CN116679756A - Parameter intelligent-adjustment space non-cooperative target adhesion obstacle avoidance guidance method - Google Patents

Abstract

The invention discloses a space non-cooperative target adhesion obstacle avoidance guidance method with intelligent parameter adjustment, and belongs to the technical field of spacecraft guidance and control. In the invention, a detector attachment dynamics model is established under an attachment point fixed coordinate system; on the basis of the optimal adhesion guidance law containing obstacle avoidance items, constructing a neural network model for updating guidance parameters online, obtaining the optimal guidance parameters of the detector in different states through an offline optimization method, training the neural network to fit the mapping relation between the optimal guidance parameters and the detector states, and calculating and updating the optimal guidance parameters in real time by taking the position of the detector relative to the obstacle and the speed of the detector as inputs so as to realize autonomous obstacle avoidance adhesion of the detector. The invention meets the calculation performance constraint of the spaceborne computer by constructing the optimal guidance law in the analytic form. According to the invention, guidance parameters can be adjusted on line according to the state and obstacle information of the detector observed in real time, so that the autonomous obstacle avoidance attachment of the celestial body detector is realized, and certain robustness is realized.

Description

Parameter intelligent-adjustment space non-cooperative target adhesion obstacle avoidance guidance method

Technical Field

The invention relates to a spacecraft guidance method, in particular to a space non-cooperative target attachment obstacle avoidance guidance method with intelligent parameter adjustment, and belongs to the technical field of spacecraft guidance and control.

Background

The small celestial body detection is a research hotspot in the current deep space detection field, and research and investigation of the small celestial body detection is beneficial to human exploration of important scientific problems such as the evolution process of a solar system, the monitoring and early warning of the collision of the small celestial body near the earth and the like, and promotes the development of the deep space detection technology. As a typical space non-cooperative target, the surface appearance of a celestial body is complex, barriers are numerous, and the surface barrier distribution is difficult to accurately predict in advance, so that the safety attachment of a detector is a great threat. In addition, limited burnup and on-board computing resources also bring practical engineering constraints to the design of the probe attachment guidance method. Therefore, there is a need for a method for autonomous obstacle avoidance and attachment guidance for a non-cooperative target in development space, and the guidance method is required to have the characteristics of low fuel consumption, compact form and the like.

Related researches are designed for an adhesion obstacle avoidance guidance method of space non-cooperative targets such as a celestial body and the like, and the method mainly comprises an optimal guidance law containing obstacle avoidance items and a track optimization-tracking guidance method. The optimal guidance law containing obstacle avoidance items mostly depends on parameters set offline, and when facing unknown obstacles, the detector can increase fuel consumption due to conservation of parameter setting and can collide with the obstacles due to excitation of parameter setting; the track optimization-tracking guidance method cannot plan the optimal track in real time due to large calculation amount, and is difficult to be applied to actual attachment tasks. Aiming at the problems that the distribution of the surface obstacle of the celestial body is difficult to predict in advance, the burnup and the calculation resources on the satellite are limited, the traditional adhesion obstacle avoidance guidance method needs to be improved, and a novel autonomous intelligent obstacle avoidance guidance method is developed.

Disclosure of Invention

Aiming at the problem that the surface morphology of a small celestial body is difficult to know in advance, starting from the requirements of meeting the requirements of safe and accurate attachment of a detector, reducing fuel consumption and improving the solving efficiency of guidance instructions, the main purpose of the invention is to provide a space non-cooperative target attachment obstacle avoidance guidance method with intelligently adjusted parameters, based on the optimal attachment guidance law containing obstacle avoidance items, a neural network model for online updating guidance parameters is constructed, the optimal guidance parameters of the detector in different states are obtained, the neural network model is trained to construct an intelligent mapping function relation between the optimal guidance parameters and the detector states, and the intelligent mapping function relation is called to online update the guidance parameters, so that the detector can avoid obstacles according to real-time detector state information, obstacle information and attachment point information, and the safe and accurate attachment of the detector is ensured.

The aim of the invention is achieved by the following technical scheme.

The invention discloses a space non-cooperative target adhesion obstacle avoidance guidance method with intelligent parameter adjustment, which comprises the following steps:

step one, under the condition that the attachment points are fixedly connected with a coordinate system, a detector attachment dynamics model is established; an optimal adherence guidance law containing obstacle avoidance items is constructed, and the risk of collision of the detector with the obstacle is reduced by changing the curvature of the adherence track.

The specific implementation method of the first step is as follows:

defining an attachment point fixation coordinate system O-XYZ: the attachment point is taken as an origin O of a coordinate system, the external normal direction of the surface of the small celestial body at the attachment point is taken as a Z axis, the X axis points to the projection point of the initial position of the detector in the tangent plane of the surface of the small celestial body at the attachment point from the attachment point, and the Y axis direction follows the right rule.

The detector attachment dynamics model is established under the attachment point fixed coordinate system and is shown as (1):

wherein r and v respectively represent a position vector and a velocity vector of the detector under the attachment point fixing coordinate system, ζ=g- ω×ω×r-2ω×v represents a sum of gravitational acceleration and inertial acceleration of the detector under the attachment point fixing coordinate system, g is a gravitational acceleration vector borne by the detector under the attachment point fixing coordinate system, ω is a celestial body rotation angular velocity vector under the attachment point fixing coordinate system, and a represents a gravitational acceleration and inertial acceleration of the detector under the attachment point fixing coordinate systemT represents the detector thrust vector, m represents the detector mass, I _sp Denote engine specific impulse, g ₀ A value representing the gravitational acceleration of the earth's sea level.

Taking thrust acceleration and detector position representing burnup as performance indexes, as shown in formula (2):

wherein t is _f Represents the time of flight, Λ= [ Λ ] _x Λ _y Λ _z ] ^T Representing obstacle avoidance coefficients, where Λ _x ,Λ _y ＞0，Λ _z And < 0. When the curvature of the adhesion track has a convex change trend, the performance index function is reduced, so that the risk of collision between the detector and the small celestial body surface obstacle is reduced.

The thrust acceleration expression is obtained as shown in the formula (3):

wherein t is _go Indicating the remaining time of flight.

In order to consider the burnup performance and the obstacle avoidance performance of the detector, the obstacle avoidance coefficient Λ in the formula (3) is modified into a dynamic obstacle avoidance coefficient Λ' shown as the formula (4):

in the method, in the process of the invention,is defined as: the current position of the overdetector is taken as two vertical planes Σ _x Sum sigma _y Is parallel to OXZ plane and OYZ plane respectively, Σ _x Sum sigma _y The planar cut obstacles each form an obstacle profile from the current position of the detector to the edges of the two obstacle profilesTangent line, Σ _x The included angle between the tangent line in the plane and the negative direction of the OX axis isThe counterclockwise rotation direction is positive; sigma and method for producing the same _y The included angle between the tangent line in the plane and the negative direction of the OY axis is +.>The counterclockwise rotation direction is positive; the probe speed is at Σ _x Sum sigma _y The projections in the plane are v _xz And v _yz ，v _xz The included angle between the direction of (2) and the negative direction of the axis OX isThe counterclockwise rotation direction is positive; v _yz The angle between the direction of (2) and the negative direction of the OY axis is +.>The counterclockwise rotation direction is positive; if v _xz 0, then corresponding->Is 0; if v _yz 0, then corresponding->Is 0.

The simultaneous formulas (3) and (4) form an optimal adhesion guidance law containing obstacle avoidance terms, and the risk of collision of the detector with the obstacle is reduced by changing the curvature of the adhesion track according to the optimal adhesion guidance law.

And secondly, taking the burnup and the distance from the detector to the obstacle as performance indexes, and taking the guidance parameters as optimization variables to respectively establish a guidance parameter optimization model with adjustable flight time and fixed flight time. And respectively carrying out off-line optimization under different initial states of the detector to obtain optimal guidance parameters under each initial state, and forming a training data set of the neural network model for updating the guidance parameters.

The specific implementation method of the second step is as follows:

and establishing a guidance parameter optimization model with adjustable flight time. Selecting a performance index shown as a formula (5):

J＝Δm+μΔh _max (5)

wherein Δm represents fuel consumption, μ represents collision penalty weight, and Δh _max Represents the maximum distance of the detector from the surface of the obstacle along the Z-axis, and μ=0 when the detector does not collide with the obstacle.

The boundary constraint condition of the detector state quantity is shown as a formula (6):

wherein r is ₀ And v ₀ Respectively representing a position vector and a speed vector of the initial moment of the detector under the fixed coordinate system of the attachment point, m ₀ Indicating the mass of the probe at the initial moment.

The thrust amplitude constraint is shown in formula (7):

0≤||T||≤T _max (7)

wherein T is _max Representing the maximum thrust amplitude.

The optimization variable is the time of flight t _f And the obstacle avoidance coefficient lambda, wherein the constraint of the optimized variable value range is shown in the formula (8):

combining the formulas (1), (5), (6), (7) and (8) to obtain a guidance parameter optimization model with adjustable flight time:

the relative position vector defining the detector and the obstacle is shown in formula (10):

r _rel ＝[x _rel y _rel z _rel ] ^T (10)

wherein x is _rel And y _rel Representing the distance of the detector along the X-axis and the Y-axis to the edge of the obstacle, z _rel Representing the distance of the detector along the Z-axis to the upper surface of the obstacle. Under the fixed coordinate system of the attachment point, different initial positions r of the detector are selected ₀ Respectively calculating relative position vectors of the detector and the obstacle to form a relative position grid; at each initial position r ₀ Respectively selecting initial speeds v with different sizes and directions ₀ The velocity grid is formed, and the relative position grid and the velocity grid together form a detector state quantity grid. And (3) at each grid point of the state quantity grid, optimizing and solving the optimal guidance parameters corresponding to the state quantity of the grid point by using the parameter optimization model of the formula (9), and recording the state quantity of each grid point and the optimal guidance parameters corresponding to the state quantity of each grid point to form a first training data set.

The optimization variable of the guidance parameter optimization model with fixed flight time is an obstacle avoidance coefficient lambda, and the parameter optimization model is shown in a formula (11):

based on the detector state quantity grids of the first training data set, selecting a series of fixed flight times at grid points of each state quantity grid, wherein the selection mode of the flight times is shown as a formula (12):

where k is the magnification, and a series of fixed times of flight are obtained by varying the value of k. And under different initial states and flight times, obtaining the optimal obstacle avoidance coefficient corresponding to each state quantity-flight time by using the parameter optimization model shown in the formula (11), and forming a second training data set.

And thirdly, based on the training data set generated in the neural network learning step two, establishing an intelligent mapping function relation between the detector state quantity and the guidance parameters. Two neural network models with flight time at an input end and an output end are built, two intelligent mapping function relations with adjustable flight time and fixed flight time are obtained, and guidance parameters are generated through the two intelligent mapping function relations at the initial time and the non-initial time of attachment. And calling an intelligent mapping function relation to update guidance parameters on line by combining with an optimal attachment guidance law, so that the detector can avoid obstacles according to real-time detector state information, obstacle information and attachment point information, and the safe and accurate attachment of the detector is ensured.

The specific implementation method of the third step is as follows:

taking the state quantity in the first training data set generated in the second step as an input quantity, taking the corresponding optimal guidance parameters, namely the optimal flight time and the optimal obstacle avoidance coefficient, as output quantities, and establishing an intelligent mapping function relation of the guidance parameters with adjustable flight time based on neural network learning, wherein an expression of the intelligent mapping function relation is shown as a formula (13):

[t _f ,Λ]＝f ₁ (r _rel ,v) (13)

wherein f ₁ (. Cndot.) is a time-of-flight tunable intelligent mapping function. The intelligent mapping function relation is utilized to calculate guidance parameters on line in the attaching process, so that the problems of large calculated amount, low calculation efficiency and the like caused by an optimization algorithm are avoided.

Taking the state quantity and the flight time in the second training data set generated in the second step as input quantity, taking the corresponding optimal guidance parameter, namely the optimal obstacle avoidance coefficient, as output quantity, and establishing an intelligent mapping function relation of the guidance parameter with fixed flight time based on neural network learning, wherein the intelligent mapping function relation is shown as a formula (14):

Λ＝f ₂ (r _rel ,v,t _f ) (14)

wherein f ₂ (. Cndot.) is an intelligent mapping function with fixed time of flight.

During the probe attachment process, the guidance law versus the remaining time of flight t is known from equation (3) _go Is more sensitive. When (when)Remaining time of flight t _g o uses an intelligent mapping function f ₁ Output value t of (-) () _f At the time of occurrence of the remaining flight time t _g o oscillate, and then cause the thrust output to oscillate, causing instability of the detector. For this purpose, a total time of flight t is calculated at the initial instant _f After which the remaining time of flight t _go By t _go ＝t _f -t, and taking the remaining flight time as an input quantity, and calculating an optimal obstacle avoidance coefficient in real time by an intelligent mapping function with fixed flight time, as shown in a formula (15).

And calling an intelligent mapping function relation to update guidance parameters on line by combining with an optimal attachment guidance law, so that the detector can avoid obstacles according to real-time detector state information, obstacle information and attachment point information, and the safe and accurate attachment of the detector is ensured.

The beneficial effects are that:

1. aiming at the problem that the obstacle information of the surface of a small celestial body is difficult to predict in advance, the intelligent parameter-adjusting spatial non-cooperative target adhesion obstacle avoidance guidance method disclosed by the invention is characterized in that a neural network model for online updating guidance parameters is constructed on the basis of the optimal adhesion guidance law containing obstacle avoidance items, the optimal guidance parameters of a detector in different states are obtained through an offline optimization method, the neural network is trained to fit the mapping relation between the optimal guidance parameters and the states of the detector, the position of the detector relative to the obstacle and the speed of the detector are taken as inputs, and the optimal guidance parameters are calculated and updated in real time, so that the autonomous obstacle avoidance adhesion of the detector is realized. According to the invention, guidance parameters can be adjusted on line according to the state and obstacle information of the detector observed in real time, so that the autonomous obstacle avoidance attachment of the celestial body detector is realized, and certain robustness is realized.

2. Aiming at engineering reality constraint conditions such as limited computing and storage capacity of a small celestial body detector computer, the intelligent parameter-adjusting space non-cooperative target attachment obstacle avoidance guidance method disclosed by the invention is designed based on an analysis form optimal guidance method, so that the problems of complex computation and the like caused by a large-scale optimization algorithm are avoided.

3. According to the space non-cooperative target adhesion obstacle avoidance guidance method with the intelligent parameter adjustment, disclosed by the invention, on the basis of realizing the beneficial effects 1 and 2, the adhesion guidance of the small celestial body detector meeting the thrust amplitude constraint can be realized, meanwhile, the obstacle avoidance performance and the burnup performance of the detector are also considered, and the double-zero adhesion condition required by the small celestial body adhesion is realized.

Drawings

FIG. 1 is a schematic flow chart of a spatial non-cooperative target attachment obstacle avoidance guidance method with intelligent parameter adjustment disclosed by the invention;

FIG. 2 is a plot of a three-dimensional trajectory of a celestial detector attachment;

FIG. 3 is a three-axis velocity profile of a celestial detector;

FIG. 4 is a graph of obstacle avoidance coefficients output by the intelligent mapping function;

FIG. 5 is a plot of the three axis thrust of the celestial detector.

Detailed Description

For a better description of the objects and advantages of the present invention, the following description will be given with reference to the accompanying drawings and examples.

In order to verify the feasibility of the method, taking the task of attaching a detector on the celestial body 433Eros as an example, simulation of a space non-cooperative target attaching obstacle avoidance guidance method with intelligent parameter adjustment is performed. The detector is moved from an initial position r ₀ ＝[800,0,600] ^T The attachment detection task is started from the hover state at (unit: km), with an initial speed of 0. The initial mass m=500 kg of the detector and the maximum thrust T of the engine _max =25n, engine specific impulse I _sp Small celestial body rotation speed ω= 3.3117 ×10 =300 s ^{- 4} rad/s, earth sea level gravity acceleration g ₀ ＝9.80665m/s ² 。

As shown in fig. 1, the method for performing obstacle avoidance guidance by attaching a spatial non-cooperative target with intelligent parameter adjustment disclosed in this embodiment specifically includes the following implementation steps:

step one, under the condition that the attachment points are fixedly connected with a coordinate system, a detector attachment dynamics model is established; an optimal attachment guidance framework containing obstacle avoidance items is designed, and the risk of collision of the detector with the obstacle is reduced by changing the curvature of an attachment track.

The specific implementation method of the first step is as follows:

defining an attachment point fixation coordinate system O-XYZ: the attachment point is taken as an origin O of a coordinate system, the external normal direction of the surface of the small celestial body at the attachment point is taken as a Z axis, the X axis points to the projection point of the initial position of the detector in the tangent plane of the surface of the small celestial body at the attachment point from the attachment point, and the Y axis direction follows the right rule.

The detector attachment dynamics model is established under the attachment point fixed coordinate system and is shown as (1):

wherein r and v respectively represent a position vector and a velocity vector of the detector under the attachment point fixing coordinate system, ζ=g- ω×ω×r-2ω×v represents a sum of gravitational acceleration and inertial acceleration of the detector under the attachment point fixing coordinate system, g represents a gravitational acceleration vector of the detector under the attachment point fixing coordinate system, ω represents a small celestial rotation angular velocity vector under the attachment point fixing coordinate system, a represents a thrust acceleration vector under the attachment point fixing coordinate system, T represents a detector thrust vector, m represents a detector mass, and I _sp ＝300s，g ₀ ＝9.80665m/s ² 。

Taking thrust acceleration and detector position representing burnup as performance indexes, as shown in formula (2):

wherein t is _f Represents the time of flight, Λ= [ Λ ] _x Λ _y Λ _z ] ^T Representing obstacle avoidance coefficients, where Λ _x ,Λ _y ＞0，Λ _z And < 0. When the curvature of the adhesion track has a convex change trend, the performance index function is reduced, so that the risk of collision between the detector and the small celestial body surface obstacle is reduced.

The thrust acceleration expression is obtained as shown in the formula (3):

wherein t is _go Indicating the remaining time of flight.

In order to consider the burnup performance and the obstacle avoidance performance of the detector, the obstacle avoidance coefficient Λ in the formula (3) is modified into a dynamic obstacle avoidance coefficient Λ' shown as the formula (4):

in the method, in the process of the invention,is defined as: the current position of the overdetector is taken as two vertical planes Σ _x Sum sigma _y Is parallel to OXZ plane and OYZ plane respectively, Σ _x Sum sigma _y Plane cutting obstacles each form an obstacle profile, tangential to the edges of the two obstacle profiles from the current position of the probe, Σ _x The included angle between the tangent line in the plane and the negative direction of the OX axis isThe counterclockwise rotation direction is positive; sigma and method for producing the same _y The included angle between the tangent line in the plane and the negative direction of the OY axis is +.>The counterclockwise rotation direction is positive; the probe speed is at Σ _x Sum sigma _y The projections in the plane are v _xz And v _yz ，v _xz The included angle between the direction of (2) and the negative direction of the axis OX isThe counterclockwise rotation direction is positive; v _yz The angle between the direction of (2) and the negative direction of the OY axis is +.>The counterclockwise rotation direction is positive; if v _xz 0, then corresponding->Is 0; if v _yz 0, then corresponding->Is 0;

the simultaneous formulas (3) and (4) form an optimal adhesion guidance law containing obstacle avoidance terms, and the risk of collision of the detector with the obstacle is reduced by changing the curvature of the adhesion track according to the optimal adhesion guidance law.

And secondly, taking the burnup and the distance from the detector to the obstacle as performance indexes, and taking the guidance parameters as optimization variables to respectively establish a guidance parameter optimization model with adjustable flight time and fixed flight time. And respectively carrying out off-line optimization under different initial states of the detector to obtain optimal guidance parameters under each initial state so as to form a training data set of the neural network.

The specific implementation method of the second step is as follows:

and establishing a guidance parameter optimization model with adjustable flight time. Selecting a performance index shown as a formula (5):

J＝Δm+μΔh _max (5)

wherein Δm represents fuel consumption, μ represents collision penalty weight, and Δh _max Representing the maximum distance of the detector from the surface of the obstacle along the Z-axis direction, and when the detector does not collide with the obstacle, making mu=0; otherwise, μ=2000.

The boundary constraint condition of the detector state quantity is shown as a formula (6):

wherein r is ₀ And v ₀ Respectively representing a position vector and a speed vector of the initial moment of the detector under the fixed coordinate system of the attachment point, m ₀ =500 kg represents the mass at the initial moment of the probe.

The thrust amplitude constraint is shown in formula (7):

0≤||T||≤T _max (7)

wherein T is _max =25n represents the thrust amplitude maximum value.

The optimization variable is the time of flight t _f And the obstacle avoidance coefficient lambda, wherein the constraint of the optimized variable value range is shown in the formula (8):

combining the formulas (1), (5), (6), (7) and (8) to obtain a guidance parameter optimization model with adjustable flight time:

the relative position vector defining the detector and the obstacle is shown in formula (10):

r _rel ＝[x _rel y _rel z _rel ] ^T (10)

wherein x is _rel And y _rel Representing the distance of the detector along the X-axis and the Y-axis to the edge of the obstacle, z _rel Representing the distance of the detector along the Z-axis to the upper surface of the obstacle. Under the fixed coordinate system of the attachment point, different initial positions r of the detector are selected ₀ Respectively calculating relative position vectors of the detector and the obstacle to form a relative position grid, wherein grid points of the relative position grid form a set { r }, in the example _rel ＝[x _rel y _rel z _rel ] ^T |x _rel ,z _rel ∈{50,100,150,…,1000},y _rel ∈{10,20,30, …,100 }. At each initial position r ₀ Respectively selecting initial speeds v with different sizes and directions ₀ A velocity grid is constructed, in this example, the grid points of the velocity grid form a set { v= [ v ] _x v _y v _z ] ^T |v _x ,v _y ,v _z E {0, -1, -2, …, -5 }. The relative position grid and the velocity grid together form a detector state quantity grid. And (3) at each grid point of the state quantity grid, optimizing and solving the optimal guidance parameters corresponding to the state quantity of the grid point by using the parameter optimization model of the formula (9), and recording the state quantity of each grid point and the optimal guidance parameters corresponding to the state quantity of each grid point to form a first training data set.

The optimization variable of the guidance parameter optimization model with fixed flight time is an obstacle avoidance coefficient lambda, and the parameter optimization model is shown in a formula (11):

based on the detector state quantity grids of the first training data set, a series of fixed flight times are selected at each state quantity grid point, and the selection mode of the flight times is shown as a formula (12):

where k is the magnification, a series of fixed times of flight are obtained by varying the value of k, in this example k=1, 2,3,4,5. And under different initial states and flight times, obtaining the optimal obstacle avoidance coefficient corresponding to each state quantity-flight time by using the parameter optimization model shown in the formula (11), and forming a second training data set.

And thirdly, based on the training data set generated in the neural network learning step two, establishing an intelligent mapping function relation between the detector state quantity and the guidance parameters. Two neural network models with flight time at an input end and an output end are built, two intelligent mapping function relations with adjustable flight time and fixed flight time are obtained, and guidance parameters are generated through the two intelligent mapping function relations at the initial time and the non-initial time of attachment. And calling an intelligent mapping function relation to update guidance parameters on line by combining with an optimal attachment guidance law, so that the detector can avoid obstacles according to real-time detector state information, obstacle information and attachment point information, and the safe and accurate attachment of the detector is ensured.

The specific implementation method of the third step is as follows:

the guidance parameters are related to the relative position of the probe-obstacle, the speed of the probe, etc. during probe attachment. Taking the state quantity in the first training data set generated in the second step as an input quantity, taking the corresponding optimal guidance parameters, namely the optimal flight time and the optimal obstacle avoidance coefficient, as output quantities, and establishing an intelligent guidance parameter mapping function with adjustable flight time based on neural network learning, wherein the intelligent mapping function expression is shown in a formula (13):

[t _f ,Λ]＝f ₁ (r _rel ,v) (13)

wherein f ₁ (. Cndot.) is a time-of-flight tunable intelligent mapping function. The intelligent mapping function is utilized to calculate guidance parameters on line in the attaching process, so that the problems of large calculated amount, low calculation efficiency and the like caused by an optimization algorithm are avoided.

Taking the state quantity and the flight time in the second training data set generated in the second step as input quantity, taking the corresponding optimal guidance parameter, namely the optimal obstacle avoidance coefficient, as output quantity, and establishing an intelligent guidance parameter mapping function with fixed flight time based on neural network learning, wherein the intelligent guidance parameter mapping function is shown as a formula (14):

Λ＝f ₂ (r _rel ,v,t _f ) (14)

wherein f ₂ (. Cndot.) is a time-of-flight tunable intelligent mapping function.

During the probe attachment process, the guidance law versus the remaining time of flight t is known from equation (3) _go Is more sensitive. When the remaining flight time t _g o uses an intelligent mapping function f ₁ Output value t of (-) () _f At the time of occurrence of the remaining flight time t _g o concussionAnd then causes the thrust output to oscillate, resulting in instability of the detector. For this purpose, a total time of flight t is calculated at the initial instant _f After which the remaining time of flight t _go By t _go ＝t _f -t, and taking the remaining flight time as an input quantity, and calculating an optimal obstacle avoidance coefficient in real time by an intelligent mapping function with fixed flight time, as shown in a formula (15). After the guidance parameters are generated in real time, the spatial non-cooperative target attached obstacle avoidance guidance method with the intelligent parameter adjustment is obtained, and the autonomous obstacle avoidance of the detector is realized.

And calling an intelligent mapping function relation to update guidance parameters on line by combining with an optimal attachment guidance law, so that the detector can avoid obstacles according to real-time detector state information, obstacle information and attachment point information, and the safe and accurate attachment of the detector is ensured.

FIG. 2 shows a plot of the adhesion trace of the celestial detector, and FIG. 3 shows a plot of the triaxial speed of the celestial detector, where v _x 、v _y 、v _z The three-axis speed components of the detector under the fixed coordinate system of the attachment point are respectively represented, and it can be seen that the detector realizes the precise attachment of the obstacle avoidance. Fig. 4 shows the obstacle avoidance coefficient curve of the intelligent mapping function output, and it can be seen that the obstacle avoidance coefficient increases when the detector approaches an obstacle, and the obstacle avoidance coefficient decreases rapidly after flying over the obstacle (about 800 seconds), so as to save fuel consumption. FIG. 5 shows a three-axis thrust curve of a celestial detector, where T _x 、T _y 、T _z And respectively representing triaxial thrust components of the detector under the attachment point fixed coordinate system, and meeting the thrust constraint can be seen. Simulation shows that the designed guidance method is suitable for obstacle avoidance attachment tasks of space non-cooperative targets.

While the foregoing is directed to embodiments of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof, and the scope thereof is determined by the claims that follow.

Claims (4)

1. The method for guiding the adhesion avoidance of the space non-cooperative target by intelligently adjusting parameters is characterized by comprising the following steps of: comprises the following steps of the method,

step one, under the condition that the attachment points are fixedly connected with a coordinate system, a detector attachment dynamics model is established; constructing an optimal attachment guidance law containing obstacle avoidance items, and reducing the risk of collision between the detector and the obstacle by changing the curvature of an attachment track;

step two, taking the burnup and the distance from the detector to the obstacle as performance indexes, and taking the guidance parameters as optimization variables, respectively establishing guidance parameter optimization models with adjustable flight time and fixed flight time; respectively carrying out off-line optimization under different initial states of the detector to obtain optimal guidance parameters under each initial state, and forming a training data set of a neural network model for updating the guidance parameters;

step three, based on the training data set generated in the neural network learning step two, establishing an intelligent mapping function relation between the detector state quantity and the guidance parameters; constructing two neural network models with flight time at an input end and an output end respectively, obtaining two intelligent mapping function relations with adjustable flight time and fixed flight time, and generating guidance parameters at an initial time and a non-initial time of attachment through the two intelligent mapping function relations respectively; and calling an intelligent mapping function relation to update guidance parameters on line by combining with an optimal attachment guidance law, so that the detector can avoid obstacles according to real-time detector state information, obstacle information and attachment point information, and the safe and accurate attachment of the detector is ensured.

2. The method for guiding the adhesion obstacle avoidance of the spatial non-cooperative target with intelligently-adjusted parameters according to claim 1, wherein the method comprises the following steps: the first implementation method of the step is that,

the specific implementation method of the first step is as follows:

defining an attachment point fixation coordinate system O-XYZ: taking the attachment point as an origin O of a coordinate system, taking the external normal direction of the surface of the small celestial body at the attachment point as a Z axis, pointing the X axis from the attachment point to a projection point of the initial position of the detector in the tangential plane of the surface of the small celestial body at the attachment point, and keeping the Y axis direction following a right rule;

the detector attachment dynamics model is established under the attachment point fixed coordinate system and is shown as (1):

wherein r and v respectively represent a position vector and a velocity vector of the detector under the attachment point fixing coordinate system, ζ=g- ω×ω×r-2ω×v represents a sum of gravitational acceleration and inertial acceleration of the detector under the attachment point fixing coordinate system, g represents a gravitational acceleration vector of the detector under the attachment point fixing coordinate system, ω represents a small celestial rotation angular velocity vector under the attachment point fixing coordinate system, a represents a thrust acceleration vector under the attachment point fixing coordinate system, T represents a detector thrust vector, m represents a detector mass, and I _sp Denote engine specific impulse, g ₀ A value representing the earth sea level gravitational acceleration;

taking thrust acceleration and detector position representing burnup as performance indexes, as shown in formula (2):

wherein t is _f Represents the time of flight, Λ= [ Λ ] _x Λ _y Λ _z ] ^T Representing obstacle avoidance coefficients, where Λ _x ,Λ _y ＞0，Λ _z < 0; when the curvature of the adhesion track has a convex change trend, the performance index function is reduced, so that the risk of collision between the detector and the small celestial body surface obstacle is reduced.

The optimal adhesion guidance law containing obstacle avoidance items is obtained as shown in the formula (3):

wherein t is _go Representing a remaining time of flight;

in order to consider the burnup performance and the obstacle avoidance performance of the detector, the obstacle avoidance coefficient Λ in the formula (3) is modified into a dynamic obstacle avoidance coefficient Λ' shown as the formula (4):

in the method, in the process of the invention,is defined as: the current position of the overdetector is taken as two vertical planes Σ _x Sum sigma _y Is parallel to OXZ plane and OYZ plane respectively, Σ _x Sum sigma _y Plane cutting obstacles each form an obstacle profile, tangential to the edges of the two obstacle profiles from the current position of the probe, Σ _x The angle between the tangent line in the plane and the negative direction of the OX axis is +.>The counterclockwise rotation direction is positive; sigma and method for producing the same _y The included angle between the tangent line in the plane and the negative direction of the OY axis is +.>The counterclockwise rotation direction is positive; the probe speed is at Σ _x Sum sigma _y The projections in the plane are v _xz And v _yz ，v _xz The angle between the direction of (2) and the negative direction of the axis of OX is +.>The counterclockwise rotation direction is positive; v _yz The angle between the direction of (2) and the negative direction of the OY axis is +.>The counterclockwise rotation direction is positive; if v _xz 0, then corresponding->Is 0; if v _yz 0, then corresponding->Is 0;

the simultaneous formulas (3) and (4) form an optimal adhesion guidance law containing obstacle avoidance terms, and the risk of collision of the detector with the obstacle is reduced by changing the curvature of the adhesion track according to the optimal adhesion guidance law.

3. The method for guiding the adhesion obstacle avoidance of the spatial non-cooperative target with intelligently-adjusted parameters according to claim 2, wherein the method comprises the following steps: the implementation method of the second step is that,

establishing a guidance parameter optimization model with adjustable flight time; selecting a performance index shown as a formula (5):

J＝Δm+μΔh _max (5)

wherein Δm represents fuel consumption, μ represents collision penalty weight, and Δh _max Representing the maximum distance of the detector from the surface of the obstacle along the Z-axis direction, and when the detector does not collide with the obstacle, making mu=0;

the boundary constraint condition of the detector state quantity is shown as a formula (6):

wherein r is ₀ And v ₀ Respectively representing a position vector and a speed vector of the initial moment of the detector under the fixed coordinate system of the attachment point, m ₀ Representing the quality of the initial moment of the detector;

the thrust amplitude constraint is shown in formula (7):

0≤||T||≤T _max (7)

wherein T is _max Representing a thrust amplitude maximum;

the optimization variable is the time of flight t _f And the obstacle avoidance coefficient lambda, wherein the constraint of the optimized variable value range is shown in the formula (8):

combining the formulas (1), (5), (6), (7) and (8) to obtain a guidance parameter optimization model with adjustable flight time:

the relative position vector defining the detector and the obstacle is shown in formula (10):

r _rel ＝[x _rel y _rel z _rel ] ^T (10)

wherein x is _rel And y _rel Representing the distance of the detector along the X-axis and the Y-axis to the edge of the obstacle, z _rel Representing the distance of the detector along the Z-axis to the upper surface of the obstacle; under the fixed coordinate system of the attachment point, different initial positions r of the detector are selected ₀ Respectively calculating relative position vectors of the detector and the obstacle to form a relative position grid; at each initial position r ₀ Respectively selecting initial speeds v with different sizes and directions ₀ Forming a speed grid, wherein the relative position grid and the speed grid jointly form a detector state quantity grid; at each grid point of the state quantity grid, optimizing and solving the optimal guidance parameters corresponding to the state quantity of the grid point by using a parameter optimization model of the formula (9), and recording the state quantity of each grid point and the optimal guidance parameters corresponding to the state quantity of each grid point to form a first training data set;

the optimization variable of the guidance parameter optimization model with fixed flight time is an obstacle avoidance coefficient lambda, and the parameter optimization model is shown in a formula (11):

based on the detector state quantity grids of the first training data set, selecting a series of fixed flight times at grid points of each state quantity grid, wherein the selection mode of the flight times is shown as a formula (12):

wherein k is the magnification, and a series of fixed flight times are obtained by changing the value of k; and under different initial states and flight times, obtaining the optimal obstacle avoidance coefficient corresponding to each state quantity-flight time by using the parameter optimization model shown in the formula (11), and forming a second training data set.

4. The method for spatial non-cooperative target attachment obstacle avoidance guidance with intelligent parameter adjustment according to claim 3, wherein: the implementation method of the third step is that,

the specific implementation method of the third step is as follows:

taking the state quantity in the first training data set generated in the second step as an input quantity, taking the corresponding optimal guidance parameters, namely the optimal flight time and the optimal obstacle avoidance coefficient, as output quantities, and establishing an intelligent mapping function relation of the guidance parameters with adjustable flight time based on neural network learning, wherein an expression of the intelligent mapping function relation is shown as a formula (13):

[t _f ,Λ]＝f ₁ (r _rel ,v) (13)

wherein f ₁ (. Cndot.) is an intelligent mapping function with adjustable flight time; the intelligent mapping function relation is utilized to calculate guidance parameters on line in the attaching process, so that the problems of large calculated amount, low calculation efficiency and the like caused by an optimization algorithm are avoided;

taking the state quantity and the flight time in the second training data set generated in the second step as input quantity, taking the corresponding optimal guidance parameter, namely the optimal obstacle avoidance coefficient, as output quantity, and establishing an intelligent mapping function relation of the guidance parameter with fixed flight time based on neural network learning, wherein the intelligent mapping function relation is shown as a formula (14):

Λ＝f ₂ (r _rel ,v,t _f ) (14)

wherein f ₂ (. Cndot.) is an intelligent mapping function with fixed flight time;

during the probe attachment process, the guidance law versus the remaining time of flight t is known from equation (3) _go Is more sensitive; when the remaining flight time t _g o uses an intelligent mapping function f ₁ Output value t of (-) () _f At the time of occurrence of the remaining flight time t _g o, further, the thrust output is oscillated, so that the detector is unstable; for this purpose, a total time of flight t is calculated at the initial instant _f After which the remaining time of flight t _go By t _go ＝t _f -t is calculated, the remaining flight time is used as an input quantity, and the optimal obstacle avoidance coefficient is calculated in real time by an intelligent mapping function with fixed flight time, as shown in a formula (15);

and calling an intelligent mapping function relation to update guidance parameters on line by combining with an optimal attachment guidance law, so that the detector can avoid obstacles according to real-time detector state information, obstacle information and attachment point information, and the safe and accurate attachment of the detector is ensured.