Mars probe landing phase navigation method and system
Optimizing the landmark distribution through Fisher information matrix and greedy algorithm has solved the problem of insufficient navigation accuracy in the landing section of the Mars rover, and achieved high-precision autonomous navigation and efficient resource utilization.
Patent Information
- Application Number
- CN202411138607.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-08-19
AI Technical Summary
The existing technology cannot achieve the highest navigation accuracy in the landing section of the Mars rover, mainly due to the characteristics of landmark distribution and limited space-based storage and computing resources.
The observability matrix of the navigation system is constructed using Fisher information matrix, and the observability of the landmark is described in combination with the CR lower boundary. The preferred landmark is adaptively determined through the greedy algorithm, which maximizes the traces of the observable matrix, optimizes the landmark distribution configuration, and reduces calculations. Resource load to improve navigation accuracy.
It realizes high-precision autonomous navigation during the landing of the Mars rover, overcomes the problems of redundancy of computing resources and low navigation accuracy caused by excessive land marks, dynamically obtains the optimal land marks, and improves the landing accuracy.
Smart Images

Figure CN119085664A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of planetary probe technology, and in particular to a navigation method and system for a landing phase of a Mars probe. Background Art
[0002] Autonomous precision landing is the key to the Mars exploration mission. Its primary premise is that the Mars probe's lander can autonomously obtain high-precision position, velocity, attitude and other state information, that is, autonomous navigation, so as to move to the predetermined landing point. The commonly used inertial navigation is characterized by the long time and long distance of the Mars exploration mission. Its cumulative error is the main error source affecting the accuracy of inertial navigation. In order to achieve precise landing, the error must be accurately estimated and compensated during the powered descent. During the Mars orbit exploration process, the Mars probe can obtain the precise position information of the landmarks on the surface of Mars, which provides an effective measurement information source for autonomous navigation in the landing phase. Therefore, combined navigation based on landmark information has become a hot research direction for autonomous navigation in the landing phase. Generally speaking, the more landmark information there is, the higher the accuracy of the state estimation of the autonomous navigation system. However, in deep space exploration, the onboard storage and computing resources are very limited, and it is impossible to store and solve a large number of landmarks.
[0003] In order to ensure navigation accuracy on a probe with limited onboard storage and computing resources, the number of observed landmarks should be reduced as much as possible. Therefore, it is necessary to select the landmarks that contribute most to navigation accuracy for observation.
[0004] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art:
[0005] There are many methods for optimizing landmarks, but due to the characteristics of the landmark distribution during the landing phase of the Mars rover, all known solutions cannot achieve the highest navigation accuracy at each stage. Therefore, how to improve the landing accuracy of the Mars rover is a problem that needs to be solved. Summary of the invention
[0006] An embodiment of the present invention provides a Mars rover landing stage navigation method, which takes into account the characteristics of landmark distribution and can effectively improve the landing accuracy of the Mars rover.
[0007] To achieve the above-mentioned purpose, on the one hand, an embodiment of the present invention provides a Mars rover landing segment navigation method, including: establishing a measurement model of visible landmarks based on visible landmarks at the current moment; obtaining the Fisher information matrix of each visible landmark according to the measurement model; calculating the trace of the Fisher information matrix corresponding to each visible landmark; when the number of visible landmarks is not less than 3, constructing a preferred configuration of landmark distribution, and the constraint conditions of the preferred configuration of landmark distribution are: selecting multiple visible landmarks as preferred landmarks, and maximizing the sum of the traces of the preferred landmarks while minimizing the number of preferred landmarks; solving the preferred configuration using a greedy algorithm to determine the preferred landmark; and performing landing navigation on the Mars rover through the preferred landmarks.
[0008] On the other hand, an embodiment of the present invention provides a Mars rover landing segment navigation system, including: a Fisher information matrix construction module, used to establish a measurement model of visible landmarks based on visible landmarks at a current moment; obtain the Fisher information matrix of each visible landmark according to the measurement model; calculate the trace of the Fisher information matrix corresponding to each visible landmark; a preferred configuration construction module, used to construct a preferred configuration of landmark distribution when the number of visible landmarks is not less than 3, and the constraint conditions of the preferred configuration of landmark distribution are: select multiple visible landmarks as preferred landmarks, and maximize the sum of the traces of the preferred landmarks while minimizing the number of preferred landmarks; a preferred landmark acquisition module, used to solve the preferred configuration using a greedy algorithm to determine the preferred landmark; a first navigation module, used to perform landing navigation on the Mars rover through the preferred landmarks.
[0009] The above technical solution has the following beneficial effects:
[0010] The technical solution of the present application uses the Fisher information matrix to construct the observability matrix of the navigation system, combines the CR lower bound to describe the observability of the landmarks during the Mars landing process, and maximizes the scalar function of the observable matrix to obtain the configuration of the optimal landmark distribution. At the same time, the greedy algorithm is introduced to adaptively determine the number of preferred landmarks. Compared with other algorithms, the greedy algorithm has a significant advantage in computational complexity. It can be achieved by adding and subtracting only a few vectors, and can better cope with the problem of severe limitation of onboard storage and computing resources in deep space exploration missions. However, the traditional greedy algorithm is prone to fall into local optimality. In order to, taking into account the characteristics of the reduction in the number of visible landmarks during the landing of the probe, the convergence threshold in the greedy algorithm is set to a floating value that changes with time, so that the solution result is as close to the global optimal as possible, and the optimal number of landmarks at each moment is dynamically obtained to achieve high-precision autonomous navigation, and overcome the problem of redundant onboard computing resources caused by too many landmarks and the problem of low navigation accuracy caused by insufficient number of landmarks. BRIEF DESCRIPTION OF THE DRAWINGS
[0011] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0012] Figure 1 It is a flow chart of a Mars probe landing stage navigation method of the present invention;
[0013] Figure 2 It is a composition diagram of a navigation system for the landing stage of a Mars probe of the present invention;
[0014] Figure 3 is a schematic diagram of the coordinate system used by the navigation system of the present invention;
[0015] Figure 4 is a schematic diagram of the relationship between the positions of the Mars probe and the landmarks in the present invention;
[0016] Figure 5 It is a schematic diagram of the landing track of the probe and the randomly distributed landmark images in a specific embodiment of the present invention;
[0017] Figure 6 is a schematic diagram of the change in the number of visible landmarks during the descent of the detector in a specific embodiment of the present invention;
[0018] Figure 7 A schematic diagram of the number of preferred landmarks at different times in a specific embodiment of the present invention;
[0019] Figure 8 is a comparison chart of the position error RMSE of the preferred landmark and the randomly selected landmark on each axis in a specific embodiment of the present invention;
[0020] Fig. 9 is a comparison diagram of the RMSE of the attitude errors of the preferred landmarks and the randomly selected landmarks on each axis in a specific embodiment of the present invention;
[0021] Fig.10 is a comparison diagram of the RMSE of the speed error of the preferred landmark and the randomly selected landmark on each axis in a specific embodiment of the present invention;
[0022] Fig.11 It is a plan distribution diagram of some preferred landmarks in a specific embodiment of the present invention. DETAILED DESCRIPTION
[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0024] like Figure 1 As shown, an embodiment of the present invention provides a Mars probe landing stage navigation method, characterized in that it includes:
[0025] S101, based on the visible landmarks at the current moment, establish a measurement model of the visible landmarks (refer to formula 16);
[0026] S102, obtaining the Fisher information matrix (i.e., observability matrix, see Formula 23) of each visible landmark according to the measurement model;
[0027] S103, calculating the trace of the Fisher information matrix corresponding to each visible landmark (refer to formula 40);
[0028] S104. When the number of the visible landmarks is not less than 3, construct a preferred configuration of the landmark distribution, wherein the constraint condition of the preferred configuration of the landmark distribution is: selecting a plurality of the visible landmarks as preferred landmarks, and maximizing the sum of the traces of the preferred landmarks while minimizing the number of the preferred landmarks (see Formula 37);
[0029] S105, using a greedy algorithm to solve the preferred configuration (refer to the subsequent formula 38 and algorithm 1), and determining the preferred landmark;
[0030] S106. Perform landing navigation for the Mars probe through the preferred landmarks.
[0031] In order to improve the landing navigation accuracy, in this application, the Fisher information is used to construct the observability matrix of the navigation system, and the scalar function of the observable matrix is maximized to obtain the optimal configuration (preferred configuration) of the landmark distribution. On the basis of determining the optimal configuration, the greedy algorithm is combined to adaptively determine the number of landmarks selected at each moment. Considering that the number of observable landmarks gradually decreases during the landing process, the convergence threshold of the greedy algorithm is set to a dynamic value related to the landing time to obtain the optimal number of landmarks.
[0032] Furthermore, the Mars probe landing stage navigation method also includes:
[0033] S107, when the number of the visible landmarks is greater than zero and less than 3, taking all the visible landmarks as the preferred landmarks;
[0034] S108: When the number of the visible landmarks is zero, perform landing navigation by inertial navigation.
[0035] When the number of observable landmarks is less than 3, there is no need to select the landmarks optimally, and all observable landmarks can be used for pose calculation. In the late stage of landing, when the number of visible landmarks drops to zero, the landmark-assisted inertial navigation system is transformed into a pure inertial navigation system. Since the accumulated error of inertial navigation cannot be effectively corrected in the late stage of landing, the navigation accuracy diverges in the late stage of landing but remains within a tolerance range.
[0036] Furthermore, in step S105, the convergence threshold of the greedy algorithm is a variable value.
[0037] Furthermore, the convergence threshold is in negative proportion to the landing time, and the convergence threshold is in positive proportion to the minimum value of the traces of all visible landmarks; wherein the landing time refers to the time elapsed from the start of the landing process to the current moment.
[0038] At the beginning of the landing process, there are many observable landmarks, and there are a large number of high-quality landmarks that contribute greatly to navigation accuracy. At this time, only a small number of landmarks can be selected from the observable landmarks to achieve a high navigation accuracy. Therefore, a larger convergence threshold should be selected at the beginning of landing to avoid selecting too many landmarks and generating computing resource load and precision redundancy. As the height of the Mars rover decreases, the closer it is to the landing point, the fewer landmarks are visible, and the lower the contribution of landmarks to navigation accuracy. If a larger number is still selected when the number of observable landmarks decreases, the number of preferred landmarks may be too small, resulting in insufficient navigation accuracy. Therefore, in the process of solving with the greedy algorithm, the convergence threshold should be a decreasing function of the landing time, constructed with an inverse proportional function as shown in Formula 39.
[0039] Furthermore, the step S106 further includes:
[0040] S1061. Establish a dynamic model of the landing process of the Mars probe (see formula 13);
[0041] S1062, linearizing the dynamic model through an extended Kalman filter (EKF) to obtain a linearized state equation;
[0042] S1063, performing landing navigation on the Mars probe in a combined navigation manner according to the linearized state equation and the preferred landmark, wherein the combined navigation refers to a navigation manner combining inertial navigation with landmark navigation;
[0043] The step S102 specifically includes:
[0044] S1021, converting the measurement model into a measurement equation;
[0045] S1022, linearizing the measurement equation by using an extended Kalman filter to obtain a measurement matrix;
[0046] S1023. Obtain a Fisher information matrix of each visible landmark according to the measurement matrix.
[0047] like Figure 2 As shown, an embodiment of the present invention provides a navigation system for the landing stage of a Mars probe, characterized in that it includes:
[0048] Fisher information matrix construction module 21, used to establish a measurement model of visible landmarks based on the visible landmarks at the current moment; obtain the Fisher information matrix of each visible landmark according to the measurement model; calculate the trace of the Fisher information matrix corresponding to each visible landmark;
[0049] A preferred configuration building module 22 is used to build a preferred configuration of the landmark distribution when the number of the visible landmarks is not less than 3, wherein the constraint condition of the preferred configuration of the landmark distribution is: selecting a plurality of the visible landmarks as preferred landmarks, and maximizing the sum of the traces of the preferred landmarks while minimizing the number of the preferred landmarks;
[0050] A preferred landmark acquisition module 23, used for solving the preferred configuration using a greedy algorithm to determine the preferred landmark;
[0051] The first navigation module 24 is used to perform landing navigation for the Mars probe through the preferred landmarks.
[0052] Furthermore, the Mars rover landing stage navigation system also includes:
[0053] A second preferred landmark acquisition module 25, configured to use all the visible landmarks as the preferred landmarks when the number of the visible landmarks is greater than zero and less than 3;
[0054] The second navigation module 26 is used to perform landing navigation in an inertial navigation manner when the number of the visible landmarks is zero.
[0055] Furthermore, in the preferred landmark acquisition module 23, the convergence threshold adopted by the greedy algorithm is a variable value.
[0056] Furthermore, the convergence threshold is in negative proportion to the landing time, and the convergence threshold is in positive proportion to the minimum value of the traces of all visible landmarks; wherein the landing time refers to the time elapsed from the start of the landing process to the current moment.
[0057] Further, the first navigation module 24 is specifically used to establish a dynamic model of the landing process of the Mars probe; linearize the dynamic model through an extended Kalman filter to obtain a linearized state equation; and perform landing navigation for the Mars probe in a combined navigation mode according to the linearized state equation and the preferred landmark, wherein the combined navigation refers to a navigation mode combining inertial navigation with landmark navigation;
[0058] The Fisher information matrix construction module 21 is also used to: convert the measurement model into a measurement equation; linearize the measurement equation through an extended Kalman filter to obtain a measurement matrix; and obtain the Fisher information matrix of each visible landmark based on the measurement matrix.
[0059] The method of the present invention is described below with a specific embodiment, and the implementation process of the specific embodiment is as follows:
[0060] 1. Mars landing navigation system model:
[0061] Landing a probe on Mars is an important step in realizing Mars exploration. It is necessary to determine the position, attitude, and velocity of the probe relative to the selected reference system. For ease of description, the following three coordinate systems are introduced: landing coordinate system, carrier coordinate system, and camera coordinate system. Figure 3 shown.
[0062] The landing coordinate system (L system) is centered on the predetermined landing point, with the z-axis perpendicular to the landing plane and pointing upward, the y-axis located on the landing plane pointing in the direction of the planet's rotation, and the x-axis pointing in the direction of the planet's rotation. L Axis, y L Axis and z L The axes together form a right-handed coordinate system.
[0063] Carrier coordinate system (B system) The carrier coordinate system is a coordinate system fixed to the detector body, and its origin is the center of mass of the detector. B Axis, y B Axis and z B The axes are usually defined about the principal axes of inertia of the detector.
[0064] The camera coordinate system (C system) takes the optical center of the camera as the origin of the coordinate system, x C y C The axis is parallel to the x and y axes of the image coordinate system, and the optical axis of the camera is z C Axis, the coordinate system satisfies the right-hand rule.
[0065] 1.1 Kinetic model
[0066] In view of the problem that it is difficult to obtain an accurate planetary landing mechanics model using traditional dynamics modeling, this paper directly uses the IMU (inertial navigation system) system to establish a dynamics model and gives the probe landing dynamics equation based on the IMU output. Assuming that the probe is subject to gravity and thrust, the probe landing dynamics model can be described as
[0067]
[0068] in L v and L r is the position and velocity relative to the landing coordinate system (L system); is the quaternion from L to B, b a and b ω is the drift bias of the accelerometer and gyroscope; L a thrust is the control acceleration vector; L g is the gravitational force, which is calculated as a constant; n g is the unmodeled gravitational acceleration; n wa and n wω is the drift deviation noise; ω=[ω x ω y ω z ] is the angular velocity of the detector rotation in frame B. Ω(ω) is defined as
[0069]
[0070] in is ω=[ω x ω y ω z ] is defined as
[0071]
[0072] The output value of IMU is a imu and ω imu It can be expressed as
[0073]
[0074] Among them C q is the attitude rotation matrix corresponding to the quaternion q, L ω G is the angular velocity of the planet in the L system, n a and n ω is the drift noise of acceleration and angular velocity, is zero-mean Gaussian white noise with covariance. Q is the covariance matrix
[0075]
[0076] 1.2 Landmark-based measurement model (or observation model)
[0077] Using the pinhole projection model, the camera is at t k Moment, for landmarks j The measurement is
[0078]
[0079] Where f is the focal length of the camera; is the noise-free measurement; is the measurement noise, and its covariance matrix is is the vector from the spaceborne camera to the landmark, and the calculation formula is as follows:
[0080]
[0081] L p j is the coordinate of the landmark in the L system, which can be obtained by comparing the captured image with the landmark database in the landing map; represents the map error, and its covariance is Represents the coordinates of the detector in the L system.
[0082] Then the lander navigation state vector There are the following forms
[0083]
[0084] Calculate the landmark p based on the state vector of the navigation system j The expected measurement value of
[0085]
[0086] in
[0087]
[0088] Combining equations (1) and (6), the state equation and measurement equation of landmark navigation can be written as
[0089]
[0090] where z imu =(a imu ,ω imu )’s IMU inertial measurement, To observe the coordinates of the landmarks, we need to Composition, N is the number of observed landmarks.
[0091]
[0092] Linearize the state equation:
[0093]
[0094] and G is the Jacobian matrix, defined as follows, where
[0095]
[0096] Linearize the measurement equation:
[0097]
[0098] Measurement Matrix The definition is as follows:
[0099]
[0100] in
[0101]
[0102] Then the covariance matrix of the navigation state error can be updated by numerical integration
[0103]
[0104] in Update the state and state variance of the navigation system using the Extended Kalman Filter (EKF)
[0105]
[0106] 2. Landmark selection method based on Fisher information matrix
[0107] 2.1 Fisher Information Matrix and CRLB
[0108] The Fisher (FIM) information matrix is defined as the inverse of CRLB (Cramer-Rao Lower Bound), which is the lowest lower bound that the state estimate covariance can achieve under ideal conditions. CRLB characterizes the uncertainty of the minimum target that can be obtained for a specified set of measurements in the state system. The inverse of CRLB is FIM, which represents the amount of information contained in a specified set of measurements in the state estimate. Mathematically, CRLB is defined as
[0109]
[0110] Among them, x is the vector to be estimated, is the estimated value of x obtained based on the measured data z, P is the covariance matrix of the estimated error, C is the lowest lower bound of the state error covariance matrix, i.e. CRLB, and its inverse J is FIM.
[0111] Consider a nonlinear system where the state equation and measurement equation are as follows:
[0112]
[0113] where x k represents the system state at time k, z k represents the measurement information at time k; v k is the systematic error, r k is the measurement error, assuming that v k With r k is Gaussian white noise, r k ~N(0,R k ),v k ~N(0,Q k ), f(·) and h(·) represent the measurement functions of the state model and the measurement model, and their Jacobian matrices are in is the gradient operator of x
[0114] The recursive formula for the FIM of this nonlinear system is:
[0115]
[0116] FIM has the following two important properties:
[0117] 1) In the landmark selection, for the observable landmarks of the detector at a certain moment, the scalar function of FIM will increase monotonically with the increase of the number of landmarks, that is, N→∞, U(J N )→∞.
[0118] 2) Under the condition that the landmarks can be observed, the FIM of multiple landmarks can be calculated by superposition. For the system model in formula (22), assuming that the number of candidate landmarks is N, the FIM of N landmarks can be obtained by recursion.
[0119]
[0120] 2.2. Landmark Optimization Scheme
[0121] In the landmark selection problem, if we only consider the current position of the detector, assuming that the current position of the detector is fixed, then for the landmarks that can be observed at this moment, the accumulated FIM information benefits of all observed landmarks are
[0122]
[0123] From the measurement matrix in (17) and (18), we can know
[0124]
[0125] Assume that the measurement noise of each landmark is uncorrelated and has the same variance, σ1 = σ2 = ... = σ N
[0126] but
[0127]
[0128] FIM can evaluate the performance of landmark navigation. Its mathematical expression is in matrix form. Different scalars M need to be selected to optimize FIM. However, using scalars to replace matrices may lead to information loss. Therefore, the selection of scalar functions is crucial. Currently, the indicators for measuring the optimal configuration of landmarks mainly include D-optimal, A-optimal, E-optimal, etc.
[0129] The D-optimal correspondence is the determinant of FIM, namely:
[0130] Μ D (J) = det(J) (28)
[0131] Since FIM is a symmetric positive definite matrix, its determinant is the product of all eigenvalues, which is proportional to the volume of the hyperellipsoid defined by the quadratic form accompanying the FIM matrix. Therefore, the physical meaning of maximizing det(J) is to minimize the volume of the hyperellipsoid.
[0132] The E-optimal correspondence is the minimum eigenvalue of FIM, that is:
[0133] Μ E (J) = min{eig(J)} (29)
[0134] From the above analysis, we can see that the physical meaning of maximizing min{eig(J)} is to minimize the maximum axis length of the hyperellipsoid.
[0135] The A-optimal correspondence is the trace of FIM, that is:
[0136] Μ A (J) = tr(J) (30)
[0137] The physical meaning of maximizing tr(J) is to minimize the variance of the average estimate.
[0138] The existing technology uses numerical simulation to verify that the A-optimal index has better effect than D-optimal and E-optimal for the target selection problem. This paper uses the trace of A-optimal FIM as a scalar function to optimize FIM.
[0139]
[0140] Since the matrix trace operation is commutative in matrix multiplication, we have
[0141]
[0142] Notice is an antisymmetric matrix with a main diagonal of 0. and At the current moment, it can be approximately regarded as a fixed value and is not affected by the choice of landmarks. The trace of FIM can be written as
[0143]
[0144] Where C0 is a constant, so maximizing the trace of the FIM matrix is equivalent to maximizing From (18), we can get
[0145]
[0146] Then maximizing the trace of the FIM matrix is equivalent to
[0147]
[0148] is the vector from the satellite camera to the landmark. Figure 4 As shown, for the optimal observation configuration, the value of the corresponding FIM track is only related to the number of landmarks and the relative position of the landmarks and the detector.
[0149] The landmarks observed by the onboard camera of the probe at different times are different. The closer the probe is to the landing point, the fewer landmarks the onboard camera can observe. The landmarks that the probe can observe at time k must meet the following conditions:
[0150]
[0151] α is the field of view angle, d is the reference direction vector of the camera optical axis in the L system, and δβ is the maximum camera error caused by the reference direction vector.
[0152] 2.3 Determination of the number of optimal landmarks
[0153] Use U=tr(J N) is used to describe the influence of landmark selection on navigation accuracy. Different landmark configurations and numbers will result in different U. Generally, the more landmarks are selected, the higher the navigation accuracy. However, too many landmarks will seriously overload the onboard computing resources. In the previous landmark selection problem, the number of preferred landmarks is generally a fixed value, and the number of preferred landmarks at each moment is the same. The contribution of landmarks to navigation accuracy at different moments is different. The above method cannot achieve the highest navigation accuracy at each stage. The landmark selection problem for the landing phase of the probe can be reduced to finding the maximum value of (35) under the constraint of (36) and the number of preferred landmarks is as small as possible, that is:
[0154]
[0155] Greedy algorithm is a common method for finding the optimal solution. It divides the solution process into several steps, and each step applies the greedy principle to select the best option in the current state, hoping that the final stacked result is also the optimal solution. This paper uses FIM combined with greedy algorithm to determine the configuration and number of the optimal landmarks.
[0156] Let B represent the set of all available landmarks. The goal is to find a subset D containing the fewest landmarks so that at each imaging time t k , the following constraints always hold
[0157]
[0158] in The set of candidate landmark points is S(t k )=(BD)∩A(t k ), A(t k ) indicates that at t k The landmarks visible at all times. U(D) is the landmark navigation accuracy obtained from the landmark dataset D. The algorithm flow is shown in Table 1. Convergence threshold U min This is an optional parameter that can be set to a fixed value or a dynamic value depending on the number of visible landmarks.
[0159] If U min If U is too small, too many landmarks will be selected, which will cause a serious load on the onboard computing resources. min If the value is too large, too few landmarks will be selected, resulting in insufficient navigation accuracy. At the beginning of landing, there are many landmarks that can be observed, and there are a large number of high-quality landmarks that contribute greatly to navigation accuracy. At this time, only a small number of landmarks can be selected from the observable landmarks to achieve higher navigation accuracy. Therefore, a larger U should be selected at the beginning of landing. minAvoid selecting too many landmarks, which will cause computing resource load and precision redundancy. As the probe descends in altitude, the closer it gets to the landing point, the fewer landmarks are visible, and the contribution of landmarks to navigation accuracy is low. If a larger U is still selected when the number of observable landmarks decreases, min , the number of preferred landmarks may be too small, resulting in insufficient navigation accuracy. When the number of observable landmarks is less than 3, there is no need to optimize the landmarks, and all observable landmarks can be used for pose calculation.
[0160] In summary min It should be a decreasing function of the landing time t. Use the simplest inverse proportional function to construct U min , as follows. is the trace of the FIM information matrix of the observable landmark that contributes the least to navigation accuracy. a is an optional parameter as follows, which is used to control U min The range of change:
[0161]
[0162]
[0163] Table 1. Greedy algorithm flow for landmark selection based on Fisher information matrix
[0164] 3. Simulation
[0165] In order to evaluate the preferred landmark scheme proposed in this specific embodiment, numerical simulation was performed using MATLAB.
[0166] 3.1 Simulation parameter settings
[0167] The Mars landing process was simulated, assuming that the landing phase lasts 100s. The navigation camera records images up to 50m at a rate of 1Hz. The field of view (FOV) of the camera is 50, the focal length is 2000 pixels, and the image noise is 1 pixel. Assuming that the inertial navigation unit (IMU) is LN-200, providing 50Hz inertial measurement values, the actual state of the probe at the initial moment is shown in Table 2. Assuming that the initial state error of the position is 100m, the initial state error of the velocity is 1m / s, the initial state error of the attitude of each axis is 1°, and the angular velocity of rotation is [0 0 0.2]° / s. Assuming that the distance from the landmark point to the local horizontal plane follows a normal distribution with a standard deviation of 2m, an irregular landing surface is generated. Assuming that the landmarks are randomly distributed on the irregular surface with an average density of 100 / km 2 , the standard deviation of the horizontal position error is 2 m. The process noise is Q. The local gravity noise is assumed to have a standard deviation of 1×10 -1 m / s 2 Gaussian white noise.
[0168] Q=diag([2.4×10 -14 I32.5×10 -14 I32.4×10 -7 I31.2×10 -7 I31×10 -8 I3])
[0169] Table 2 The actual state of the detector at the initial moment
[0170]
[0171]
[0172] 3.2 Simulation Experiment Results
[0173] like Figure 5 As shown, the probe's Mars landing orbit and the distribution of landmarks during simulation are given. Figure 5 The number of visible landmarks at each imaging moment during the probe's descent is shown. For a given landmark density, there are about 650 landmarks in the landing area. Therefore, if all available landmarks are used for navigation, about 650 landmarks must be stored on board, which is extremely challenging for the limited storage and computing resources in planetary landing missions. In order to reduce the number of landmarks selected for landmark navigation, Algorithm 1 is used to solve the problem of landmark number selection. In Algorithm 1, U min Set as a is an optional parameter. Table 3 shows U min The relationship between the number of all landmarks selected during the landing process and the running time of the simulation program, U min The larger the value, the more landmarks are selected, and the longer the average and maximum running time of the corresponding simulation program. min This will result in too many landmarks being selected. min After decreasing to a certain value, further decreasing will not significantly improve navigation accuracy. min , at this time U min It can be regarded as infinite, that is, no landmark optimization is performed, and all observed landmarks are solved together for pose. When the navigation accuracy reaches a higher level, the increase in the number of observed landmarks has limited improvement in navigation accuracy. In the late landing period, when the number of visible landmarks in D drops to zero, the landmark-assisted inertial navigation system is transformed into a pure inertial navigation system. Since the accumulated error of inertial navigation cannot be effectively corrected in the late landing period, the navigation accuracy diverges in the late landing period but is always within a tolerance range.
[0174] Table 3. Different U min The number of selected landmarks and the running time
[0175]
[0176] In order to evaluate the accuracy of the landmark optimization scheme, FIM is combined with the greedy algorithm to obtain the number and configuration of the optimal landmarks at each moment. According to the above simulation results, the convergence threshold U in the greedy algorithm is set to min Set to min(U(D)) / 150t, the number of preferred landmarks at each moment is as follows Figure 7 , randomly select the same number of landmarks as the original one at each moment, and compare the navigation errors of the two on each axis.
[0177] Figure 8 , Fig. 9 , Fig.10 The mean square error (RMSE) of position, velocity, and attitude is shown. When the same number of landmarks is selected, the error and RMSE of position and attitude on each axis are lower when the landmark is selected at the beginning of landing than when the landmark is randomly selected. In the later stage of landing, as the number of observable landmarks decreases, the error increases, but the error is always within a tolerance range. It is worth noting that there is no obvious difference between the two for the yaw angle error. This is because the yaw angle is not observable when only known landmarks are used for measurement. Therefore, no matter how the known landmarks are selected, the yaw angle will not change much.
[0178] As for the error and RMSE of velocity on each axis, at the beginning of landing, the velocity error of the preferred landmark is lower than that of the randomly selected landmark, but both are maintained at a high level. As time goes by, the landmark measurement information corrects the inertial navigation measurement information multiple times, and the velocity error is significantly reduced. Compared with the velocity error of the randomly selected landmark, the velocity error of the preferred landmark converges faster. In the later stage of landing, as the number of visible landmarks decreases, there is no significant difference in the errors of the two, and both are at a low level. This is because the measurement information of the landmark does not directly contain the measurement information of the velocity, and the correction of the velocity error is indirect correction. The estimation of the detector velocity is mainly obtained by using the dynamic model, the measurement information is underdetermined, and the state estimation effect is unstable. In the absence of landmark measurement information, there is no significant difference in the errors between the two.
[0179] Fig.11 The plan distribution diagram of some selected landmarks is shown. The moment when there are many visible landmarks in the early stage of landing is selected. Starting from the 5th second, the sampling is performed every 5 seconds, and a total of 6 samplings are performed. From the configuration of the landmarks, the landmarks are distributed more dispersedly, and most of them are distributed at the edge of the image. For a single landmark, the trace of its FIM information matrix is:
[0180]
[0181] Projecting the point of the camera coordinate system to the pixel coordinate system, we can get
[0182]
[0183] Comparing equation (40) and equation (41), if FIM is larger, then The bigger, The smaller it is, the more it reflects (41), that is, |x in the pixel coordinate system i |,|y i | is larger, that is, the landmark points are distributed at the edge of the image. The area of the polygon enclosed by the landmark points at the edge of the image is proportional to the distance from the landmark to the projection of the probe on the landing surface. The larger the distance, the larger the area of the polygon. The area of the polygon enclosed by the landmark points at the edge of the image can be used to approximate the accuracy of the landmark measurement model. Ideally, the optimal landmark distribution should be centered on the projection point of the probe and distributed at the outermost part of the image taken by the onboard camera, and the landmark points form a regular polygon with each other. However, due to the influence of the field of view angle, focal length of the onboard camera and the distribution of landmarks on the surface of Mars, as shown in the landmark distribution diagrams at 20, 25, and 30 seconds, there may not be landmark points at the edge of the image, and the landmark points may be concentrated on one side of the probe.
[0184] The disclosed embodiments are described above to enable any person skilled in the art to implement or use the present invention. Various modifications of these embodiments are obvious to those skilled in the art, and the general principles defined in this application may also be applied to other embodiments without departing from the spirit and scope of the present disclosure. Therefore, the present disclosure is not limited to the embodiments given in this application, but is consistent with the broadest scope of the principles and novel features disclosed in this application.
[0185] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method 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. A Mars probe landing stage navigation method, characterized in that: include: Based on the visible landmarks at the current moment, a measurement model of the visible landmarks is established; According to the measurement model, a Fisher information matrix of each visible landmark is obtained; Calculate the trace of the Fisher information matrix corresponding to each visible landmark; When the number of the visible landmarks is not less than 3, constructing a preferred configuration of the landmark distribution, the constraint condition of the preferred configuration of the landmark distribution is: selecting a plurality of the visible landmarks as preferred landmarks, and maximizing the sum of the traces of the preferred landmarks while minimizing the number of the preferred landmarks; Using a greedy algorithm to solve the preferred configuration and determine the preferred landmark; The Mars probe is navigated for landing by means of the preferred landmarks.
2. The Mars probe landing stage navigation method according to claim 1, characterized in that: Also includes: When the number of the visible landmarks is greater than zero and less than 3, all the visible landmarks are used as the preferred landmarks; When the number of the visible landmarks is zero, landing navigation is performed in an inertial navigation manner.
3. The Mars probe landing stage navigation method according to claim 2, characterized in that: In the process of solving the preferred configuration using the greedy algorithm, the convergence threshold of the greedy algorithm is a variable value.
4. The Mars probe landing stage navigation method according to claim 3, characterized in that: The convergence threshold is in negative proportion to the landing time, and is in positive proportion to the minimum value of all visible landmark traces; wherein the landing time refers to the time elapsed from the start of the landing process to the current moment.
5. The Mars probe landing stage navigation method according to claim 1, characterized in that: The method of performing landing navigation on the Mars probe by using the preferred landmarks further includes: Establish a dynamic model of the Mars rover landing process; The dynamic model is linearized by an extended Kalman filter to obtain a linearized state equation; According to the linearized state equation and the preferred landmark, performing landing navigation on the Mars probe in a combined navigation manner, wherein the combined navigation refers to a navigation manner combining inertial navigation with landmark navigation; The step of obtaining the Fisher information matrix of each visible landmark according to the measurement model specifically includes: converting the measurement model into a measurement equation; The measurement equation is linearized by an extended Kalman filter to obtain a measurement matrix; According to the measurement matrix, the Fisher information matrix of each visible landmark is obtained.
6. A Mars rover landing stage navigation system, characterized in that: include: A Fisher information matrix building module is used to establish a measurement model of visible landmarks based on the visible landmarks at the current moment; obtain the Fisher information matrix of each visible landmark according to the measurement model; and calculate the trace of the Fisher information matrix corresponding to each visible landmark; A preferred configuration building module, used for building a preferred configuration of the landmark distribution when the number of the visible landmarks is not less than 3, wherein the constraint condition of the preferred configuration of the landmark distribution is: selecting a plurality of the visible landmarks as preferred landmarks, and maximizing the sum of the traces of the preferred landmarks while minimizing the number of the preferred landmarks; A preferred landmark acquisition module, used for solving the preferred configuration by using a greedy algorithm to determine the preferred landmark; The first navigation module is used to perform landing navigation for the Mars probe through the preferred landmarks.
7. The Mars probe landing stage navigation system according to claim 6, characterized in that: Also includes: A second preferred landmark acquisition module, configured to use all the visible landmarks as the preferred landmarks when the number of the visible landmarks is greater than zero and less than 3; The second navigation module is used to perform landing navigation in an inertial navigation manner when the number of the visible landmarks is zero.
8. The Mars probe landing stage navigation system according to claim 7, characterized in that: In the preferred landmark acquisition module, the convergence threshold of the greedy algorithm is a variable value.
9. The Mars probe landing stage navigation system according to claim 8, characterized in that: The convergence threshold is in negative proportion to the landing time, and is in positive proportion to the minimum value of all visible landmark traces; wherein the landing time refers to the time elapsed from the start of the landing process to the current moment.
10. The Mars probe landing stage navigation system according to claim 6, characterized in that: The first navigation module is specifically used to establish a dynamic model of the landing process of the Mars probe; linearize the dynamic model through an extended Kalman filter to obtain a linearized state equation; and perform landing navigation for the Mars probe in a combined navigation mode according to the linearized state equation and the preferred landmark, wherein the combined navigation refers to a navigation mode combining inertial navigation with landmark navigation; The Fisher information matrix construction module is also used to: convert the measurement model into a measurement equation; linearize the measurement equation through an extended Kalman filter to obtain a measurement matrix; and obtain the Fisher information matrix of each visible landmark based on the linearized measurement matrix.
Citation Information
Patent Citations
Planetary landing autonomous optical navigation landmark time sequence optimization method
CN117073681A
Landing system for an aircraft
US20160093225A1
