Intelligent prediction guidance method for Mars pneumatic auxiliary orbit descending

Abstract

The invention discloses an intelligent prediction guidance method for Mars pneumatic auxiliary orbit descending, and belongs to the field of deep space exploration. According to the method, a high-precision kinetic model of the Mars pneumatic-assisted orbit descent is constructed, a network training data set is generated by utilizing the high-precision model, a near-fire-point state intelligent forecasting model and a far-fire-point state intelligent forecasting model of the Mars pneumatic-assisted orbit descent are constructed, and meanwhile, near-fire-point heat flow constraints are constructed; designing a predictive guidance method by taking the orbit height deviation and the control input as optimization performance indexes; and optimizing and solving the control input to control the detector until the track height of the detector is reduced to be close to the preset track height, so that the detector is reduced to the preset track under the condition of meeting heat flow and control amplitude constraints, and meanwhile, the fuel consumption of the detector is reduced. According to the method, by solving the states of the near fire point and the far fire point under the two-body dynamic model, the network is constructed to predict the deviation value between the states of the near fire point and the far fire point under the high-precision dynamic model and the prediction result of the two-body model, and the training cost of the neural network is reduced.

Description

Technical Field

[0001] The present invention relates to an intelligent prediction and guidance method, in particular to an intelligent prediction and guidance method for aerodynamically assisted orbit lowering on Mars, and belongs to the technical field of deep space exploration. Background Art

[0002] Mars exploration is of great significance to the pursuit of the origin of the universe and the origin of life, as well as the search for extraterrestrial life and habitable planets. Therefore, it has always been the focus of deep space exploration by major space powers. In order to reduce fuel consumption during the capture and orbit reduction phases, the aerodynamically assisted orbit change technology that uses aerodynamic flight to adjust the orbit for celestial bodies with an atmosphere such as Mars has received widespread attention. During the aerodynamically assisted orbit reduction process on Mars, the use of predictive guidance methods can not only accurately lower the probe to the predetermined orbit under the conditions of meeting the constraints of heat flow and control amplitude, but also reduce the fuel consumption of the probe, allowing the probe to carry more scientific payloads, significantly improving the scientific benefits and resource utilization efficiency of the Mars exploration mission.

[0003] Since the Mars aerodynamically assisted orbit lowering process is complex and lasts for a long time, in order to make the predicted probe state as close to the actual state as possible, the dynamic model used in the prediction and guidance method must take into account the influence of the irregular spherical gravity perturbation of Mars. However, the use of a low-order Mars irregular spherical gravity model will lead to a large deviation between the predicted trajectory and the actual trajectory, which cannot meet the accuracy requirements; while the use of a high-order Mars irregular spherical gravity model can improve the accuracy, it will significantly reduce the calculation efficiency and it is difficult to meet the real-time requirements of orbit applications. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent prediction and guidance method for Mars aerodynamic-assisted orbit reduction, which uses a high-precision dynamic model of Mars aerodynamic-assisted orbit reduction to generate training data, uses a neural network to learn the mapping relationship between input and output, and constructs a fast and high-precision state intelligent prediction model. The constructed fast and high-precision state prediction model is used to construct a prediction and guidance method, which can accurately lower the probe to a predetermined orbit while satisfying the constraints of heat flow and control amplitude, while reducing the fuel consumption of the probe.

[0005] The object of the present invention is achieved through the following technical solutions.

[0006] The invention discloses an intelligent prediction and guidance method for Mars aerodynamically assisted orbit reduction, constructs a high-precision dynamic model for Mars aerodynamically assisted orbit reduction, takes into account that the control is mainly applied at the far-fire point, and requires that the near-fire point state and the control input in the attachment process meet the constraints of heat flow and control amplitude, and combines neural networks to respectively construct a near-fire point state intelligent prediction model for predicting the near-fire point state from the initial state and a far-fire point state intelligent prediction model for predicting the far-fire point state from the initial state. The data set used for neural network training is generated by the constructed Mars aerodynamically assisted orbit reduction high-precision dynamic model. In the process of data set generation, the value range of the initial state is defined, and sampling is performed according to uniform distribution. The initial state is recursively deduced using the Mars aerodynamically assisted orbit reduction high-precision dynamic model to collect the near-fire point state and the far-fire point state at the next moment, and the near-fire point state obtained by the high-precision dynamic model is subtracted from the near-fire point state obtained by the two-body dynamic model to obtain the near-fire point state correction amount, and the far-fire point state obtained by the high-precision dynamic model is subtracted from the far-fire point state obtained by the two-body dynamic model to obtain the far-fire point state correction amount. The initial state of the detector is used as input, and the correction of the near-fire state is used as output to form the data set for the near-fire state correction prediction network training. The initial state of the detector is used as input, and the correction of the far-fire state is used as output to form the data set for the far-fire state correction prediction network training. After completing the training of the near-fire state correction prediction network and the far-fire state correction prediction network, the near-fire state and the far-fire state obtained by two-body dynamics are corrected to construct the Mars aerodynamically assisted orbit reduction near-fire state intelligent prediction model and the far-fire state intelligent prediction model. The near-fire state intelligent prediction model is used to construct the near-fire heat flow constraint equation and the control amplitude constraint equation. At the same time, the far-fire state intelligent prediction model is used as the state prediction equation. The orbit height deviation and control input are used as optimization performance indicators to design a prediction guidance method. The current state of the detector is used as input to optimize the control quantity of the detector at the far-fire point. The detector is accurately lowered to the predetermined orbit under the conditions of satisfying the heat flow and control amplitude constraints, and the fuel consumption of the detector is reduced. This is the Mars aerodynamically assisted orbit reduction intelligent prediction guidance.

[0007] The Mars aerodynamic assisted orbit lowering intelligent prediction and guidance method disclosed in the present invention comprises the following steps:

[0008] Step 1: In the Martian inertial system, a high-precision dynamic model of the Mars aerodynamic-assisted orbit descent is constructed. In the subsequent step 2, the constructed high-precision dynamic model will be used to generate a data set for the neural network to learn.

[0009] In the Martian inertial system, a high-precision dynamic model of Mars aerodynamically assisted orbit reduction is constructed.

[0010]

[0011] Where r is the distance between the probe and the center of Mars, θ and φ represent the latitude and longitude of the probe in the Mars inertial system, V is the velocity of the probe, γ and ψ represent the track angle and heading angle of the probe, D and L represent the aerodynamic drag and aerodynamic lift acceleration, g, l and f represent the radial, longitude and latitude components of the Mars non-spherical gravity, respectively. The Mars full gradient gravity model

[0012]

[0013] The Martian atmosphere model is expressed using the Martian atmosphere correction index model as follows:

[0014]

[0015] Where h is the orbital altitude of the probe. Given the Martian atmospheric model, the aerodynamic drag and aerodynamic lift acceleration in the probe dynamics equation are expressed as

[0016]

[0017] Among them, C D and C L are the drag coefficient and lift coefficient of the probe respectively, S is the effective frontal area of ​​the probe, V is the speed of the probe, m is the mass of the probe, and ρ is the atmospheric density determined by the Martian atmosphere model.

[0018] Since the aerodynamically assisted orbit reduction process lasts for a long time, the influence of the irregular spherical gravitational perturbation of Mars needs to be considered when designing the prediction and guidance method. If a low-cost gravity model is used in the dynamic equation of the probe, the predicted trajectory will deviate greatly from the actual trajectory of the probe, and the accuracy requirements cannot be met; if a high-order gravity model is used, the calculation efficiency will be seriously reduced and it will be difficult to apply on orbit. In order to characterize the influence of the irregular spherical gravitational perturbation of Mars on the state of the probe in the high-precision dynamic model of the aerodynamically assisted orbit reduction of Mars, and at the same time improve the model accuracy and calculation efficiency, in step 2, the high-precision dynamic model of the aerodynamically assisted orbit reduction of Mars will be used to generate training data, and the mapping relationship between input and output will be learned using a neural network to build a fast and high-precision state intelligent prediction model.

[0019] Step 2: Use the high-precision dynamics model constructed in step 1 to generate the data set required for neural network training. In the process of generating the data set, define the value range of the initial state, sample according to the uniform distribution, and use the high-precision dynamics model of Mars aerodynamic assisted orbit reduction to recursively extrapolate the initial state to collect the peri-fire state and the apo-fire state at the next moment. Subtract the peri-fire state obtained by the high-precision dynamics model from the peri-fire state obtained by the two-body dynamics model to obtain the peri-fire state correction, and subtract the apo-fire state obtained by the high-precision dynamics model from the apo-fire state obtained by the two-body dynamics model to obtain the apo-fire state correction. The detector initial state is used as input, and the peri-fire state correction is used as output to form the data set for the training of the peri-fire state correction prediction network. The detector initial state is used as input, and the apo-fire state correction is used as output to form the data set for the training of the apo-fire state correction prediction network.

[0020] During the Mars aerodynamically assisted orbit reduction process, control is mainly applied at the farthest point, and the nearthest point state is required to meet the heat flow constraints. Therefore, when using neural networks, an intelligent prediction model of the nearthest point state predicted by the initial state and an intelligent prediction model of the farthest point state will be constructed. In step 2, the Mars aerodynamically assisted orbit reduction high-precision dynamic model constructed in step 1 will be used to collect the input as the initial state of the probe, and the output will be a data set of the next nearthest point state; the Mars aerodynamically assisted orbit reduction high-precision dynamic model constructed in step 1 will be used to collect the input as the initial state of the probe, and the output will be a data set of the next farthest point state.

[0021] The detector's near-fire point height, far-fire point height, and six orbital numbers are used to determine the value range of the detector's initial state, and h is defined p and h a are the perigee height and apogee height of the detector respectively. i, Ω, w, and f are defined as the orbital inclination, right ascension of ascending node, argument of perigee and true anomaly of the detector respectively. Then the initial state range of the detector is

[0022]

[0023] Among them, h p,min ,h a,min ,i min ,Ω min , w min , f min are the initial perigee height, apogee height, orbit inclination, right ascension of ascending node, argument of perigee and lower limit of true anomaly, h p,max ,h a,max ,i max ,Ω max , w max , f maxThey are respectively the detector's initial perigee altitude, apogee altitude, orbit inclination, ascending node right ascension, perigee argument and upper limit of true anomaly angle.

[0024] After determining the initial state range of the detector, sampling is performed in the range according to a uniform distribution, and the sampling results are converted into states represented by latitude, longitude and height. Define x0 = [r0, θ0, φ0, V0, γ0, ψ0] T The initial state of the selected detector is obtained by using high-precision dynamic integration to obtain the state of the detector near the fire point and the state of the detector far from the fire point.

[0025]

[0026] Where f represents the constructed high-precision dynamic model of Mars aerodynamic-assisted orbit reduction, t p and t a They respectively represent the time from the initial moment to the next near-fire point and the time from the initial moment to the next far-fire point.

[0027] The time t from the initial moment to the next fire point p and the time t from the initial moment to the next distant fire point a It is difficult to estimate accurately, but its orbital period is estimated by using the semi-major axis determined by the initial state of the probe.

[0028]

[0029] Where T is the orbital period of the probe under simple two-body dynamics, a0 is the semi-major axis of the orbit, μ m is the standard gravitational constant of Mars. After estimating the orbital period, the high-precision dynamics model of Mars aerodynamics-assisted orbit reduction is used to integrate the initial state of the probe for one cycle, and the data of the next perigee state and the next apogee state corresponding to the initial state are collected.

[0030] Considering the two-body dynamics, the next perigee and apogee states can be analytically solved according to the initial state of the probe. Under the high-precision dynamics of Mars aerodynamics-assisted orbit reduction, the mapping from the initial state to the next perigee state and from the initial state to the next apogee state can be divided into the analytical quantity obtained by the two-body dynamics and the correction quantity caused by non-spherical gravitational perturbation and atmospheric drag. Definition and In order to use the simplified two-body dynamics model to obtain the analytical part of the near-fire state and the far-fire state, the corrected part of the near-fire state δx p and the far-fire state δx a Expressed as

[0031]

[0032] According to the determined value range of the detector initial state, K samples are collected to construct a data set that maps the detector initial state to the near-fire point state correction value and the far-fire point state correction value.

[0033]

[0034] Among them, X is the initial state set of the detector, Y p and Y a They respectively represent the set of near-fire point state correction quantities and the set of far-fire point state correction quantities.

[0035] Step 3, construct a near-fire state correction prediction network, and use the data set constructed in step 2 that maps the detector's initial state to the near-fire state correction for network training. Construct a far-fire state correction prediction network, and use the data set constructed in step 2 that maps the detector's initial state to the far-fire state correction for network training. After completing the training of the near-fire state correction prediction network and the far-fire state correction prediction network, by correcting the near-fire state and the far-fire state obtained by two-body dynamics, construct a Mars aerodynamically assisted orbit lowering near-fire state intelligent prediction model and a far-fire state intelligent prediction model. Compared with the method of directly constructing a network to predict the near-fire state and the far-fire state, the method solves the near-fire and far-fire states under the two-body dynamics model, constructs a network to predict the deviation between the near-fire and far-fire states under the high-precision dynamics model and the two-body model prediction results, thereby reducing the training cost of the neural network and improving the prediction accuracy of the neural network.

[0036] Constructing a prediction network DNN for the near-fire state correction p , use the dataset constructed in step 2 that maps the detector initial state to the near-fire state correction to perform network training. Construct the far-fire state correction prediction network DNN a , use the data set constructed in step 2 that maps the initial state of the probe to the correction value of the far-fire state to perform network training. After completing the training of the near-fire state correction prediction network and the far-fire state correction prediction network, the near-fire state and far-fire state obtained by two-body dynamics are corrected to construct the Mars aerodynamic-assisted orbit-descent near-fire state intelligent prediction model and the far-fire state intelligent prediction model.

[0037]

[0038] Among them, x k represents the state of the flexible lander at time k, and The states near the fire point and far from the fire point at time k+1 under two-body dynamics, respectively, p,k+1 and x a,k+1The high-precision near-fire point status and far-fire point status are obtained by intelligent prediction.

[0039] Step 4. Use the near-fire point state intelligent prediction model to construct the near-fire point heat flow constraint equation and the control amplitude constraint equation. At the same time, use the far-fire point state intelligent prediction model as the state prediction equation, take the orbit height deviation and control input as the optimization performance indicators, design a predictive guidance method, use the current state of the probe as the input of the predictive guidance law, optimize and solve the control quantity of the probe at the far-fire point, and accurately lower the probe to the predetermined orbit while satisfying the heat flow and control amplitude constraints. At the same time, reduce the fuel consumption of the probe, that is, Mars aerodynamically assisted orbit descent intelligent predictive guidance.

[0040] During the Mars aerodynamic assisted orbit descent process, in order to protect the safety of the probe, it is necessary to limit the heat flux density of the probe at the near-fire point. According to the state x of the far-fire point of the probe at time k+i, a,k+i and the control input ΔV a,k+i The heat flow constraint equation near the fire point is constructed by the intelligent prediction model of the near-fire point state and is expressed as

[0041]

[0042] Among them, ρ p,k+i+1 is the atmospheric density near the fire point at time k+i+1, γ=[0,0,0,1,0,0] T , is the maximum allowable heat flux.

[0043] At the same time, there is an upper limit to the thrust that the detector can exert, and the velocity change at the far fire point is constrained by the control amplitude.

[0044] |ΔV a,k+i |≤ΔV max (12)

[0045] Where, ΔV max It is the upper limit of the velocity change at the far fire point.

[0046] When the probe is initially located at the farthest point, the Mars aerodynamics-assisted orbit reduction begins. The Mars aerodynamics-assisted orbit reduction prediction guidance is based on the current farthest point state x at each time k. a,k To predict the input of the guidance law, the remote firing point control input ΔV is obtained by solving the optimization problem a,k The optimization problem to be solved takes orbit height deviation and control input as optimization performance indicators, uses the intelligent prediction model of the far-fire state as the state prediction equation, and considers the heat flow constraint and control amplitude constraint at the same time.

[0047]

[0048] Among them, N is the number of prediction cycles, Rm is the radius of Mars, is the desired orbital height of the apogee. The apogee control input ΔV within the predicted number of revolutions a,k+i After optimization, select ΔV a,k As the control input at time k.

[0049] When the probe is at the aphelion, the control input is optimized to control the probe until its orbital altitude is lowered to near the predetermined orbital altitude, so that the probe can be accurately lowered to the predetermined orbit while satisfying the constraints of heat flow and control amplitude. At the same time, the fuel consumption of the probe is reduced, that is, the Mars aerodynamic-assisted orbit descent intelligent predictive guidance.

[0050] Beneficial effects:

[0051] 1. The present invention discloses an intelligent prediction and guidance method for aerodynamically assisted orbit reduction on Mars, constructs a high-precision dynamic model for aerodynamically assisted orbit reduction on Mars, generates a network training data set using the high-precision model, constructs an intelligent prediction model for the near-fire point state and an intelligent prediction model for the far-fire point state of aerodynamically assisted orbit reduction on Mars, and constructs a near-fire point heat flow constraint at the same time, and designs a prediction and guidance method with orbit height deviation and control input as optimization performance indicators; the control input is optimized and solved by the prediction and guidance method to control the probe until the orbit height of the probe is reduced to near the predetermined orbit height, so as to achieve the precise reduction of the probe to the predetermined orbit while satisfying the heat flow and control amplitude constraints, and at the same time reduce the fuel consumption of the probe, that is, intelligent prediction and guidance for aerodynamically assisted orbit reduction on Mars.

[0052] 2. Compared with the method of directly constructing a network to predict the state of the near-fire point and the far-fire point, the Mars aerodynamically assisted orbit reduction intelligent prediction and guidance method disclosed in the present invention solves the state of the near-fire point and the far-fire point under the two-body dynamic model, and constructs a network to predict the deviation between the state of the near-fire point and the far-fire point under the high-precision dynamic model and the prediction result of the two-body model, thereby reducing the training cost of the neural network and improving the prediction accuracy of the constructed Mars aerodynamically assisted orbit reduction intelligent prediction model for the near-fire point state and the far-fire point state. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of the intelligent prediction and guidance method for Mars aerodynamic assisted orbit reduction;

[0054] Figure 2 This is a schematic diagram of the training effect of the near-fire state correction prediction network;

[0055] Figure 3 This is a schematic diagram of the training effect of the network for predicting the correction value of the far-fire point state;

[0056] Figure 4 The far-fire maneuver curve provided for the predictive guidance method. DETAILED DESCRIPTION

[0057] In order to better illustrate the purpose and advantages of the present invention, the invention is further described below in conjunction with an example and corresponding drawings.

[0058] like Figure 1 As shown, the Mars aerodynamic assisted orbit lowering intelligent prediction and guidance method disclosed in this embodiment has the following specific implementation steps:

[0059] Step 1: In the Martian inertial system, a high-precision dynamic model of the Mars aerodynamic-assisted orbit descent is constructed. In the subsequent step 2, the constructed high-precision dynamic model will be used to generate a data set for the neural network to learn.

[0060] In the Martian inertial system, a high-precision dynamic model of Mars aerodynamically assisted orbit reduction is constructed.

[0061]

[0062] Where r is the distance between the probe and the center of Mars, θ and φ represent the latitude and longitude of the probe in the Mars inertial system, V is the velocity of the probe, γ and ψ represent the track angle and heading angle of the probe, D and L represent the aerodynamic drag and aerodynamic lift acceleration, g, l and f represent the radial, longitude and latitude components of the Mars non-spherical gravity, respectively. The Mars full gradient gravity model

[0063]

[0064] The Martian atmosphere model is expressed using the Martian atmosphere correction index model as follows:

[0065]

[0066] Where h is the orbital altitude of the probe. Given the Martian atmospheric model, the aerodynamic drag and aerodynamic lift acceleration in the probe dynamics equation are expressed as

[0067]

[0068] Among them, C D =2.2 and C L =0.31 are the drag coefficient and lift coefficient of the probe respectively, S=36 is the effective frontal area of ​​the probe, V is the speed of the probe, m=1225 is the mass of the probe, and ρ is the atmospheric density determined by the Martian atmosphere model.

[0069] Since the aerodynamically assisted orbit reduction process lasts for a long time, the influence of the irregular spherical gravitational perturbation of Mars needs to be considered when designing the prediction and guidance method. If a low-cost gravity model is used in the dynamic equation of the probe, the predicted trajectory will deviate greatly from the actual trajectory of the probe, and the accuracy requirements cannot be met; if a high-order gravity model is used, the calculation efficiency will be seriously reduced and it will be difficult to apply on orbit. In order to characterize the influence of the irregular spherical gravitational perturbation of Mars on the state of the probe in the high-precision dynamic model of the aerodynamically assisted orbit reduction of Mars, and at the same time improve the model accuracy and calculation efficiency, in step 2, the high-precision dynamic model of the aerodynamically assisted orbit reduction of Mars will be used to generate training data, and the mapping relationship between input and output will be learned using a neural network to build a fast and high-precision state intelligent prediction model.

[0070] Step 2: Use the high-precision dynamics model constructed in step 1 to generate the data set required for neural network training. In the process of generating the data set, define the value range of the initial state, sample according to the uniform distribution, and use the high-precision dynamics model of Mars aerodynamic assisted orbit reduction to recursively extrapolate the initial state to collect the peri-fire state and the apo-fire state at the next moment. Subtract the peri-fire state obtained by the high-precision dynamics model from the peri-fire state obtained by the two-body dynamics model to obtain the peri-fire state correction, and subtract the apo-fire state obtained by the high-precision dynamics model from the apo-fire state obtained by the two-body dynamics model to obtain the apo-fire state correction. The detector initial state is used as input, and the peri-fire state correction is used as output to form the data set for the training of the peri-fire state correction prediction network. The detector initial state is used as input, and the apo-fire state correction is used as output to form the data set for the training of the apo-fire state correction prediction network.

[0071] During the Mars aerodynamically assisted orbit reduction process, control is mainly applied at the farthest point, and the nearthest point state is required to meet the heat flow constraints. Therefore, when using neural networks, an intelligent prediction model of the nearthest point state predicted by the initial state and an intelligent prediction model of the farthest point state will be constructed. In step 2, the Mars aerodynamically assisted orbit reduction high-precision dynamic model constructed in step 1 will be used to collect the input as the initial state of the probe, and the output will be a data set of the next nearthest point state; the Mars aerodynamically assisted orbit reduction high-precision dynamic model constructed in step 1 will be used to collect the input as the initial state of the probe, and the output will be a data set of the next farthest point state.

[0072] The detector's near-fire point height, far-fire point height, and six orbital numbers are used to determine the value range of the detector's initial state, and h is defined p and h a are the perigee height and apogee height of the detector respectively. i, Ω, w, and f are defined as the orbital inclination, right ascension of ascending node, argument of perigee and true anomaly of the detector respectively. Then the initial state range of the detector is

[0073]

[0074] Among them, h p,min =8×10 4 m, h a,min =4.5×10 5 m,i min =89deg,Ω min =0deg,w min =0deg, f min =180deg are respectively the initial perigee height, apogee height, orbit inclination, right ascension of ascending node, argument of perigee and lower limit of true anomaly angle of the detector, h p,max =1.25×10 5 m, h a,max =1×10 6 m,i max =93deg,Ω max =360deg, w max =360deg, f max =180deg are the detector's initial perigee height, apogee height, orbit inclination, ascending node right ascension, perigee argument and upper limit of true anomaly angle.

[0075] After determining the initial state range of the detector, sampling is performed in the range according to a uniform distribution, and the sampling results are converted into states represented by latitude, longitude and height. Define x0 = [r0, θ0, φ0, V0, γ0, ψ0] T The initial state of the selected detector is obtained by using high-precision dynamic integration to obtain the state of the detector near the fire point and the state of the detector far from the fire point.

[0076]

[0077] Where f represents the constructed high-precision dynamic model of Mars aerodynamic-assisted orbit reduction, t p and t a They respectively represent the time from the initial moment to the next near-fire point and the time from the initial moment to the next far-fire point.

[0078] The time t from the initial moment to the next fire point p and the time t from the initial moment to the next distant fire point a It is difficult to estimate accurately, but its orbital period is estimated by using the semi-major axis determined by the initial state of the probe.

[0079]

[0080] Where T is the orbital period of the probe under simple two-body dynamics, a0 is the semi-major axis of the orbit, μm is the standard gravitational constant of Mars. After estimating the orbital period, the high-precision dynamics model of Mars aerodynamics-assisted orbit reduction is used to integrate the initial state of the probe for one cycle, and the data of the next perigee state and the next apogee state corresponding to the initial state are collected.

[0081] Considering the two-body dynamics, the next perigee and apogee states can be analytically solved according to the initial state of the probe. Under the high-precision dynamics of Mars aerodynamics-assisted orbit reduction, the mapping from the initial state to the next perigee state and from the initial state to the next apogee state can be divided into the analytical quantity obtained by the two-body dynamics and the correction quantity caused by non-spherical gravitational perturbation and atmospheric drag. Definition and In order to use the simplified two-body dynamics model to obtain the analytical part of the near-fire state and the far-fire state, the corrected part of the near-fire state δx p and the far-fire state δx a It can be expressed as

[0082]

[0083] According to the determined value range of the detector initial state, K = 2 × 10 5 samples, constructing a data set that maps the detector's initial state to the near-fire point state correction and the far-fire point state correction.

[0084]

[0085] Among them, X is the initial state set of the detector, Y p and Y a They respectively represent the set of near-fire point state correction quantities and the set of far-fire point state correction quantities.

[0086] Step 3, construct a near-fire state correction prediction network, and use the data set constructed in step 2 that maps the detector's initial state to the near-fire state correction for network training. Construct a far-fire state correction prediction network, and use the data set constructed in step 2 that maps the detector's initial state to the far-fire state correction for network training. After completing the training of the near-fire state correction prediction network and the far-fire state correction prediction network, by correcting the near-fire state and the far-fire state obtained by two-body dynamics, construct a Mars aerodynamically assisted orbit lowering near-fire state intelligent prediction model and a far-fire state intelligent prediction model. Compared with the method of directly constructing a network to predict the near-fire state and the far-fire state, the method solves the near-fire and far-fire states under the two-body dynamics model, constructs a network to predict the near-fire and far-fire states under a high-precision dynamics model, and the deviation between the two-body model prediction results reduces the training cost of the neural network and improves the prediction accuracy of the neural network.

[0087] Constructing a prediction network DNN for the near-fire state correction p , use the dataset constructed in step 2 that maps the detector initial state to the near-fire state correction to perform network training. Construct the far-fire state correction prediction network DNN a , use the data set constructed in step 2 that maps the initial state of the probe to the correction value of the far-fire state to perform network training. After completing the training of the near-fire state correction prediction network and the far-fire state correction prediction network, the near-fire state and far-fire state obtained by two-body dynamics are corrected to construct the Mars aerodynamic-assisted orbit-descent near-fire state intelligent prediction model and the far-fire state intelligent prediction model.

[0088]

[0089] Among them, x k represents the state of the flexible lander at time k, and The states near the fire point and far from the fire point at time k+1 under two-body dynamics, respectively, p,k+1 and x a,k+1 The high-precision near-fire point state and far-fire point state obtained by intelligent prediction. The constructed network has 4 hidden layers, each with 32 neurons, and the activation function is tanh. Figure 2 and Figure 3 The training results of the near-fire point state correction prediction network and the far-fire point state correction prediction network are shown respectively.

[0090] Step 5. Use the near-fire point state intelligent prediction model to construct the near-fire point heat flow constraint equation and the control amplitude constraint equation. At the same time, use the far-fire point state intelligent prediction model as the state prediction equation, take the orbit height deviation and control input as the optimization performance indicators, design a predictive guidance method, use the current state of the probe as the input of the predictive guidance law, optimize and solve the control quantity of the probe at the far-fire point, and accurately lower the probe to the predetermined orbit while satisfying the heat flow and control amplitude constraints. At the same time, reduce the fuel consumption of the probe, that is, Mars aerodynamically assisted orbit descent intelligent predictive guidance.

[0091] During the Mars aerodynamic assisted orbit descent process, in order to protect the safety of the probe, it is necessary to limit the heat flux density of the probe at the near-fire point. According to the state x of the far-fire point of the probe at time k+i, a,k+i and the control input ΔV a,k+i The heat flow constraint equation near the fire point is constructed by the intelligent prediction model of the near-fire point state and is expressed as

[0092]

[0093] Among them, ρ p,k+i+1is the atmospheric density near the fire point at time k+i+1, γ=[0,0,0,1,0,0] T , is the maximum allowable heat flux.

[0094] At the same time, there is an upper limit to the thrust that the detector can exert, and the velocity change at the far fire point is constrained.

[0095] |ΔV a,k+i |≤ΔV max (25)

[0096] Where, ΔV max =50m / s is the upper limit of the velocity change at the far fire point.

[0097] When the probe is initially located at the farthest point, the Mars aerodynamics-assisted orbit reduction begins. The Mars aerodynamics-assisted orbit reduction prediction guidance is based on the current farthest point state x at each time k. a,k To predict the input of the guidance law, the remote firing point control input ΔV is obtained by solving the optimization problem a,k The optimization problem to be solved takes orbit height deviation and control input as optimization performance indicators, uses the intelligent prediction model of the far-fire state as the state prediction equation, and considers the heat flow constraint and control amplitude constraint at the same time.

[0098]

[0099] Among them, N = 20 is the number of prediction cycles, R m =3.397×10 6 is the radius of Mars, is the desired orbital height of the apogee. The apogee control input ΔV within the predicted number of revolutions a,k+i After optimization, select ΔV a,k As the control input at time k.

[0100] When the detector is at the far fire point, the detector is controlled by optimizing the control input. The optimization result is as follows: Figure 4 As shown, until the probe's orbital altitude is lowered to near the predetermined orbital altitude, the probe can be accurately lowered to the predetermined orbit while satisfying the constraints of heat flow and control amplitude, while reducing the probe's fuel consumption, that is, Mars aerodynamic-assisted orbit descent intelligent predictive guidance.

[0101] The specific description above further illustrates the purpose, technical solutions and beneficial effects of the invention in detail. It should be understood that the above is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent prediction and guidance method for Mars aerodynamically assisted orbit lowering, characterized in that: The steps include: Step 1: In the Mars inertial system, a high-precision dynamic model of Mars aerodynamic-assisted orbit reduction is constructed. In the subsequent step 2, the constructed high-precision dynamic model will be used to generate a data set for neural network learning; Step 2, using the high-precision dynamics model constructed in step 1 to generate the data set required for neural network training; in the process of generating the data set, define the value range of the initial state, sample according to the uniform distribution, use the high-precision dynamics model of Mars aerodynamic assisted orbit reduction to recursively extrapolate the initial state to collect the perigee state and the apogee state at the next moment, subtract the perigee state obtained by the high-precision dynamics model from the perigee state obtained by the two-body dynamics model to obtain a correction for the perigee state, and subtract the apogee state obtained by the high-precision dynamics model from the apogee state obtained by the two-body dynamics model to obtain a correction for the apogee state; The initial state of the detector is used as input, and the correction value of the near-fire state is used as output to form a data set for the near-fire state correction value prediction network training; the initial state of the detector is used as input, and the correction value of the far-fire state is used as output to form a data set for the far-fire state correction value prediction network training; Step 3, construct a near-fire state correction prediction network, and use the data set constructed in step 2 that maps the detector initial state to the near-fire state correction for network training; construct a far-fire state correction prediction network, and use the data set constructed in step 2 that maps the detector initial state to the far-fire state correction for network training; after completing the training of the near-fire state correction prediction network and the far-fire state correction prediction network, correct the near-fire state and the far-fire state obtained by two-body dynamics to construct a Mars aerodynamically assisted orbit reduction near-fire state intelligent prediction model and a far-fire state intelligent prediction model; the method solves the near-fire and far-fire states under the two-body dynamics model, constructs a network to predict the deviation between the near-fire state and the far-fire state under the high-precision dynamics model and the two-body model prediction results, thereby reducing the training cost of the neural network and improving the prediction accuracy of the neural network; Among them, x k represents the state of the flexible lander at time k, and The states near the fire point and far from the fire point at time k+1 under two-body dynamics, respectively, p,k+1 and x a,k+1 High-precision near-fire point status and far-fire point status obtained by intelligent prediction; Step 4. Use the near-fire point state intelligent prediction model to construct the near-fire point heat flow constraint equation and the control amplitude constraint equation. At the same time, use the far-fire point state intelligent prediction model as the state prediction equation, take the orbit height deviation and control input as the optimization performance indicators, design a predictive guidance method, use the current state of the probe as the predictive guidance law input, optimize and solve the control quantity of the probe at the far-fire point, and accurately lower the probe to the predetermined orbit while satisfying the heat flow and control amplitude constraints. At the same time, reduce the fuel consumption of the probe, that is, Mars aerodynamically assisted orbit descent intelligent predictive guidance.

2. The Mars aerodynamic assisted orbit reduction intelligent prediction and guidance method according to claim 1, characterized in that: In step 1, In the Martian inertial system, a high-precision dynamic model of Mars aerodynamically assisted orbit reduction is constructed. Where r is the distance between the probe and the center of Mars, θ and φ represent the latitude and longitude of the probe in the Mars inertial system, V is the velocity of the probe, γ and ψ represent the track angle and heading angle of the probe, D and L represent the aerodynamic drag and aerodynamic lift acceleration, g, l and f represent the radial, longitude and latitude components of the Mars non-spherical gravity, respectively. The Mars full gradient gravity model The Martian atmosphere model is expressed using the Martian atmosphere correction index model as follows: Where h represents the orbital altitude of the probe. Given the Martian atmospheric model, the aerodynamic drag and aerodynamic lift acceleration in the probe dynamics equation are expressed as Among them, C D and C L are the drag coefficient and lift coefficient of the probe respectively, S is the effective frontal area of ​​the probe, V is the speed of the probe, m is the mass of the probe, and ρ is the atmospheric density determined by the Martian atmosphere model.

3. The Mars aerodynamic assisted orbit reduction intelligent prediction and guidance method according to claim 2, characterized in that: Step 2 is implemented as follows: During the Mars aerodynamically assisted orbit reduction process, control is mainly applied at the far-fire point, and the near-fire point state is required to meet the heat flow constraint; therefore, when using the neural network, an intelligent prediction model of the near-fire point state predicted by the initial state and an intelligent prediction model of the far-fire point state will be constructed; in step 2, the Mars aerodynamically assisted orbit reduction high-precision dynamics model constructed in step 1 will be used to collect the input as the initial state of the probe, and the output will be a data set of the next near-fire point state; the Mars aerodynamically assisted orbit reduction high-precision dynamics model constructed in step 1 will be used to collect the input as the initial state of the probe, and the output will be a data set of the next far-fire point state; The detector's near-fire point height, far-fire point height, and six orbital numbers are used to determine the value range of the detector's initial state, and h is defined p and h a are the perigee height and apogee height of the detector respectively. i, Ω, w, and f are defined as the orbital inclination, right ascension of ascending node, argument of perigee and true anomaly of the detector respectively. Then the initial state range of the detector is Among them, h p,min ,h a,min ,i min ,Ω min , w min , f min are the initial perigee height, apogee height, orbit inclination, right ascension of ascending node, argument of perigee and lower limit of true anomaly, h p,max ,h a,max ,i max ,Ω max , w max , f max They are the detector's initial perigee height, apogee height, orbit inclination, ascending node right ascension, perigee argument and true anomaly upper limit; After determining the initial state range of the detector, sampling is performed in the range according to a uniform distribution, and the sampling results are converted into states represented by latitude, longitude and height. Define x0 = [r0, θ0, φ0, V0, γ0, ψ0] T The initial state of the selected detector is obtained by using high-precision dynamic integration to obtain the state of the detector near the fire point and the state of the detector far from the fire point. Where f represents the constructed high-precision dynamic model of Mars aerodynamic-assisted orbit reduction, t p and t a They represent the time from the initial moment to the next near-fire point and the time from the initial moment to the next far-fire point respectively; The time t from the initial moment to the next fire point p and the time t from the initial moment to the next distant fire point a It is difficult to estimate accurately, but its orbital period is estimated by using the semi-major axis determined by the initial state of the probe. Where T is the orbital period of the probe under simple two-body dynamics, a0 is the semi-major axis of the orbit, μ m is the standard gravitational constant of Mars. After estimating the orbital period, the high-precision dynamics model of Mars aerodynamics-assisted orbit reduction is used to integrate the initial state of the probe for one week, and the data of the next perigee state and the next apogee state corresponding to the initial state are collected. Considering that under two-body dynamics, the next perigee and apogee states can be analytically solved according to the initial state of the probe; under the high-precision dynamics of Mars aerodynamic-assisted orbit reduction, the mapping from the initial state to the next perigee state and from the initial state to the next apogee state can be divided into the analytical quantity obtained by two-body dynamics and the correction quantity caused by non-spherical gravitational perturbation and atmospheric drag; define and In order to use the simplified two-body dynamics model to obtain the analytical part of the near-fire state and the far-fire state, the corrected part of the near-fire state δx p and the far-fire state δx a Expressed as According to the determined value range of the detector initial state, K samples are collected to construct a data set that maps the detector initial state to the near-fire point state correction value and the far-fire point state correction value. Among them, X is the initial state set of the detector, Y p and Y a They respectively represent the set of near-fire point state correction quantities and the set of far-fire point state correction quantities.

4. The Mars aerodynamic assisted orbit reduction intelligent prediction and guidance method according to claim 3, characterized in that: In step 3, Constructing a prediction network DNN for the near-fire state correction p , use the data set constructed in step 2 that maps the detector initial state to the near-fire state correction to perform network training; construct a far-fire state correction prediction network DNN a , use the data set constructed in step 2 that maps the initial state of the probe to the correction amount of the far-fire state to perform network training; after completing the training of the near-fire state correction amount prediction network and the far-fire state correction amount prediction network, correct the near-fire state and the far-fire state obtained by two-body dynamics to construct the Mars aerodynamic-assisted orbit descent near-fire state intelligent prediction model and the far-fire state intelligent prediction model 5. The Mars aerodynamic assisted orbit reduction intelligent prediction and guidance method according to claim 4, characterized in that: Step 4 is implemented as follows: During the Mars aerodynamic assisted orbit descent process, in order to protect the safety of the probe, it is necessary to limit the heat flux density of the probe at the near-fire point. According to the state x of the far-fire point of the probe at time k+i, a,k+i and the control input ΔV a,k+i The heat flow constraint equation near the fire point is constructed by the intelligent prediction model of the near-fire point state and is expressed as Among them, ρ p,k+i+1 is the atmospheric density near the fire point at time k+i+1, γ=[0,0,0,1,0,0] T , Q max is the maximum allowable heat flux density; At the same time, there is an upper limit to the thrust that the detector can exert, and the velocity change at the far fire point is constrained by the control amplitude. |ΔV a,k+i |≤ΔV max (12) Where, ΔV max is the upper limit of the velocity variation at the far fire point; When the probe is initially located at the farthest point, the Mars aerodynamics-assisted orbit descent begins; the Mars aerodynamics-assisted orbit descent prediction guidance is based on the current farthest point state x at each time k. a,k To predict the input of the guidance law, the remote firing point control input ΔV is obtained by solving the optimization problem a,k The optimization problem to be solved takes orbit height deviation and control input as optimization performance indicators, uses the intelligent prediction model of the far-fire state as the state prediction equation, and considers heat flow constraints and control amplitude constraints at the same time. Among them, N is the number of prediction cycles, R m is the radius of Mars, is the desired apogee orbit height; the apogee control input ΔV within the predicted number of laps a,k+i After optimization, select ΔV a,k As the control input at time k; When the probe is at the aphelion, the control input is optimized to control the probe until its orbital altitude is lowered to near the predetermined orbital altitude, so that the probe can be accurately lowered to the predetermined orbit while satisfying the constraints of heat flow and control amplitude. At the same time, the fuel consumption of the probe is reduced, that is, the Mars aerodynamic-assisted orbit descent intelligent predictive guidance.

Images