for a spacecraft with uncertain duration and resource requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a diagram of the autonomous avoidance architecture of spacecraft orbital threats in the present invention;

[0060] Figure 2 This is a diagram of the planning results of the spacecraft of the present invention for coping with orbital threats. DETAILED DESCRIPTION

[0061] In order to make the purpose and advantages of the present invention more clearly understood, the present invention is specifically described below in conjunction with embodiments. It should be understood that the following text is only used to describe one or several specific embodiments of the present invention, and does not strictly limit the scope of protection of the specific claims of the present invention.

[0062] like Figure 1-2 As shown, the autonomous avoidance mission planning method for spacecraft orbit threats with uncertainty includes the following steps:

[0063] Step 1: Analyze some uncertainties in the planning of autonomous spacecraft orbital threat avoidance missions;

[0064] The main motivation for this patent stems from those problems where a numerical attribute needs to be determined at the end of the planning process, rather than deciding from a set of options during the search process.

[0065] During the threat avoidance process, the spacecraft needs to store some important data, waiting to be transmitted to the ground control center after the threat avoidance is completed. At the initial stage of planning, the spacecraft needs to allocate an appropriate amount of available storage resources. However, at the initial moment of planning, it is not exactly known how many resources need to be allocated to threat avoidance. This can only be determined after the entire threat avoidance plan is established. The resource allocation action should have flexible uncertainty about the resource capacity that can be allocated, so that when the rest of the plan is generated under a series of cumulative numerical constraints, the planner can freely determine the resource capacity allocation value. Therefore, the storage space you want to allocate depends on which actions are performed subsequently. Allocating capacity values ​​in advance may result in a poor plan that is not the optimal resource capacity allocated. Therefore, determining the storage space to be allocated should be done at a later stage of plan construction (possibly after the entire action sequence has been constructed).

[0066] Uncertainty also occurs in the duration constraints of actions, which allow the duration of actions to vary. For example, camera capture of a target involves the action of turning to a certain angle at a constrained angular velocity to aim at the target. Neither the angular velocity nor the steering angle of the action can violate the maximum limit constraint. The steering angle is related to the target position of the camera, and there is uncertainty in the steering time of each camera. The attitude change action requires the control of the torque gyro, and the constraint needs to ensure that the combined power consumption of the motor does not exceed the available power output. This constraint is a constant condition that ensures that it holds throughout the execution of the action. The maximum and minimum angular rates of the attitude change maneuver are limited to a range. The duration of the attitude change action is determined by the maneuver angle and the limit angular rate given by the decision, which cannot be determined at the initial stage of planning. When the constraints on duration and power consumption are met, the action changes the attitude position of the spacecraft by the displacement rate at the end of execution. The spacecraft thrust has similar properties to the attitude change action. The duration of the trajectory change cannot be determined at the beginning of planning. It needs to be specifically determined by the trajectory change path given by the decision and the maximum thrust constraint during the avoidance process.

[0067] In summary, there are multiple action-specific flexible digital resource variables in threat avoidance, and some actions have different planning characteristics in the timing settings (such as the required concurrency, the flexible duration of each action). This patent attempts to solve the above uncertainty problem while solving the autonomous avoidance of orbital threats.

[0068] Step 2: Spacecraft have complex operational constraints and multiple concurrent subsystems. Its on-orbit operation needs to consider factors such as structure, capabilities, equipment status, and mission requirements. When planning the mission, it is necessary to describe the spacecraft's resources, subsystem functions, composition, and constraints that need to be met.

[0069] Among them, the constraints that the spacecraft needs to meet when conducting mission planning include resource constraints, causal constraints, and time constraints.

[0070] This embodiment comprehensively considers the need for spacecraft to avoid orbital threats, and the specific subsystems selected are shown in the following table:

[0071] Table 1 Subsystem names and number of states involved

[0072] Subsystem Name Number of state variables Number of states camera 3 18 radar 2 6 2D servo turntable 2 10 Attitude and orbit determination 1 5 Posture Control 1 3 Track Control 1 5 Propulsion system 1 5 Intelligent computing unit 1 26

[0073] Step 3: Establish a mission planning model for autonomous avoidance of spacecraft orbital threats;

[0074] Step 3 includes the following steps:

[0075] Step 301: Determine state variables in the spacecraft;

[0076] The state S of a spacecraft is a tuple:

[0077] S=<F,V,B,Q,P,C,T,D>

[0078] Among them: F is a fact that holds in the state, expressed as a Boolean variable; V is a resource variable in the spacecraft; B maps the numerical value of the resource variable to an upper limit and a lower limit, and the upper and lower limits of the variable value defined in V are equal to this value; Q is a set of actions that have been started but not yet completed; P is the set of steps to reach state S in the plan; C is the set of time constraints and resource constraints accumulated about the steps in P; T is the upper and lower limits of the time when the state must end; D records the upper and lower limits of all uncertain quantities associated with the actions in P in S;

[0079] The size of the resource variable V in the spacecraft system is defined as r, which can be expressed as V = {v1,…,v r V={v1,…,v r} represents a vector of rational numbers in the state (the value of each resource variable);

[0080] Step 302: define a task planning problem with uncertainty;

[0081] The problem of autonomously avoiding spacecraft orbital threats with uncertainty is represented as a tuple:

[0082] Pro=<S,A,I,G,M>

[0083] Among them: A is the duration action; the resulting action is called a continuous action, I is the initial state of the spacecraft under orbital threat; G is the target state to achieve the avoidance of orbital threat; M is the metric objective function, which is used to select the objective function such as the minimum total planning time or resource variable consumption in planning;

[0084] d a is a finite set of uncertainties (size m a ), whose upper and lower limits are recorded in D; the i-th uncertainty of the continuous action tuple a is given by Indicates that each action has several uncertain quantities, including duration and resource requirement values. If an action has only one uncertain quantity, the size of its uncertain parameter set is 1.

[0085] Step 303: Create a continuous action a∈A tuple, as follows:

[0086]

[0087] Where: dur represents the duration constraint of the continuous action tuple a, constraining the time length between the start and end of the continuous action tuple a, each of which can be expressed as <g(v,d a,dur),comp,c>, where c∈R, g(v,d a ,dur) is applied to v, d a Any function of and an uncertainty; comp∈{≥,>,≤,<} is a set of comparison operators; cont is a set of bounded uncertainty constraints on the continuous action tuple a, in the form of is applied to d a Any function of ;

[0088] pre and pre ⊥ They represent the start and end conditions of the continuous action tuple a, respectively, where each condition consists of a proposition that must be true (pre) and a resource value (pre n ) condition; the form of the digital condition is as follows: n = <f(v,d a ),comp,c>, where f(v,d a ) is applied to v and d a Any function of ; represents the propositional and numerical invariant conditions of a persistent action tuple a that must hold between the start and end of the action; eff and eff ⊥ Represents the start and end effects of a continuous action tuple a; assuming eff xx represents each given set of effects, where x∈{,⊥}; each eff xx include: represents the proposition to be added to the state, represents the proposition to be removed from the state, Represents the numerical effects acting on state variables;

[0089] Duration parameters can appear in the numerical effects of actions Assume that exp e Represents an arithmetic expression on V, op represents an assignment operator {+=, -+:=, *+, / =}; the numerical effect becomes a tuple where v i is a state variable in V; each continuous action creates two time points: the start and end of the action.

[0090] Often bounded uncertainty constraints (in the form of ) refers to a numerical precondition that can limit the bounds of an uncertainty by a constant; on the other hand, a non-constant bounded uncertainty constraint of the form <f(v,d a ),comp,c> also refers to a numerical precondition, which can limit the bounds of uncertainty using state variables, which means that the values ​​in the state and the uncertainty are interdependent.

[0091] Step 304: defining a spacecraft orbit threat autonomous avoidance mission planning solution with uncertainties;

[0092] A plan π is a time-stamped sequence of actions with duration π, as follows:

[0093]

[0094] Where: n is the planned length, ts i ∈R is the timestamp (application action a i time), Δt i ∈R is an action a i Duration of is a i Lower and upper limits of uncertainty Rational number matrix, that is The plan is to calculate the time points (t = ts i ) and (t=ts i +Δt i ) Apply action a i The start and end of the program generate a sequence of at most 2n states (each state is represented by s(t)); in addition, the following conditions must be met in the planning solution: the duration constraint dur of each action in π is satisfied; the constant bounded uncertainty constraint cont of each action in π is satisfied; pre(a i ) and pre ⊥ (a i ) respectively remain in state s(ts i ) and s(ts i +Δt i ), so that holds in every state s(t), where ts i <t<ts i +Δt i , so that The state trajectory produced by applying continuous actions can satisfy the final state of the goal.

[0095] Step 4: Based on the mission requirement of autonomously avoiding orbital threats, taking the Earth observation satellite as an example, Figure 1 The threat avoidance process shown in the figure performs autonomous mission planning on board, thereby eliminating the involvement of ground personnel and achieving threat avoidance actions;

[0096] Specifically, onboard autonomous mission planning includes the following steps:

[0097] When an Earth observation satellite encounters a potential collision or harassment threat during its orbital operation, it first uses four detection methods: global, infrared, laser, and microwave. It combines sensors according to the space environment conditions such as direct light, backlight, and earth shadow to search and capture the threat target, and perform initial distance and angle measurement on the threat target. It also realizes threat identification in a complex space environment through information complementarity between different sensor devices.

[0098] Secondly, by fusing the sensor information, we can obtain the precise speed, distance, azimuth and other information of the threat target; on this basis, we can obtain the abnormal behavior characteristics of the threat target (the abnormal behavior characteristics include the speed and behavior semantics of the target), the target orbit parameters and the collision probability; the visible light camera can achieve accurate imaging at close range, and after obtaining the distance information of the target, we can determine whether the visible light camera is needed for accurate imaging; the morphological characteristics of the threat target are extracted from the multi-layer image information collected by the visible light camera, including identifying the threat target as debris or enemy spacecraft and the payload carried by the target; then, we combine the processed target information and the spacecraft's own attitude and orbit parameters for fusion reasoning to obtain a quantitative evaluation of the threat category and threat level of the target;

[0099] Next, based on the above threat target information and the information after threat level reasoning, the specific avoidance behavior that the spacecraft should take is reasoned and decided, among which the avoidance behavior includes orbital maneuver, attitude maneuver, and normal operation. Combined with the threat target behavior information, the future action behavior of the threat target is estimated. Then, the optimal trajectory to achieve threat avoidance is solved by combining the threat behavior prediction and avoidance behavior.

[0100] Finally, the predetermined optimal avoidance trajectory is executed by the on-board actuator, causing the spacecraft to change its attitude and orbit to avoid the threat. In addition, the spacecraft's own parameters are fed back to the threat level reasoning and action sequence planning in real time to build a closed-loop system for autonomous avoidance of spacecraft orbital threats.

[0101] It is important to point out how to coordinate observation tasks with threat avoidance. Spacecraft orbital threats are a very important but occasional avoidance task. When there is no orbital threat, the spacecraft mainly performs earth observation tasks. In the process of threat avoidance, in order to minimize the impact on the execution of observation tasks, a two-stage strategy planning is proposed.

[0102] like Figure 1 As shown, the entire threat avoidance mission is divided into two stages, with the avoidance behavior decision as the midpoint, the forward stage is the first stage, and the backward stage is the second stage. In the first stage of planning, the attitude and orbital state of the spacecraft will not change, which will not affect the execution of the observation mission. Therefore, the observation mission continues in the first stage of planning, and in the avoidance behavior decision, the subsequent actions to achieve threat avoidance can be obtained;

[0103] If the current orbital threat does not cause harm to the spacecraft, there is no need to avoid it, and the entire plan is terminated, and the second phase of the plan will not be implemented;

[0104] If the evasive action decision requires certain actions of the spacecraft to avoid orbital threats, the observation mission needs to be interrupted immediately and the second stage of planning needs to be carried out because the second stage of evasive action will change the attitude and orbit of the spacecraft, thereby affecting the execution of the observation mission.

[0105] In the two-stage planning strategy, the first-stage planning and observation tasks are executed in parallel, and the interruption of the observation task is determined based on whether the second-stage planning is needed. Therefore, this strategy can effectively reduce the impact of threat avoidance on the observation task and improve the efficiency of task execution.

[0106] Step 5: Use LP to solve the uncertain constraint bounds of the threat avoidance action duration and resource constraints, and check the consistency of the time and resource values ​​in the state;

[0107] Specifically, the spacecraft state represents a set of temporal states, where time and resource numerical state variables interact with each other. When a new continuous action tuple a starts, the state process adds newly created variables corresponding to the uncertainties in a to D, whose upper and lower limits are derived from the preconditions of a (or -∞ or +∞ when there are no boundaries in a). Corresponding constraints are added to C, representing the preconditions and postconditions requirements for these uncertainties.

[0108] During planning, the search chooses whether to start by adding a new action to extend the state, or to complete the action being executed (which ends in Q); in both cases, constraints can be added to C to ensure that the preconditions and invariants are met. Based on C, by solving the upper and lower bounds of each variable, the bounds of B, T, and D can be found; in this way, branching on infinite choices of duration values ​​is avoided, and constraints on possible values ​​of duration are managed, ensuring the feasibility of constraints in each state process.

[0109] Specifically, step 5 includes the following steps:

[0110] Step 501: Use LP to check the temporal and numerical consistency of the spacecraft state. The variables that capture the evolution of resource numerical variables along the planned trajectory are defined as follows:

[0111] At each step i in the planning, construct a variable vector pair V i and V i '; each v i ∈V i Record the value of the state variable v just before step i; similarly, each v i '∈Vi 'Record the value of the state variable v immediately after step i. For example, the state variable v can have a discrete instantaneous value change at step i, in which case the constraint v i '=v i +c is added to LP to record the change, where v i In step i a constant value c is added.

[0112] Any discrete effect on a variable v at step i will change its value from v i Update to v i '; in step i and step i+1 Any subsequent influence on v will cause its value to change from v i 'Continuously update to v i+1 ; Assume Δv now Indicates time step i and step now The cumulative rate of successive changes in v between now Given the value of v between steps, it is calculated as follows:

[0113] v now =Δv now ×(step now -step i )

[0114] Note that the duration of the action is a free variable, while the rate of change of the resource value remains constant.

[0115] Step 502: In addition to duration, uncertainty also needs to be reasoned about in the plan. Like duration, uncertainty is a variable that appears only as part of the action structure in the plan, not as part of the state. This means that when a new action is applied, a variable can be added to the constraint set in the state, representing not only the value of the state variable, but also the newly introduced uncertainty.

[0116] The main characteristics of uncertainty in actions are that each uncertainty is a local variable, whose scope is limited to the action in which it is defined, and is implemented in the plan through global variables; each uncertainty introduces constraints only when the action is applied; subsequent constraints can affect the value of the uncertainty by imposing restrictions on variables that also appear in the constraints related to the uncertainty;

[0117] In LP, we now introduce the formal structure of linear constraints on uncertain quantities in planning, as follows: Let a be the index of the action that appears in step s, with m a uncertainties, each of which has an index of j, where j = {0,…,m a-1}; Assume v is the vector of state variables, w v is the vector of constant coefficients of the LP state variable v, d a A set of uncertain LP variables in the continuous action tuple a is defined, and each variable can be expressed as d(j) (a,s) , w d Yes a constant coefficient vector, c is a constant, f(v,d a )and is applied to r resource numeric variables (v) and m a The uncertainty (d a ) is a linear function; the function can be expressed as:

[0118] f(v,d a )=w v ·v+w d ·d a +c

[0119]

[0120] In addition to the time and value constraints that may appear in step 501, the following constraints are inserted into LP during the state process:

[0121] Any numeric precondition given in the following form:

[0122] <f(v,d a ),comp,c>(very bounded control parameter constraints)

[0123] (often bounded controller parameter constraints)

[0124] Any numerical effect of the form:

[0125] <v,op,f(v,d a )>

[0126]

[0127] The main difference between the constraint formulations encoded in the LP and the planning problem (given in step 302) is that the uncertainties are defined as global variables in the LP using unique identifiers (e.g., unique actions, steps, and uncertainty indices); in contrast, in the planning problem definition, they are defined as local action variables; this plays a key role when propagating constraints with such local parameters in the state process;

[0128] After expanding the constraint set C as part of the state progression, C is tested for feasibility; if not feasible, then the state corresponds to an empty set of temporal states, so the state needs to be pruned and the search is restored to the previous state; the set C is solved for minimum and maximum bounds T, B, and D.

[0129] Step 6: Obtain the estimated value of the uncertainty within the constraint interval through controlled planning management;

[0130] The duration of an action may not be fixed, but may be determined by the value of a resource variable, or it may be constrained to a range of values. Similarly, the uncertainty defined in an action is usually not fixed, but is constrained within an interval and may be subject to multiple constraints. State variables affected by continuity or duration are constrained to the range [v min ,v max ]; the planning process is free to choose variable values ​​within its range, these choices may not be independent, and each choice must be propagated through the constraint set C to determine how it affects the bounds of other variables;

[0131] The presence of uncertainty can lead to complex branching choices in the search space, as in the case of duration parameters; when planning branches on an infinite set of states, it is possible to avoid exploring every state by leaving the choice to the LP constraint space;

[0132] Although the solver can help plan many possible states in the pruned state (in other words, it helps reduce the size of the state), the intervals of the uncertainties are usually not tight enough to easily make decisions about their values. During the planning process, estimates of the parameters are retained until a partially well-founded plan is reached (a planning solution is found), and a series of constraints are accumulated in the process; each uncertainty has an interval of feasible values; otherwise, the plan reached is numerically infeasible; at this point, there are two options to continue with the estimation; the first option is to define a metric objective function to optimize a linear function so that the uncertainties are optimized in order (or possibly in the desired order of importance), and the optimality is subject to the defined metric objective function; the second option is to leave the decision to the spacecraft control system during execution, which is called controlled plan management.

[0133] The bounds of uncertain quantities can depend on each other (i.e., the value decisions made by the control system may change the feasible bounds of other parameters). This problem makes the assignment process slightly more complicated and requires constant modification of the feasible bounds of these parameters. In this step, this problem is solved by iteratively optimizing each parameter subject to temporal numerical constraints that are propagated through the state trajectory that produces the flexible plan (or controlled plan π). This process can be considered a post-processing step and is in principle independent of the plan execution.

[0134] Let S FINAL is the target state reached at the end of the forward state space search, C FINAL Is the cumulative state S FINAL At the end of planning, a controlled plan π is returned, consisting of a time-stamped sequence of activities and flexible intervals of their uncertainties, if any; choosing a value within any of these intervals may affect the following properties of the controlled plan:

[0135] 1. Timestamps and durations of actions. This happens if the duration parameter of an action is linearly dependent on an uncertainty, whose value is specified most recently. As a result, the order of actions may change.

[0136] 2. The interval boundaries of other uncertain quantities.

[0137] Where: C FINAL It includes the timestamp of the action and its uncertainty decision variables, as well as the time order constraints between actions and the duration constraints of the actions in π. Therefore, it is mandatory to give C FINAL Any numerical decision of can easily update these properties of timestamps, durations, and action sequences in the plan π, including bounds on other uncertain quantities.

[0138] The choice made for each parameter must be propagated through the constraint space (from C FINAL start) so that the uncertainty bounds (and other properties of the plan) can be iteratively updated. This process is actually a bound search, where the state propagation decisions are made by the spacecraft control system; this post-processing approach is called controlled plan management.

[0139] Iteratively solve C using a mathematical solver (e.g. LP or MIP) FINAL (plus value assignment constraints) until a boundary is reached (until all uncertain quantities are assigned a fixed value);

[0140] Controlled planning management can be viewed as a search problem, where nodes represent mathematical models (linear or mixed integer models) consisting of temporal numerical constraints (expressed as E), and edges represent state transitions based on execution decisions for uncertain quantities (expressed as r). Each node E is called a model state, forming a mathematical model to be solved. The initial state of the search is equivalent to C. FINAL ; Although the process is modeled as a search problem, it does not employ any state space search method; instead, the branching scheme is subjective to the spacecraft control system and is based on the sequence of decisions made by the control system; if an infeasible state is reached, this indicates that the control system has the right to backtrack in the search space to any state model that is recorded as feasible.

[0141] In principle, controlled planning management can be implemented using Algorithm 1. This algorithm uses C FINAL , using it as the initial state, and finally returning a list of execution value decisions for each flexible uncertainty quantity (denoted as R(E)), which propagates all temporal numerical constraints through the model state (i.e., each cumulative allocation decision denoted as r and the initial constraint set C FINAL ) until it is in model state E, where each uncertainty is assigned a fixed value (i.e., D(E) = 0); the algorithm then reports a list of uncertainty assignments as planning solution. Controlled plan management builds a theoretical foundation for managing partially stranded plans before or during plan execution.

[0142] Note that the search graph is only expanded over model states that are considered to have feasible solutions. The search is based entirely on the choice of model states drawn by the control system (from List E). The termination criterion of the algorithm is that it will terminate only when there are no more flexible uncertainties in the model states, which is entirely dependent on the efficiency of the control system in the decision-making process. In general, the control system is the only mechanism that makes the choice of state propagation (and termination) over the feasible region modeled by the algorithm.

[0143]

[0144]

[0145] Multiple flexible uncertainties can be defined in the same action, but the application of controlled planning management is no different (i.e. they are still assigned one by one), and these parameter domains can still be dependent on each other.

[0146] Finally, formalize the new plan achieved after this process. Modify the definition of the controlled plan so that it reports the specified value of each uncertainty quantity for each action in units of π. Call the updated plan the controlled plan with execution decision (π exec ), as follows:

[0147]

[0148] in It indicates an action i A vector of values ​​specifying the uncertain quantities.

[0149] Step 7: Planning uses a heuristic algorithm based on the Time Relaxation Planning Graph (TRPG) to guide the spacecraft system to search towards the goal in the planning space. The Time Relaxation Planning Graph (TRPG) associates a timestamp with each action and fact layer.

[0150] If a persistent action tuple a has an uncertain effect on a resource variable v, then the uncertainty is relaxed to whichever bound on the corresponding control variable D has the largest effect (increasing the upper bound of v's achievable value or decreasing the lower bound of v's achievable value). If the bounds on the control variables depend on the values ​​of the state variables, the heuristic constructs the LP at the fact level where the conditions occur, using only the time-independent numerical constraints on the actions to compute the bounds on the heuristic before extracting the relaxed plan.

[0151] Through the above steps, the final planning result is output, and the spacecraft autonomous orbit threat avoidance mission planning with uncertain duration and resource requirements is completed. In order to prove the effectiveness and feasibility of the above steps, this technical solution discloses a simulation experiment, which is as follows:

[0152] Simulation settings:

[0153] In the space environment, there are three lighting conditions, including front light, back light, and ground shadow. When the spacecraft encounters an orbital threat, different combinations of four sensors are selected to perceive the threat target according to the illumination. Table 2 shows the sensors used in the three types of illumination. The actions of each subsystem in the spacecraft are performed to autonomously avoid orbital threats. Table 3 shows the functions of each subsystem.

[0154] Table 2 Data acquisition sensors under three lighting conditions

[0155]

[0156]

[0157] Table 3 Overview of spacecraft orbit threat avoidance subsystems

[0158]

[0159] All steps in the simulation are implemented in Linux operating system using C++ programming language, with computer memory limited to 8GB and frequency of 3.60GHz.

[0160] Simulation results analysis:

[0161] Figure 2 The task planning solution for autonomously avoiding orbital threats in the shadow of the Earth is shown. The global camera is always turned on when performing observation tasks, and the threat target is found at 2 seconds. Since the spacecraft is currently in the shadow of the Earth, three other sensor devices are used to collect threat data. When the distance requirement is met, the visible light camera is also used for data collection. After the threat data is collected, data fusion and software action decision-making are performed in the intelligent computing unit to analyze the threat target. Finally, the propulsion system is used to push the spacecraft to change its orbit to achieve threat avoidance.

[0162] from Figure 2 It can be seen that the spacecraft can autonomously plan a series of onboard action sequences from threat target detection to collision avoidance. Through the proposed autonomous mission planning strategy, the spacecraft can plan a series of threat avoidance actions. In the absence of ground measurement and control, the spacecraft can autonomously respond to orbital threats.

[0163] The simulation results show that the spacecraft can perform autonomous onboard collision avoidance when it is threatened by an orbit. The entire planning result derives the action sequence between the spacecraft subsystems and software modules, and coordinates a series of actions of the spacecraft in response to threats. The effectiveness and advantages of the strategy have been verified.

[0164] The above is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be considered as the protection scope of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention shall be implemented according to the conventional means in the art unless otherwise specified and limited.

Claims

1. A method for autonomously avoiding spacecraft orbital threats with uncertainty, characterized by: The following steps are involved: Step 1: Analyze the uncertainty issues in the planning of autonomous spacecraft orbital threat avoidance missions; Step 2: When planning the mission, describe the spacecraft's resources, subsystem functions, composition, and constraints that need to be met; Step 3: Establish a mission planning model for autonomous avoidance of spacecraft orbital threats; Step 4: Based on the mission requirement of autonomously avoiding orbital threats, when observing the earth, the threat avoidance process is adopted to carry out autonomous mission planning on board and implement threat avoidance actions; Step 5: Use the mathematical solver LP to solve the uncertain constraint bounds of the threat avoidance action duration and resource constraints, and check the time and resource numerical consistency in the state; Step 6: Obtain the estimated value of the uncertainty within the constraint interval through controlled planning management; Step 7: Use a heuristic algorithm based on a time-relaxed planning graph to guide the spacecraft system to search towards the goal in the planning space. The time-relaxed planning graph associates a timestamp with each action and fact layer. The step 3 comprises the following steps: Step 301: Determine state variables in the spacecraft; Step 302: define a task planning problem with uncertainty; Step 303: Create a persistent action tuple; Step 304: defining a spacecraft orbit threat autonomous avoidance mission planning solution with uncertainties; The state variable of step 301 is the state of the spacecraft S， is a tuple: S=<F,V,B,Q,P,C,T,D> in: F is a fact that holds true in the state, represented as a Boolean variable; V is the resource variable in the spacecraft; B Map the value of the resource variable to upper and lower limits, V The upper and lower limits of the variable values ​​defined in are equal to this value; Q It is a set of actions that have been started but not yet completed; P It is the planned arrival state S A collection of steps; C yes P The set of time constraints and resource constraints on the accumulation of steps; T are the upper and lower limits of the time by which the state must end; D will with P The upper and lower limits of all uncertainties associated with the actions in S middle; The task planning problem with uncertainty in step 302 is represented as a tuple: Pro=<S,A,I,G,M> in: A is a duration action; the resulting action is called a continuous action, I It is the initial state of the spacecraft under orbital threat; G It is to achieve the target state in order to avoid orbital threats; M It is a metric objective function, which is used to select the objective function with the minimum total planning time or resource variable consumption in planning; The step 5 comprises the following steps: Step 501: Use LP to check the temporal and numerical consistency of the spacecraft state. The variables that capture the evolution of resource numerical variables along the planned trajectory are defined as follows: Each step in the planning i , construct variable vector pairs and sequence; each Record in step i The value of the vector v of the previous resource numeric variable; similarly, each In step i Immediately afterwards, the values ​​of the vector v of resource numerical variables are recorded; V i Indicates in step i The value of the previous resource vector, V i ' indicates that in step i The resource vector value after that; In step i Any discrete effect on a vector v of resource numeric variables changes its value from Updated to ;exist and Between v Any continuous impact on Continuously updated to Assumptions Indicates time and between v The cumulative rate of continuous change on the variable Give the steps between v The value is calculated as follows: The duration of the action is a free variable, while the rate of change of the resource value remains constant. step i represents the time from the initial planning time 0 to step i, step now Represents the time from the initial planning time 0 to the current step; Step 502: Reasoning about uncertainties, which are variables that appear only as part of the action structure in the plan and not as part of the state; when a new action is applied, variables can be added to the constraint set in the state, representing not only the value of the state variable, but also the newly introduced uncertainty; In LP, the linear constraint structure of uncertain quantities is as follows: set up a index is the index of the action that appears in step s, with uncertain quantities, and the index of each uncertain quantity is j ,in Assumptions v is a vector of resource numeric variables, is a vector of LP resource numerical variables v The vector of constant coefficients of A set of uncertain LP variables in the continuous action tuple a is defined, and each uncertain LP variable can be expressed as , yes The constant coefficient vector of c is a constant, and Is applied to r A vector of resource numeric variables v and The linear function of the uncertain LP variables can be expressed as: In addition to the time and value constraints that may appear in step 501, the following constraints are inserted into LP during the state process: Any numeric precondition given in the following form: Any numerical effect of the form: is a set of comparison operators; In the constraint formulation of LP, uncertainties are defined as global variables in LP using unique identifiers; in contrast, in the planning problem definition, they are defined as local action variables; Constraint Set C After expansion as part of the state process, C Perform feasibility test; if it is not feasible, then the state corresponds to an empty set of time states, so the state needs to be pruned and the search is restored to the previous state; the constraint set C is for minimum and maximum boundaries T , B and D Solved.

2. The method for autonomously avoiding a spacecraft orbital threat with uncertainty according to claim 1, characterized in that: The constraints that the spacecraft needs to meet when performing mission planning in step 2 include resource constraints, causal constraints and time constraints.

3. The method for autonomously avoiding a spacecraft orbital threat with uncertainty according to claim 1, characterized in that: The persistent action tuple in step 303 is The details are as follows: in, eff n Indicates the numerical resource impact of the action, eff x expresses the propositional influence of an action, pre n x Indicates the resource value conditions required for the action to start or end. Represents a continuous action tuple a Duration constraints, constraining continuous action tuples a The length of time between the start and end of each In the form of , c is a constant, R represents the set of real numbers, Is applied to v , Any function of ; is a set of bounded uncertainty constraints on the continuous action tuple a, in the form of ; and Represent the start and end conditions of the continuous action tuple a, where each condition consists of a proposition that must be true pre and resource values Condition composition; resource value The conditions have the following form: ; Represents propositional and numerical invariant conditions for a persistent action tuple a that must hold between the start and end of the action; and Represents the start and end effects of the continuous action tuple a; assuming Denotes each given set of effects, where x∈{tb,te}, x Represents a symbol set, which has two elements, tb and te, tb represents the time when the action starts, and te represents the time when the action ends; each include: represents the proposition to be added to the state, represents the proposition to be removed from the state, Represents a numerical effect acting on a state variable.

4. The method for autonomously avoiding a spacecraft orbital threat with uncertainty according to claim 3 is characterized in that: The autonomous avoidance mission planning of the spacecraft orbit threat with uncertainty in step 304 is a planning , is a sequence of actions with timestamps and duration , as shown below: in: n To plan the length, is the timestamp, i.e. the action is applied time, Is an application action Duration of yes The lower and upper limits of the uncertainty in the matrix are rational numbers, that is, ; The plan is implemented by and Apply Action The beginning and end of the n A sequence of states, each of which is represented by express.

Images