Space tethered robot releasing method considering safety constraint

Abstract

The embodiment of the invention discloses a space tethered robot releasing method and device considering safety constraints, a medium and equipment. The collision avoidance problem of a space tethered robot and space debris is decomposed into an offline training part and an online execution part. In the off-line stage, firstly, a control compression mapping theory is used for deducing conditions which need to be met by a tracking controller, and a neural network is used for searching the controller meeting the conditions; and then a trajectory planner is constructed based on an optimization method. In the online stage, when the space tethered robot detects space debris, the trajectory planner generates a reference trajectory meeting safety constraints, and then the tracking controller tracks the reference trajectory, so that the space tethered robot does not collide with the space debris in the releasing process, and safe releasing of the space tethered robot is achieved.

Description

Technical Field

[0001] The present application relates to the field of space tethered robots, and in particular to a method, device, medium and equipment for releasing a space tethered robot taking safety constraints into consideration. Background Art

[0002] Space tethered robots are widely used in artificial gravity, asteroid exploration and other space missions due to their flexibility, safety and low cost. However, with the increasing number of space debris in Earth orbit, the safety of various space robots is seriously affected. In addition, collisions between space robots and these space debris will generate more debris, further deteriorating the environment in Earth orbit. Summary of the invention

[0003] The main purpose of the present application is to provide a method, device, medium and equipment for releasing a space tethered robot taking into account safety constraints, aiming to avoid collision between the space tethered robot and space debris or other spacecraft during the release process, so as to ensure the safety of the space tethered robot.

[0004] To achieve the above-mentioned objectives, the first aspect of the present application provides a method for releasing a space tethered robot taking safety constraints into consideration, comprising: obtaining an initial state value and an initial control value of the space tethered robot; in an offline stage, determining the controller constraints that the tracking controller of the space tethered robot needs to satisfy according to the control compression mapping theory, and constructing the tracking controller that satisfies the controller constraints based on a deep neural network, wherein the tracking controller is used to track an arbitrary trajectory; determining a performance indicator function of a trajectory planner according to control variables and state variables of the tracking controller, wherein the control variables satisfy the control constraints of the tracking controller, and the state variables satisfy the state constraints and safety constraints of the tracking controller; in an online stage, detecting space debris around the space tethered robot, updating the safety constraints based on the position of the space debris, and solving the performance indicator function based on the safety constraints, the control variables and the state variables to generate a reference trajectory; determining a tracking trajectory of the tracking controller according to the reference trajectory, and releasing the space tethered robot based on the tracking trajectory.

[0005] Optionally, the expression of the controller constraint condition includes:

[0006]

[0007] Among them, b i is the i-th column of matrix B, u i is the i-th element of the system control input u, M is a positive definite real symmetric matrix, B ⊥ satisfy

[0008] Optionally, the controller includes a real symmetric matrix M and a controller gain k, and constructing the tracking controller that satisfies the controller constraints based on a deep neural network includes: determining a positive definite real symmetric matrix M based on a first deep neural network; determining the controller gain k based on a second deep neural network, wherein the expression of the real symmetric matrix M and the expression of the controller gain k are respectively:

[0009]

[0010] In the formula, a is the activation function, are the parameters of the first deep neural network, are parameters of the second deep neural network, x* is a reference state variable, and x is a state variable; the loss function of the tracking controller is determined based on the expression of the controller constraint; the first deep neural network and the second deep neural network are trained based on the loss function of the tracking controller to obtain the tracking controller that satisfies the controller constraint.

[0011] Optionally, the expression based on the controller constraint condition determines a loss function of the tracking controller, the loss function comprising an average of a first sub-loss function to a fourth sub-loss function;

[0012] The expression of the first sub-loss function is:

[0013]

[0014] Among them, φ(x, x * ,u * θ M ,θ u ) represents the left side of the first inequality in the expression of the controller constraint, represents the expected value, Indicates in space The uniform distribution on σ is a very small positive number. is a non-positive definite penalty function if and only if A>0,

[0015] The expression of the second sub-loss function is:

[0016]

[0017] Wherein, C1 represents the left side of the first equation in the expression of the controller constraint, |||| F represents the F-norm;

[0018] The expression of the third sub-loss function is:

[0019] M(x;θ M )= α I+m(x;θ M ) T m(x;θ M )

[0020] in, α is the lower bound of the condition number of the matrix M, m(x; θ M ) is a first deep neural network;

[0021] The expression of the fourth sub-loss function is:

[0022]

[0023] The average expression of the first to fourth sub-loss functions is:

[0024]

[0025] in, It's space The data is uniformly sampled.

[0026] Optionally, the space tethered robot comprises a platform, a space tether and a terminal maneuvering unit connected in sequence, and the step of determining a performance index function of a trajectory planner according to a control variable and a state variable of the tracking controller comprises:

[0027] Get the expression of the security constraint, where the expression of the security constraint is:

[0028]

[0029] Among them, D p (x) represents the distance from the end maneuvering unit to the center of the circle p, d p (x) represents the distance from the space tether to the center of the circle p, and h represents the size of the safety domain;

[0030] The optimization problem of the trajectory planner is determined based on the following expression

[0031]

[0032] Among them, g is the performance index function, x0 and x f denote the initial state and final state of the space tethered robot, respectively; x(t) and u(t) denote the state variable and the control variable, respectively; represents the state constraints of the space tethered robot, denotes the control constraints of the space tethered robot, t0 and t f Represent the initial time and final time respectively;

[0033]

[0034] z1=θ,z2=ξ-1, x = [z1, z2, z3, z4] T , u=[u1,u2] T .

[0035] θ is the in-plane angle between the orbital coordinate system Oxyz and the space tethered robot body coordinate system Ox2y2z2, ξ represents the normalized space tether length, represents the derivative of ξ, represents the derivative of θ, and u1 and u2 represent the controller inputs respectively.

[0036] Optionally, the expression of the performance indicator function is:

[0037] g(x(t),u(t))=1 / 2·[x(t) T Qx(t)+u(t) T Ru(t)]

[0038] in,

[0039] In order to achieve the above-mentioned purpose, the second aspect of the present application provides a space tethered robot releasing device considering safety constraints, including: an acquisition module, used to acquire the initial state value and initial control value of the space tethered robot; a controller construction module, used to determine the controller constraints that the tracking controller of the space tethered robot needs to satisfy according to the control compression mapping theory in an offline stage, and construct the tracking controller that satisfies the controller constraints based on a deep neural network, and the tracking controller is used to track an arbitrary trajectory; a trajectory planner construction module, used to determine the performance indicator function of the trajectory planner according to the control variables and state variables of the tracking controller, wherein the control variables satisfy the control constraints of the tracking controller, and the state variables satisfy the state constraints and safety constraints of the tracking controller; a reference trajectory generation module, used to detect space debris around the space tethered robot in an online stage, update the safety constraints based on the position of the space debris, and solve the performance indicator function based on the safety constraints, the control variables and the state variables to generate a reference trajectory; a space robot releasing module, used to determine the tracking trajectory of the tracking controller according to the reference trajectory, the initial state value and the initial control value, and release the space tethered robot based on the tracking trajectory.

[0040] Optionally, the controller includes a real symmetric matrix M and a controller gain k, and the controller construction module is further used to: determine a positive definite real symmetric matrix M based on a first deep neural network; determine a controller gain k based on a second deep neural network, wherein the expression of the real symmetric matrix M and the expression of the controller gain k are respectively:

[0041]

[0042] In the formula, a is the activation function, are the parameters of the first deep neural network, is the parameter of the second deep neural network, x * is the reference state variable, x is the state variable;

[0043] The loss function of the tracking controller is determined based on the expression of the controller constraint condition; and the first deep neural network and the second deep neural network are trained based on the loss function of the tracking controller to obtain the tracking controller that satisfies the controller constraint condition.

[0044] To achieve the above-mentioned objectives, the third aspect of the present application further provides a computer-readable storage medium, which includes instructions, which, when executed on a computer, enables the computer to execute the space tethered robot release method considering safety constraints provided in the first aspect.

[0045] To achieve the above-mentioned purpose, the fourth aspect of the present application also provides an electronic device, which includes: at least one processor, a memory and an input-output unit; wherein the memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the space tethered robot release method considering safety constraints provided in the first aspect.

[0046] The embodiments of the present application propose a method, device, medium and equipment for releasing a space tethered robot that takes safety constraints into consideration, by decomposing the collision avoidance problem between the space tethered robot and space debris into an offline training part and an online execution part. In the offline stage, the control compression mapping theory is first used to derive the conditions that the tracking controller must meet, and a neural network is used to find a controller that meets the conditions; then a trajectory planner is constructed based on an optimization method. In the online stage, when the space tethered robot detects space debris, the trajectory planner generates a reference trajectory that meets the safety constraints, and then the tracking controller tracks this reference trajectory, so that the space tethered robot will not collide with the space debris during the release process, thereby achieving the safe release of the space tethered robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic flow chart of an embodiment of a method for releasing a space tethered robot taking safety constraints into consideration in this application;

[0048] Figure 2 A schematic diagram of a release trajectory and a reference trajectory of a space tethered robot provided in accordance with an embodiment of a method for releasing a space tethered robot taking safety constraints into consideration in this application;

[0049] Figure 3 A structural block diagram is provided for an embodiment of a space tethered robot release device taking safety constraints into consideration for the present application.

[0050] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0051] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0052] The purpose of the present application method is to design a safe release method for a space tethered robot to avoid collision with space debris or other spacecraft during the release process, so as to ensure the safety of the space tethered robot.

[0053] To achieve the above purpose, the technical solution adopted in this application includes the following steps:

[0054] (1) Establish the dynamics of the space tethered robot;

[0055] (2) Based on the control compression theory, the conditions that the tracking controller must meet are derived, and a tracking controller is constructed using a neural network. The neural network is trained to obtain a controller that can track any feasible trajectory.

[0056] (3) An online trajectory planner is constructed based on the optimization method, which can plan a reference trajectory that can avoid collision with space debris in real time while considering safety constraints;

[0057] (4) A tracking controller is used to track the reference trajectory so that the tethered space robot can avoid collision with space debris when released, thus ensuring safety.

[0058] The method for releasing a space tethered robot taking safety constraints into consideration is further described below in conjunction with the accompanying drawings and embodiments.

[0059] Reference Figure 1 The first embodiment of the present application provides a space tethered robot release method considering safety constraints. The space tethered robot release method considering safety constraints may include:

[0060] S10, obtaining an initial state value and an initial control value of the space tethered robot;

[0061] S20, in the offline stage, determining the controller constraints that the tracking controller of the space tethered robot needs to satisfy according to the control compression mapping theory, and constructing a tracking controller that satisfies the controller constraints based on a deep neural network, where the tracking controller is used to track an arbitrary trajectory;

[0062] S30, determining a performance index function of the trajectory planner according to the control variables and state variables of the tracking controller, wherein the control variables satisfy the control constraints of the tracking controller, and the state variables satisfy the state constraints and safety constraints of the tracking controller;

[0063] S40, in the online stage, detecting space debris around the space tethered robot, updating safety constraints based on the positions of the space debris, and solving a performance indicator function based on the safety constraints, control variables, and state variables to generate a reference trajectory;

[0064] S50, determining a tracking trajectory of the tracking controller according to the reference trajectory, the initial state value and the initial control value, and releasing the space tethered robot based on the tracking trajectory.

[0065] The general method of space release considering safety constraints proposed in this embodiment decomposes the collision avoidance problem between the space tethered robot and space debris into an offline training part and an online execution part. In the offline stage, the control compression mapping theory is first used to derive the conditions that the tracking controller must meet, and a neural network is used to find a controller that meets the conditions; then a trajectory planner is constructed based on the optimization method. In the online stage, when the space tethered robot detects space debris, the trajectory planner generates a reference trajectory that meets the safety constraints, and then the tracking controller tracks this reference trajectory, so that the space tethered robot will not collide with the space debris during the release process, thereby achieving the safe release of the space tethered robot. Compared with existing research, the method of this application has the following advantages: the collision avoidance problem between the space tethered robot and space debris is decoupled into an offline training part and an online execution part, so that the method has greater flexibility; the tracking controller is found using a neural network, without the need for a complex controller design process; the trajectory planner is constructed based on the optimization method, so that the reference trajectory can be planned in real time while considering safety constraints.

[0066] In an embodiment of the present application, the expression of the controller constraint condition includes:

[0067]

[0068] Among them, b i is the i-th column of matrix B, u i is the i-th element of the system control input u, M is a positive definite real symmetric matrix, B ⊥ satisfy

[0069] Specifically, Figure 1 As shown in the figure, the space tethered robot is mainly composed of a platform, a space tether and a terminal maneuvering unit. When the space tethered robot is released, the platform and the terminal maneuvering unit can be regarded as mass points, and the space tether is always kept in a tensioned state. The inertial coordinate system is EXYZ, the orbital coordinate system is Oxyz, and the space tethered robot body coordinate system is Ox2y2z2. The angle between the latter two coordinate systems is defined as the in-plane angle θ and the out-of-plane angle β.

[0070] In-plane motion of a tethered space robot Its kinetic equation can be written as:

[0071]

[0072] Among them, z1=θ, z2=ξ-1, x = [z1, z2, z3, z4] T , u=[u1,u2] T , d represents a bounded external disturbance, and

[0073]

[0074] in, n=4, m=2.

[0075] After obtaining the dynamic equation of the space tethered robot, the processor can construct a tracking controller based on the dynamic equation. The controller includes a real symmetric matrix M and a controller gain k. Step S20 can include the following execution process:

[0076] S201, determining a positive definite real symmetric matrix M based on a first deep neural network;

[0077] S202: Determine controller gain k based on the second deep neural network

[0078] Among them, the expression of the real symmetric matrix M and the expression of the controller gain k are:

[0079]

[0080] In the formula, a is the activation function, are the parameters of the first deep neural network, is the parameter of the second deep neural network, x * is the reference state variable, x is the state variable;

[0081] S203, determining a loss function of the tracking controller based on an expression of the controller constraint condition;

[0082] S204: Train the first deep neural network and the second deep neural network based on the loss function of the tracking controller to obtain a tracking controller that satisfies the controller constraints.

[0083] The processor uses a neural network to find a tracking controller by executing steps S201 to S204, without the need for a complicated controller design process, thereby speeding up the design process of the tracking controller and saving time. The space tethered robot is decoupled from the space debris collision avoidance problem, making the method more flexible.

[0084] Next, four sub-loss functions are constructed based on the conditions that the tracking controller needs to meet.

[0085] Determine a loss function of the tracking controller based on an expression of the controller constraint, wherein the loss function includes an average of first to fourth sub-loss functions;

[0086] The expression of the first sub-loss function is:

[0087]

[0088] Among them, φ(x, x * ,u * θ M ,θ u ) represents the left side of the first inequality in the expression of the controller constraint. represents the expected value, Indicates in space uniform distribution on the surface, σ is a very small positive number, is a non-positive definite penalty function if and only if A>0,

[0089] The expression of the second sub-loss function is:

[0090]

[0091] Where C1 represents the left side of the first equation in the expression of the controller constraint, |||| F represents the F-norm;

[0092] The expression of the third sub-loss function is:

[0093] M(x;θ M )= α I+m(x;θ M ) T m(x;θ M )

[0094] in, α is the lower bound of the condition number of the matrix M, m(x; θM ) is a first deep neural network;

[0095] The expression of the fourth sub-loss function is:

[0096]

[0097] in, is the upper bound of the condition number of the matrix M.

[0098] Combining the above four sub-loss functions, we get the loss function for training the first deep neural network and the second deep neural network.

[0099] The average expression of the first to fourth sub-loss functions is:

[0100]

[0101] in, It's space The data is uniformly sampled.

[0102] By executing the above steps, the processor obtains the relevant parameters of the first deep neural network and the second deep neural network, that is, and Thus, the real symmetric matrix M and controller gain k of the tracking controller are determined, and the tracking controller is also determined.

[0103] In the offline stage, it is also necessary to build a trajectory planner for the space robot based on the optimization method, so that the reference trajectory can be planned in real time while considering safety constraints.

[0104] In an embodiment of the present application, step S30 may include the following execution process:

[0105] S301. Obtain an expression of a security constraint, where the expression of the security constraint is:

[0106]

[0107] Where Dp(x) represents the distance from the terminal maneuvering unit to the center p of the circle, dp(x) represents the distance from the space tether to the center p of the circle, and h represents the size of the safety domain;

[0108] S302: Determine the optimization problem of the trajectory planner based on the following expression

[0109]

[0110] Among them, g is the performance index function, x0 and x f They represent the initial state and final state of the space tethered robot, respectively. x(t) and u(t) represent the state variable and control variable, respectively. represents the state constraints of the space tethered robot, represents the control constraints of the space tethered robot, t0 and t f Represent the initial time and final time respectively;

[0111]

[0112] z1=θ,z2=ξ-1, x = [z1, z2, z3, z4] T , u=[u1,u2] T .

[0113] θ is the in-plane angle between the orbital coordinate system Oxyz and the space tethered robot body coordinate system Ox2y2z2, ξ represents the normalized space tether length, represents the derivative of ξ, represents the derivative of θ, and u1 and u2 represent the controller inputs respectively.

[0114] It can be understood that before obtaining the safety constraints, it can be assumed that the space debris detected when the space robot is released is located in a circle with the center of the circle being p, and that neither the terminal maneuvering unit nor the space tether can collide with the space debris.

[0115] In this embodiment, by solving the above optimization problem online, a reference trajectory (x * ,u * ).

[0116] Exemplarily, the expression of the performance indicator function is:

[0117] g(x(t),u(t))=1 / 2·[x(t) T Qx(t)+u(t) T Ru(t)]

[0118] in,

[0119] After the online phase is over, the processor can enter the online phase, in which the processor can use the tracking controller to track the reference trajectory. That is, during the release process of the space tethered robot, its sensors continuously detect whether there are space debris around it. If there are debris, the safety constraints are updated, and a collision-free reference trajectory (x*, u*) is generated by solving the optimization problem of the trajectory planner online. Then, the tracking controller is used to track the reference trajectory to achieve the safe release of the space tethered robot.

[0120] In order to verify the effectiveness of the method of this application, numerical simulation is carried out below.

[0121] refer to Figure 2 , assuming that the total length of the space tether is 2000m, the equivalent mass of the space tethered robot is 10kg, and the orbital angular velocity is 0.00117rad / s; the initial release position is (0m, -2000m), and the target position of the release is (0m, 0m); the center position of the circle formed by the detected space debris is (300m, -1200m), its radius is 80m, and the size of the safety area is 40m (h=40m); the relevant parameters of the tracking controller are λ=2.6748, The performance indicator function in the trajectory planner is defined as g(x(t),u(t))=1 / 2·[x(t) T Qx(t)+u(t) T Ru(t)], The upper bound of the interference is ζ=1×10 -2 The number of simulation steps is 1500. The simulation results are as follows Figure 2 As shown in Figure 2, it can be seen that the reference trajectory planned by the trajectory planner complies with the safety constraints, and the tracking controller found by the neural network can track the reference trajectory well, realizing the safe release of the space tethered robot.

[0122] refer to Figure 3 On the basis of the above embodiments, the present application further provides a space tethered robot releasing device considering safety constraints. The space tethered robot releasing device 100 may include an acquisition module 101, a controller construction module 102, a trajectory planner construction module 103, a reference trajectory generation module 104 and a space robot releasing module 105.

[0123] The acquisition module 101 is used to acquire the initial state value and initial control value of the space tethered robot;

[0124] The controller construction module 102 is used to determine the controller constraints that the tracking controller of the space tethered robot needs to satisfy according to the control compression mapping theory in the offline stage, and to construct a tracking controller that satisfies the controller constraints based on a deep neural network, and the tracking controller is used to track any trajectory;

[0125] The trajectory planner construction module 103 is used to determine the performance index function of the trajectory planner according to the control variables and state variables of the tracking controller, wherein the control variables satisfy the control constraints of the tracking controller, and the state variables satisfy the state constraints and safety constraints of the tracking controller;

[0126] The reference trajectory generation module 104 is used to detect space debris around the space tethered robot in the online stage, update the safety constraint conditions based on the positions of the space debris, and solve the performance indicator function based on the safety constraint conditions, control variables and state variables to generate a reference trajectory;

[0127] The space robot releasing module 105 is used to determine the tracking trajectory of the tracking controller according to the reference trajectory, the initial state value and the initial control value, and release the space tethered robot based on the tracking trajectory.

[0128] In an embodiment of the present application, the controller includes a real symmetric matrix M and a controller gain k, and the controller construction module 102 can be specifically used for:

[0129] Determine a positive definite real symmetric matrix M based on the first deep neural network;

[0130] Determine the controller gain k based on the second deep neural network

[0131] Among them, the expression of the real symmetric matrix M and the expression of the controller gain k are:

[0132]

[0133] In the formula, a is the activation function, are the parameters of the first deep neural network, is the parameter of the second deep neural network, x * is the reference state variable, x is the state variable;

[0134] Determine the loss function of the tracking controller based on the expression of the controller constraints;

[0135] Based on the loss function of the tracking controller, the first deep neural network and the second deep neural network are trained to obtain a tracking controller that meets the controller constraints.

[0136] Based on the above embodiments, the present application also provides a computer-readable storage medium including instructions, which, when executed on a computer, enables the computer to execute the space tethered robot release method considering safety constraints provided in any of the above embodiments.

[0137] Based on the above embodiments, the present application also provides an electronic device, which includes: at least one processor, a memory and an input-output unit; wherein the memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the space tethered robot release method considering safety constraints provided in the above embodiments.

[0138] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for releasing a space tethered robot considering safety constraints, characterized in that: include: Obtaining the initial state value and initial control value of the space tethered robot; In the offline stage, the controller constraints that the tracking controller of the space tethered robot needs to satisfy are determined according to the control compression mapping theory, and the tracking controller satisfying the controller constraints is constructed based on a deep neural network, and the tracking controller is used to track an arbitrary trajectory; Determining a performance indicator function of a trajectory planner according to control variables and state variables of the tracking controller, wherein the control variables satisfy control constraints of the tracking controller, and the state variables satisfy state constraints and safety constraints of the tracking controller; In an online stage, space debris around the space tethered robot is detected, the safety constraint condition is updated based on the position of the space debris, and the performance indicator function is solved based on the safety constraint condition, the control variable and the state variable to generate a reference trajectory; A tracking trajectory of the tracking controller is determined according to the reference trajectory, the initial state value and the initial control value, and the spatial tethered robot is released based on the tracking trajectory.

2. The method for releasing a space tethered robot considering safety constraints as claimed in claim 1, characterized in that: The expressions of the controller constraints include: Among them, b i is the i-th column of matrix B, u i is the i-th element of the system control input u, M is a positive definite real symmetric matrix, B ⊥ satisfy 3. The method for releasing a space tethered robot considering safety constraints as claimed in claim 1, characterized in that: The controller includes a real symmetric matrix M and a controller gain k, and the tracking controller that satisfies the controller constraint condition is constructed based on a deep neural network, including: Determine a positive definite real symmetric matrix M based on the first deep neural network; Determine the controller gain k based on the second deep neural network Among them, the expression of the real symmetric matrix M and the expression of the controller gain k are respectively: In the formula, a is the activation function, are the parameters of the first deep neural network, are the parameters of the second deep neural network, x* is the reference state variable, and x is the state variable; Determining a loss function of the tracking controller based on an expression of the controller constraints; The first deep neural network and the second deep neural network are trained based on the loss function of the tracking controller to obtain the tracking controller that satisfies the controller constraints.

4. The method for releasing a space tethered robot considering safety constraints as claimed in claim 1, characterized in that: The loss function includes the average of the first sub-loss function to the fourth sub-loss function; The loss function of the tracking controller is determined based on the expression of the controller constraint condition, including Construct the first sub-loss function, the expression of the first sub-loss function is: Among them, φ(x, x*, u*; θ M ,θ u ) represents the left side of the first inequality in the expression of the controller constraint, represents the expected value, Indicates in space The uniform distribution on σ is a very small positive number. is a non-positive definite penalty function if and only if hour, Construct the second sub-loss function, the expression of which is: Wherein, C1 represents the left side of the first equation in the expression of the controller constraint, |||| F Represents the F-norm: Construct a third sub-loss function, the expression of the third sub-loss function is: M(x;θ M )= α I+m(x;θ M ) T m(x;θ M ) Among them, α is the lower bound of the condition number of the matrix M, m(x; θ M ) is a first deep neural network; Construct a fourth sub-loss function, the expression of the fourth sub-loss function is: in, is the upper bound of the condition number of the matrix M; The loss function is constructed based on the first to fourth sub-loss functions, and the expression of the loss function is: in, It's space The data is uniformly sampled.

5. The method for releasing a space tethered robot considering safety constraints as claimed in claim 1, characterized in that: The space tethered robot comprises a platform, a space tether and a terminal maneuvering unit connected in sequence, and the performance index function of the trajectory planner is determined according to the control variables and state variables of the tracking controller, including: Get the expression of the security constraint, where the expression of the security constraint is: In the above formula, Dp(x) represents the distance from the terminal maneuvering unit to the center p of the circle, dp(x) represents the distance from the space tether to the center p of the circle, and h represents the size of the safety zone; The optimization problem of the trajectory planner is determined based on the following expression: x(0)=x0, x(t f )=x f In the above formula, g is the performance index function, x0 and x f denote the initial state and final state of the space tethered robot, respectively; x(t) and u(t) denote the state variable and the control variable, respectively; represents the state constraints of the space tethered robot, denotes the control constraints of the space tethered robot, t0 and t f Represent the initial time and final time respectively. Among them, z1=θ, z2=ξ-1, x=[z1, z2, z3, z4]T, u=[u1, u2]T. is the in-plane angle between the orbital coordinate system Oxyz and the space tethered robot body coordinate system Ox2t2z2, ξ represents the normalized space tether length, represents the derivative of ξ, express The derivatives of , u1 and u2 represent the controller inputs respectively.

6. The method for releasing a space tethered robot considering safety constraints as claimed in claim 5, characterized in that: The expression of the performance indicator function is: g(x(t),u(t))=1 / 2·[x(t) T Qx(t)+u(t) T Ru(t)] in, 7. A space tethered robot release device considering safety constraints, characterized in that: include: An acquisition module, used for acquiring an initial state value and an initial control value of the space tethered robot; A controller construction module, used for determining, in an offline stage, controller constraints that the tracking controller of the space tethered robot needs to satisfy according to the control compression mapping theory, and constructing the tracking controller satisfying the controller constraints based on a deep neural network, wherein the tracking controller is used to track an arbitrary trajectory; A trajectory planner construction module, used to determine a performance indicator function of the trajectory planner according to control variables and state variables of the tracking controller, wherein the control variables satisfy the control constraints of the tracking controller, and the state variables satisfy the state constraints and safety constraints of the tracking controller; A reference trajectory generation module is used to detect space debris around the space tethered robot in an online stage, update the safety constraint condition based on the position of the space debris, and solve the performance indicator function based on the safety constraint condition, the control variable and the state variable to generate a reference trajectory; The space robot release module is used to determine the tracking trajectory of the tracking controller according to the reference trajectory, the initial state value and the initial control value, and release the space tethered robot based on the tracking trajectory.

8. The space tethered robot release device considering safety constraints as claimed in claim 1, characterized in that: The controller includes a real symmetric matrix M and a controller gain k, and the controller building block is further used for: Determine a positive definite real symmetric matrix M based on the first deep neural network; Determine the controller gain k based on the second deep neural network Among them, the expression of the real symmetric matrix M and the expression of the controller gain k are respectively: In the formula, a is the activation function, are the parameters of the first deep neural network, is the parameter of the second deep neural network, x * is the reference state variable, x is the state variable; Determining a loss function of the tracking controller based on an expression of the controller constraints; The first deep neural network and the second deep neural network are trained based on the loss function of the tracking controller to obtain the tracking controller that satisfies the controller constraints.

9. A computer-readable storage medium, characterized in that: The method comprises instructions, which, when executed on a computer, enable the computer to execute the method for releasing a space tethered robot taking safety constraints into consideration as claimed in any one of claims 1 to 6.

10. An electronic device, characterized in that: The electronic device comprises: at least one processor, memory, and input-output unit; The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the space tethered robot release method considering safety constraints according to any one of claims 1 to 6.

Images