High-precision pose positioning system for strip mine electric shovel based on factor graph optimization

Abstract

The invention relates to a strip mine electric shovel high-precision pose positioning system based on factor graph optimization, and belongs to the technical field of electric shovel positioning. Comprising a data acquisition module used for acquiring an original GNSS observation sequence and an original IMU observation sequence, and performing space-time alignment to obtain a to-be-corrected GNSS observation sequence and a to-be-denoised IMU observation sequence; the de-noising module is used for de-noising the IMU observation sequence to be de-noised by adopting a pre-trained de-noising diffusion model; the working condition recognition module is used for recognizing the current working condition of the electric shovel; the GNSS data processing module is used for correcting the GNSS observation sequence to be corrected according to the current working condition of the electric shovel; the model construction module is used for constructing a multi-source fusion factor graph optimization model according to the de-noised IMU observation sequence and the corrected GNSS observation sequence; and the pose determination module is used for solving the factor graph optimization model to obtain the pose of the electric shovel. According to the invention, high-precision pose positioning of the strip mine electric shovel can be realized.

Description

Technical Field

[0001] This invention relates to the field of electric shovel positioning technology, and in particular to a high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization. Background Technology

[0002] In open-pit mining operations, determining the position of the electric shovel is crucial for both operational efficiency and safety. Traditional positioning methods based on a single GNSS (Global Navigation Satellite System) or IMU (Inertial Measurement Unit) require initial alignment in dynamic environments, but this initialization process is time-consuming. In static environments, gyroscope drift causes divergence in the heading angle, leading to decreased positioning accuracy or even failure. Furthermore, IMUs experience high measurement noise and nonlinear amplification of measurement errors under conditions of strong vibration and impact.

[0003] Patent CN118311629B proposes a seamless indoor / outdoor positioning method based on UWB / GNSS / IMU. It achieves accurate positioning in different indoor and outdoor environments through loose coupling between GNSS and IMU, combined with an extended Kalman filter algorithm. Patent CN120491132A proposes a CTG neural network-assisted navigation method and device for weak GNSS signal scenarios. In scenarios with stable GNSS signals, it simultaneously collects raw IMU data and GNSS satellite positioning data from the mobile vehicle, constructing a spatiotemporal feature dataset including dynamic acceleration, angular velocity, and position increment. Then, it builds a CTG network and uses the spatiotemporal feature dataset to train the CTG network offline, obtaining the optimal network weights and model structure. A smooth positioning result is obtained through a tightly coupled algorithm.

[0004] However, the above methods are all implemented in vehicle-mounted environments with relatively low vibration. For open-pit mine electric shovels operating in complex environments, the strong vibrations and impacts during operation make it difficult to calculate the position and posture of the shovel body and bucket in real time. Therefore, for complex open-pit mine operating environments, there is an urgent need for a solution that can achieve high-precision positioning, simultaneous measurement of the posture of multiple components, strong anti-interference capabilities, and reasonable cost. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization. The technical solution of this invention is as follows: A high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization, comprising: The data acquisition module is used to acquire the original GNSS observation sequence and the original IMU observation sequence pre-installed on the electric shovel by the antenna and IMU, and to perform spatiotemporal alignment on the original GNSS observation sequence and the original IMU observation sequence to obtain the GNSS observation sequence to be corrected and the IMU observation sequence to be denoised. The denoising module is used to denoise the IMU observation sequence to be denoised using a pre-trained denoising diffusion model, so as to obtain the denoised IMU observation sequence. The working condition identification module is used to identify the current working condition of the electric shovel based on the GNSS observation sequence to be corrected and the denoised IMU observation sequence; The GNSS data processing module is used to correct the GNSS observation sequence to be corrected according to the current working condition of the electric shovel, and obtain the corrected GNSS observation sequence. The model building module is used to construct a multi-source fusion factor graph optimization model based on the denoised IMU observation sequence and the corrected GNSS observation sequence. The pose determination module is used to solve the factor graph optimization model to obtain the electric shovel pose.

[0006] Preferably, the working condition identification module includes: The statistical feature calculation unit is used to calculate the horizontal velocity, rate of change of heading angle, and variance of vertical acceleration of the electric shovel at the current moment based on the denoised IMU observation sequence. The walking condition determination unit is used to determine the current working condition of the electric shovel as the walking condition when it is determined that the horizontal speed of the electric shovel at the current moment is greater than the preset horizontal speed threshold, the rate of change of the heading angle at the current moment is less than the preset rate of change of the heading angle threshold, and the variance of the vertical acceleration of the electric shovel at the current moment is less than the preset vertical acceleration variance threshold. The slewing condition determination unit is used to determine the current working condition of the electric shovel as a slewing condition when the current horizontal speed of the electric shovel is less than or equal to a preset horizontal speed threshold and the current heading angle change rate is greater than or equal to a preset heading angle change rate threshold. The excavation condition determination unit is used to determine the current working condition of the electric shovel as excavation condition when the signal-to-noise ratio of the electric shovel at the current moment in the GNSS observation sequence to be corrected continuously decreases or is interrupted.

[0007] Preferably, the GNSS observation sequence to be corrected includes a pseudorange sequence to be corrected, and the GNSS data processing module includes: The pseudorange correction unit for the walking mode is used to calculate the pseudorange mean and pseudorange standard deviation of a first preset number of pseudoranges in the pseudorange sequence to be corrected when the current working mode of the electric shovel is the walking mode. Based on the pseudorange mean and pseudorange standard deviation, it determines whether the pseudorange is an abnormal pseudorange. If the pseudorange is an abnormal pseudorange, it replaces the pseudorange with the previous pseudorange. After obtaining the pseudorange sequence after preliminary correction, it fits the pseudorange sequence after preliminary correction through a quadratic polynomial fitting model to obtain the corrected pseudorange as the corrected GNSS observation sequence. The slewing condition pseudorange correction unit is used to perform pseudorange smoothing correction on any pseudorange in the pseudorange sequence to be corrected when the current working condition of the electric shovel is slewing condition, based on the second preset number of pseudoranges in its neighborhood and the corresponding carrier phase, to obtain the corrected pseudorange sequence as the corrected GNSS observation sequence. The excavation-condition pseudorange correction unit is used to interpolate the pseudorange sequence to be corrected when the current working condition of the electric shovel is excavation, to obtain the interpolated pseudorange sequence, and to calculate the position change of the electric shovel through the denoised IMU observation sequence. Based on the preliminary corrected pseudorange sequence and the position change of the electric shovel, the corrected pseudorange sequence is calculated as the corrected GNSS observation sequence.

[0008] Preferably, the model building module includes: The state definition unit is used to define the global state vector to be estimated in the factor graph optimization model. The factor node construction unit is used to construct inertial factor, GNSS factor, dual-antenna baseline factor, and motion constraint factor based on the denoised IMU observation sequence, the corrected GNSS observation sequence, and the sliding window. The model building unit is used to combine the global state vector to be estimated, the inertia factor, the GNSS factor, the dual-antenna baseline factor, and the motion constraint factor to obtain the factor graph optimization model.

[0009] Preferably, the global state vector to be estimated Represented as: ; ; in, Indicates the size of the sliding window. This represents the state at time k. , and Let b represent the position, velocity, and attitude of the b-frame relative to the e-frame at time k, respectively. The b-frame is the carrier coordinate system, and the e-frame is the geocentric Cartesian coordinate system. and These represent the deviation vectors of the accelerometer and gyroscope in the IMU at time k, respectively. This represents the inter-system deviation vector of GNSS systems at time k. This represents the tropospheric wet delay error component in the zenith direction of the mobile station at time k. and These represent inter-station differential and inter-satellite differential, respectively. This represents the integer ambiguity after double difference. This indicates the number of consecutive ambiguities within the sliding window. and These represent the rover and the base station, respectively, with IF indicating an ionospheric-free combination. This represents the double-difference ambiguity between the rover and the base station after the ionosphere is removed.

[0010] Preferably, the factor node construction unit, when constructing the inertia factor, is used for: Based on the denoised IMU observation sequence, the time interval is recursively calculated using the following formula. IMU pre-integration term , and : ; in, , The time interval between adjacent observations in the denoised IMU observation sequence; , and Time interval The IMU pre-integration term within, , and Time interval The IMU pre-integration term within the memory, A rotation matrix representation of quaternions; and These are the specific force sequence and angular velocity sequence from the denoised IMU observation sequence, respectively. Representing quaternion multiplication, Indicates the accelerometer in the IMU The deviation vector at time; According to time interval IMU pre-integration term , and The inertia factor is calculated using the following formula. : ; in, This represents the set of denoised IMU observations within the sliding window; This represents the state change from time k to time k+1 obtained by IMU pre-integration in system b; yes The rotation matrix representation; , and These represent the position, velocity, and attitude of frame b relative to frame e at time k+1, respectively. This represents the gravity vector in the e-frame. ; This represents the rotation vector for extracting quaternions. and These represent the deviation vectors of the accelerometer and gyroscope in the IMU at time k+1, respectively.

[0011] Preferably, the factor node construction unit, when constructing GNSS factors, is used for: GNSS factors, including pseudorange factor, carrier phase factor, and tropospheric-inter-system deviation factor, are constructed based on the corrected GNSS observation sequence. , is represented as: ; Where G represents the corrected GNSS observation sequence, and g represents a corrected GNSS observation value in the corrected GNSS observation sequence. This represents the GNSS residual term constructed from the corrected GNSS observation sequence. P, L, and T represent the time series and satellite sequence sets of pseudorange, carrier phase, and tropospheric-system bias, respectively, and i and j represent the indices of the observed satellites. and Let i and j represent observation satellites. Represents a robust loss function. This represents the covariance matrix corresponding to the GNSS factor; , and These represent the pseudorange factor, carrier phase factor, and tropospheric-system deviation factor, respectively. , and These represent the covariance matrices corresponding to the pseudorange factor, carrier phase factor, and tropospheric-system deviation factor at time k, respectively. ; in, This indicates that the satellite is being observed. The single difference between the rover and the base station predicted by INS is shown below. This indicates the observation satellites predicted by INS. With mobile station The distance between them; Indicates observation satellite With the base station The distance between them; This indicates the location of the mobile station as predicted by INS; Indicates observation satellite Location; Indicates the base station Location; Indicates observation of satellites The single difference between the rover and the base station predicted by INS is shown below. This represents the double-difference tropospheric residual; , represents the double-difference pseudo-range after the combination without ionosphere; Indicates information about observation satellites and The geometric distance between the mobile station and the base station; This indicates the noise of the double-difference pseudorange observation; ; in, Indicates wavelength; Indicates the double-difference ambiguity after ionosphere-free combination; , indicating the double-difference carrier phase after ionospheric combination; Indicates the wavelength after IF combination; This indicates double-difference carrier phase observation noise; ; in, , and These represent the inter-system bias parameters for BDS, Galileo, and GLONASS, respectively. It is the tropospheric wet delay error component in the zenith direction of the mobile station; k and k+1 represent time k and time k+1, respectively.

[0012] Preferably, the factor node construction unit, when constructing the dual-antenna baseline factor, is used for: Dual-antenna baseline factors were constructed based on the corrected GNSS observation sequence. , is represented as: ; in, and They represent observation satellites. and The single difference between mobile station antennas A and B; This represents the double-difference tropospheric residual between mobile station antennas A and B; Indicates the wavelength after IF combination; This represents the double-difference ambiguity between mobile station antennas A and B after they are combined without an ionosphere; Indicates observation of satellites and The double-difference carrier phase between mobile station antennas A and B; , and They represent observation satellites. and The unit observation vector at the receiver. This is the baseline vector between the two antennas A and B of the mobile station; Indicates wavelength; and Representing mobile stations The two antennas A and B are related to the observation satellite. and The double-difference ambiguity and carrier phase observation noise.

[0013] Preferably, the factor node construction unit is used to: Construct motion constraint factors including rotational motion factors, walking motion factors, and mining motion factors. , is represented as: ; in, This represents the set of motion constraint factors within the sliding window. This represents one of the motion constraint factors. This represents the covariance matrix corresponding to the motion constraint factor; This represents the motion constraint residuals constructed based on different motion modes; S, W, and D represent the sets of time series during the rotation, walking, and digging processes, respectively. , and These represent the rotational motion factor, the walking motion factor, and the digging motion factor, respectively. , and Let represent the covariance matrices corresponding to the rotational motion factor, walking motion factor, and mining motion factor at time k, respectively. ; in, and Let K and K+1 represent the vertical positions of the cab and bucket in the e-frame at time k and time k+1, respectively. This indicates the expected vertical velocity during walking; This represents converting a quaternion into a Lie algebra vector. Indicates the expected change in attitude; and Let b represent the attitude of the b-frame relative to the e-frame at time k and k+1, respectively.

[0014] Preferably, the pose determination module includes: The global optimization unit is used to construct the global optimization objective function of the factor graph optimization model, and solve the global optimization objective function by the maximum a posteriori probability to obtain the optimal state estimate of the global state vector to be estimated. The marginalization unit is used to transform the optimal state estimate and its corresponding observation constraints at the oldest time step in the sliding window into prior information after the global optimization is completed, and add it to the global optimization objective function at the next time step.

[0015] All of the above-mentioned optional technical solutions can be combined arbitrarily, and the present invention will not provide a detailed description of the structure after each combination.

[0016] By means of the above solution, the beneficial effects of the present invention are as follows: By performing electric shovel pose localization based on the original GNSS and IMU observation sequences, and simultaneously fusing data acquired by the antenna and IMU, multi-directional data can be provided for pose localization, thus initially improving positioning accuracy. By employing a pre-trained denoising diffusion model to denoise the IMU observation sequence, targeted denoising can be achieved, preventing noise data from affecting the accuracy of the positioning results, thereby further improving the accuracy of the positioning results. By identifying the current operating condition of the electric shovel based on the GNSS observation sequence to be corrected and the denoised IMU observation sequence, and then correcting the GNSS observation sequence based on the current operating condition, the obtained corrected GNSS observation sequence better reflects the actual pose of the electric shovel, thus further improving the accuracy of the positioning results. By constructing a multi-source fusion factor graph optimization model, and including motion constraint factors in the factor graph optimization model, drift when the GNSS signal is lost can be effectively suppressed. In summary, this invention can improve the accuracy of positioning results from multiple perspectives, thereby achieving high-precision pose localization of open-pit mine electric shovels.

[0017] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the composition structure of the present invention; Figure 2 This is a data processing flowchart of the present invention. Detailed Implementation

[0019] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0020] like Figure 1 and Figure 2 As shown in the figure, the high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization provided in this embodiment of the invention includes: The data acquisition module is used to acquire the original GNSS observation sequence and the original IMU observation sequence pre-installed on the electric shovel by the antenna and IMU, and to perform spatiotemporal alignment on the original GNSS observation sequence and the original IMU observation sequence to obtain the GNSS observation sequence to be corrected and the IMU observation sequence to be denoised. The denoising module is used to denoise the IMU observation sequence to be denoised using a pre-trained denoising diffusion model, so as to obtain the denoised IMU observation sequence. The working condition identification module is used to identify the current working condition of the electric shovel based on the GNSS observation sequence to be corrected and the denoised IMU observation sequence; The GNSS data processing module is used to correct the GNSS observation sequence to be corrected according to the current working condition of the electric shovel, and obtain the corrected GNSS observation sequence. The model building module is used to construct a multi-source fusion factor graph optimization model based on the denoised IMU observation sequence and the corrected GNSS observation sequence. The pose determination module is used to solve the factor graph optimization model to obtain the electric shovel pose.

[0021] In this embodiment of the invention, a main antenna (A1) and a secondary antenna (A2) are pre-installed above the electric shovel's cab, and antennas A3 and A4 are installed at the tail of the boom and on the bucket, respectively. Hardware interface adaptation and integration development for the multiple antennas are completed using an embedded development board. An IMU is installed in the cab, and the original GNSS observation sequences (including original pseudorange sequences, original carrier phase sequences, original Doppler shift sequences, and original signal-to-noise ratio sequences) and original IMU observation sequences (including original specific force sequences and original angular velocity sequences) from the bucket are acquired through the antennas and the IMU. Based on this, a unified time reference and carrier coordinate system (b-frame) are established, and hardware-level spatiotemporal synchronization is performed on the original GNSS observation sequences and the original IMU observation sequences to obtain the GNSS observation sequences to be corrected and the IMU observation sequences to be denoised.

[0022] In one specific embodiment, the noise reduction module includes: The normalization unit is used to normalize the IMU observation sequence to be denoised, so as to obtain the standard IMU observation sequence. The fragment acquisition unit is used to process the standard IMU observation sequences into a B×L×H standard IMU observation sequencing matrix, where B is the batch size, L is the time step, and H is the number of IMU channels; The denoising unit is used to input the standard IMU observation matrix into the pre-trained denoising diffusion model, output the denoised IMU observation matrix from the denoising diffusion model, and concatenate the denoised IMU observation matrices into a denoised IMU observation sequence.

[0023] In this embodiment of the invention, the denoising and diffusion model needs to be trained before inputting the standard IMU observation matrix into the pre-trained denoising and diffusion model. Specifically, training the denoising and diffusion model includes the following steps: (1) Dataset Construction: An HWT605 inertial measurement unit was installed in the cab of the electric shovel to collect raw IMU observation sequences under various working conditions, including walking, digging, and slewing, covering scenarios with strong vibration and impact. The raw force comparison sequences output by the IMU were... and the original angular velocity sequence Standardize them separately, the formula is: ,in, The mean, Let the standard deviation be such that the specific force and angular velocity are distributed in the range of The processed data is divided into training, validation, and test sets in a 7:2:1 ratio.

[0024] (2) Diffusion model design: A denoising network based on the U-Net architecture is constructed, which includes an input layer, encoder, temporal attention module, decoder and output layer, to capture temporal dependencies: ① Noise Learning - Forward Diffusion: To clean IMU data Gaussian noise is gradually added to generate noisy data. Let the number of diffusion steps be T (in this embodiment, T=1000), and the noisy data at step t... satisfy: ,in It is Gaussian noise. , Let be the diffusion coefficient, satisfying . .

[0025] ② Data cleaning - back diffusion (U-Net network): Input layer: Input noisy data fragments (shape: (B=16 is the batch size, L=200 is the time step, H=6 is the number of IMU channels) and the diffusion step number encoding t; The encoder contains four convolutional blocks, each consisting of a 1D convolution (kernel_size=3), batch normalization, GELU activation function, and downsampling with a stride of 2, progressively compressing the temporal dimension and extracting features. Temporal attention module: Embedded with a Transformer-based multi-head self-attention and feedforward neural network, it captures long-term temporal dependencies and performs nonlinear transformations and feature enhancements on the attention output. Attention weights are calculated as follows: Where Q, K, and V are the query, key, and value matrices, respectively. For key dimensions; Decoder: Upsamples the compressed features extracted by the encoder to gradually restore the original sequence length. Its structure is symmetrical with the encoder and contains 4 deconvolution blocks. It uses skip connections to concatenate the output of the encoder at the corresponding stage with the upsampled output of the current decoder to gradually restore the time dimension. Output layer: Output is a noise prediction tensor with the same shape as the input. , indicating that the model is for the th Add noise to the prediction step.

[0026] (3) Model training: IMU data of electric shovel under typical strong vibration conditions are collected in advance to construct a dataset. The training model learns the reverse distribution process from "Gaussian noise" to "normal movement data of electric shovel", so that it can identify and remove non-Gaussian strong vibration noise.

[0027] After training the denoising diffusion model through the above steps, the standard IMU observation matrix is ​​input into the pre-trained denoising diffusion model. The denoising diffusion model performs the reverse denoising process and outputs the denoised IMU observation matrix. The denoised IMU observation matrices are then concatenated to form the denoised IMU observation sequence.

[0028] In one specific embodiment, the working condition identification module includes: The statistical feature calculation unit is used to calculate the horizontal velocity, rate of change of heading angle, and variance of vertical acceleration of the electric shovel at the current moment based on the denoised IMU observation sequence. The walking condition determination unit is used to determine the current working condition of the electric shovel as the walking condition when it is determined that the horizontal speed of the electric shovel at the current moment is greater than the preset horizontal speed threshold, the rate of change of the heading angle at the current moment is less than the preset rate of change of the heading angle threshold, and the variance of the vertical acceleration of the electric shovel at the current moment is less than the preset vertical acceleration variance threshold. The slewing condition determination unit is used to determine the current working condition of the electric shovel as a slewing condition when the current horizontal speed of the electric shovel is less than or equal to a preset horizontal speed threshold and the current heading angle change rate is greater than or equal to a preset heading angle change rate threshold. The excavation condition determination unit is used to determine the current working condition of the electric shovel as excavation condition when the signal-to-noise ratio of the electric shovel at the current moment in the GNSS observation sequence to be corrected continuously decreases or is interrupted.

[0029] In this embodiment of the invention, the horizontal velocity of the electric shovel at the current moment is directly extracted from the denoised IMU observation sequence. The rate of change of the shovel's heading angle at the current moment is calculated. When, the calculation formula is: ,in, and Let be the heading angles at time k and time k-1, respectively. This is the sampling time interval for the IMU.

[0030] Calculate the variance of the vertical acceleration of the electric shovel at the current moment. When, the calculation formula is: Where E represents the number of sampling points within the time window, This represents the vertical acceleration value at time k. This represents the average vertical acceleration within the time window.

[0031] Specifically, when the signal-to-noise ratio of the electric shovel in the GNSS observation sequence to be corrected decreases continuously, such as when it is less than 25dB, the current working condition of the electric shovel is determined to be digging condition.

[0032] In one specific embodiment, the GNSS observation sequence to be corrected includes a pseudorange sequence to be corrected, and the GNSS data processing module includes: The pseudorange correction unit for the walking mode is used to calculate the pseudorange mean and pseudorange standard deviation of a first preset number of pseudoranges in the pseudorange sequence to be corrected when the current working mode of the electric shovel is the walking mode. Based on the pseudorange mean and pseudorange standard deviation, it determines whether the pseudorange is an abnormal pseudorange. If the pseudorange is an abnormal pseudorange, it replaces the pseudorange with the previous pseudorange. After obtaining the pseudorange sequence after preliminary correction, it fits the pseudorange sequence after preliminary correction through a quadratic polynomial fitting model to obtain the corrected pseudorange as the corrected GNSS observation sequence. The slewing condition pseudorange correction unit is used to perform pseudorange smoothing correction on any pseudorange in the pseudorange sequence to be corrected when the current working condition of the electric shovel is slewing condition, based on the second preset number of pseudoranges in its neighborhood and the corresponding carrier phase, to obtain the corrected pseudorange sequence as the corrected GNSS observation sequence. The excavation-condition pseudorange correction unit is used to interpolate the pseudorange sequence to be corrected when the current working condition of the electric shovel is excavation, to obtain the interpolated pseudorange sequence, and to calculate the position change of the electric shovel through the denoised IMU observation sequence. Based on the preliminary corrected pseudorange sequence and the position change of the electric shovel, the corrected pseudorange sequence is calculated as the corrected GNSS observation sequence.

[0033] Specifically, during the movement of the electric shovel, the vibration caused by the uneven ground is relatively small, and the GNSS signal in the open-pit mine is not affected by obstruction. Therefore, embodiments of the present invention can calculate the average pseudorange over a first preset number of recent times (e.g., 10 moments). and pseudorange standard deviation If the current pseudorange satisfy If so, the current pseudorange is determined to be an abnormal pseudorange, and the pseudorange from the previous time step is used. Instead, for the dynamic lag error caused by the walking process, this embodiment of the invention uses a quadratic polynomial fitting model to correct the pseudorange, the formula being: ; in, The time difference between the current time and the starting time. The coefficients obtained by fitting the error of the most recent 20 time steps using the least squares method are used to compensate for dynamic lag. For the corrected pseudorange, This is the original pseudorange.

[0034] Furthermore, since the vibration of the electric shovel body and bucket is minimal during rotation, this embodiment of the invention utilizes the carrier phase change rate to smooth the pseudorange, converting the "number of phase change cycles" into the actual distance change, thereby improving pseudorange accuracy. Specifically, pseudorange smoothing correction is performed using the following formula: ; in, This indicates the second preset quantity (usually 6 to 9). For the corrected pseudorange, For the original pseudorange, For carrier wavelength, for The carrier phase observation at time [time]. for The carrier phase observation value at time t.

[0035] Furthermore, due to the severe vibrations caused by the impact of large pieces of material during excavation, the GNSS signal of the bucket is easily blocked at the initial stage of excavation. Therefore, when a GNSS signal interruption is detected, such as a signal-to-noise ratio (SNR) <25dB or a pseudorange change rate >1m / s for three consecutive time points, this embodiment of the invention calculates the position change of the electric shovel through interpolation and denoising of the IMU observation sequence. To fill in the pseudorange, the following formula is used: ; ; ; in, These are the indices of the valid times before and after the interruption. , for The original pseudorange of time, for The original pseudorange of time, for pseudorange interpolation at time t, This represents the change in position of the electric shovel. express Comparison in a system (navigation coordinate system), This represents the gravity vector in the n-system (the rotation matrix can be calculated using the longitude, latitude, and elevation of the target location, and then transformed to the geocentric coordinate system). For observing satellites The unit line-of-sight vector, For the corrected pseudorange, Let n be the velocity of the electric shovel in the n-system at time k-1. Let n be the posture of the electric shovel at time k-1. It is a rotation matrix representation of quaternions.

[0036] In one specific embodiment, the model building module includes: The state definition unit is used to define the global state vector to be estimated in the factor graph optimization model. The factor node construction unit is used to construct inertial factor, GNSS factor, dual-antenna baseline factor, and motion constraint factor based on the denoised IMU observation sequence, the corrected GNSS observation sequence, and the sliding window. The model building unit is used to combine the global state vector to be estimated, the inertia factor, the GNSS factor, the dual-antenna baseline factor, and the motion constraint factor to obtain the factor graph optimization model.

[0037] In this embodiment of the invention, the global state vector to be estimated includes navigation state and sensor error terms. In one specific embodiment, the global state vector to be estimated... Represented as: ; ; in, Indicates the size of the sliding window. This represents the state at time k. , and Let b represent the position, velocity, and attitude of the b-frame relative to the e-frame at time k, respectively. The b-frame is the carrier coordinate system, and the e-frame is the geocentric Cartesian coordinate system. and These represent the deviation vectors of the accelerometer and gyroscope in the IMU at time k, respectively. This represents the inter-system deviation vector of GNSS systems at time k. This represents the tropospheric wet delay error component in the zenith direction of the mobile station at time k. and These represent inter-station differential and inter-satellite differential, respectively. This represents the integer ambiguity after double difference. This indicates the number of consecutive ambiguities within the sliding window. and These represent the rover and the base station, respectively, with IF indicating an ionospheric-free combination. This represents the double-difference ambiguity between the rover and the base station after the ionosphere is removed.

[0038] In one specific embodiment, the factor node construction unit, when constructing the inertia factor, is used for: Based on the denoised IMU observation sequence, the time interval is recursively calculated using the following formula. IMU pre-integration term , and : ; in, , The time interval between adjacent observations in the denoised IMU observation sequence; , and Time interval The IMU pre-integration term within the memory, , and Time interval The IMU pre-integration term within the memory, A rotation matrix representation of quaternions; and These are the specific force sequence and angular velocity sequence from the denoised IMU observation sequence, respectively. Representing quaternion multiplication, Indicates the accelerometer in the IMU The deviation vector at time; According to time interval IMU pre-integration term , and The inertia factor is calculated using the following formula. : ; in, This represents the set of denoised IMU observations within the sliding window; This represents the state change from time k to time k+1 obtained by IMU pre-integration in system b; yes The rotation matrix representation; , and These represent the position, velocity, and attitude of frame b relative to frame e at time k+1, respectively. This represents the gravity vector in the e-frame. ; This represents the rotation vector for extracting quaternions. and These represent the deviation vectors of the accelerometer and gyroscope in the IMU at time k+1, respectively.

[0039] The time-varying accelerometer and gyroscope biases are included in the residual vector and are modeled as a random walk process.

[0040] In one specific embodiment, the factor node construction unit, when constructing GNSS factors, is used for: GNSS factors, including pseudorange factor, carrier phase factor, and tropospheric-inter-system deviation factor, are constructed based on the corrected GNSS observation sequence. , is represented as: ; Where G represents the corrected GNSS observation sequence, and g represents a corrected GNSS observation value in the corrected GNSS observation sequence. This represents the GNSS residual term constructed from the corrected GNSS observation sequence. P, L, and T represent the time series and satellite sequence sets of pseudorange, carrier phase, and tropospheric-system bias, respectively, and i and j represent the indices of the observed satellites. and Let i and j represent observation satellites. Represents a robust loss function. This represents the covariance matrix corresponding to the GNSS factor; , and These represent the pseudorange factor, carrier phase factor, and tropospheric-system deviation factor, respectively. , and Let represent the covariance matrices corresponding to the pseudorange factor, carrier phase factor, and tropospheric-system deviation factor at time k, respectively.

[0041] Specifically, this embodiment of the invention uses denoised IMU observation sequences and corrected GNSS observation sequences, combined with ionospheric-free (IF) combined observations and double-difference RTK technology. Most error terms are eliminated sequentially through station-satellite double-difference, and the geometric model is optimized. The double-differenced rover station observation model is as follows: ; In the formula, and Inter-station differential and inter-satellite differential; and These represent the double-difference pseudorange and double-difference carrier phase after the ionosphere-free combination, respectively; Indicates information about observation satellites and The geometric distance between the mobile station and the base station, where, and It consists of the inter-satellite differential directional cosine matrix and the position error of the rover station; and These are the double-difference tropospheric residuals and the double-difference ambiguity between the rover and the base station after IF combination, respectively. Indicates the wavelength after IF combination; and These are double-difference pseudorange observation noise and double-difference carrier phase observation noise, respectively.

[0042] After IF combination and double difference observation pseudo-distance factor at time and carrier phase factor Defined as: ;

[0043] in, This indicates that the satellite is being observed. The single difference between the rover and the base station predicted by INS is shown below. This indicates the observation satellites predicted by INS. With mobile station The distance between them; Indicates observation satellite With the base station The distance between them; This indicates the location of the mobile station as predicted by INS; Indicates observation satellite Location; Indicates the base station Location; Indicates observation of satellites The single difference between the rover and the base station predicted by INS is shown below. This represents the double-difference tropospheric residual; , represents the double-difference pseudo-range after the combination without ionosphere; Indicates information about observation satellites and The geometric distance between the mobile station and the base station; This indicates the noise of the double-difference pseudorange observation; ; in, Indicates wavelength; Indicates the double-difference ambiguity after ionosphere-free combination; , indicating the double-difference carrier phase after ionospheric combination; Indicates the wavelength after IF combination; This represents the double-difference carrier phase observation noise.

[0044] Furthermore, since the receiver clock offsets differ between different GNSS systems, this invention uses the GPS system as a reference and only models other non-reference systems, as follows: ; ; in, This represents the receiver clock bias after taking into account inter-system biases; Indicates the receiver clock bias when using the GPS system as a reference; , and These represent the inter-system bias (ISB) parameters for BDS(C), Galileo(E), and GLONASS(R), respectively.

[0045] The base station receives the GNSS wide-area error model and precise orbit and clock products from the GNSS server, generates accurate tropospheric delay hourly, and broadcasts it to the rover as atmospheric correction. Enhanced information about the current coarse position is obtained using an interpolation method based on an improved linear combination method (MLCM). After generating tropospheric wet delay correction information for the rover, it is incorporated as an additional constraint in the state estimation into the factor graph optimization model. Constraining the estimation of tropospheric wet delay accelerates positioning convergence and improves positioning accuracy, as expressed in: ; ; in, It is a mobile site ( Interpolation tropospheric wet delay correction for the current approximate location; It is a base station ( Tropospheric wet delay information; It corresponds to The interpolation coefficients for each reference station, since this invention only involves one reference station, therefore ; Indicates satellite Initial guesses about the tropospheric wet delay of mobile stations at the first frequency; It is the error of the tropospheric wet delay component in the zenith direction of the mobile station.

[0046] Since the inter-system bias and tropospheric wet delay change slowly in the short term, the ISB parameter and the tropospheric wet delay component are modeled as a random walk process, and the following tropospheric and inter-system bias factors are constructed between adjacent time points: ; in, , and These represent the inter-system bias parameters for BDS, Galileo, and GLONASS, respectively. It is the tropospheric wet delay error component in the zenith direction of the mobile station; k and k+1 represent time k and time k+1, respectively.

[0047] In one specific embodiment, the observability of gyroscope deviations in the vibration environment of electric shovel excavation by the IMU is weak, resulting in a large heading angle deviation. GNSS attitude measurement, on the other hand, has the advantage of accuracy remaining constant over time, and is not limited by field of view or day / night cycles, nor does it suffer from gyroscope drift issues. Therefore, this embodiment of the invention employs a dual-antenna orientation principle, calculating and observing satellite... and The baseline vector is calculated by resolving the phase difference of the two carriers arriving at the two antennas, thereby achieving attitude measurement. The calculation formula is as follows: ;; in, and These represent common-view observation satellites. and The unit observation vector at the receiver. This is the baseline vector between the two antennas of the mobile station; Indicates wavelength; and They represent mobile stations ( The two antennas A and B are related to the observation satellite. and The double-difference ambiguity and carrier phase observation noise. Assume the mobile station antenna A is a known reference point. The approximate coordinates of mobile station antenna B are: Its coordinate correction is Two observation satellites and Using the reference satellite, the linearized form of the double-difference observation equation is: ; in, , and They represent observation satellites. and The line-of-sight vector difference for antenna B. and These represent the approximate positions of antenna B and the observation satellite, respectively. and The geometric distance.

[0048] pass The double-difference observation equation at time t can be solved to obtain This yields the baseline vector. Therefore, the dual-antenna baseline factor... The baseline factor between different antennas A and B can be defined as the dual-antenna baseline factor. , is represented as: ; in, and They represent observation satellites. and The single difference between mobile station antennas A and B; This represents the double-difference tropospheric residual between mobile station antennas A and B; Indicates the wavelength after IF combination; This represents the double-difference ambiguity between mobile station antennas A and B after they are combined without an ionosphere; Indicates observation of satellites and The double-difference carrier phase between mobile station antennas A and B; , and They represent observation satellites. and The unit observation vector at the receiver. This is the baseline vector between the two antennas A and B of the mobile station; Indicates wavelength; and Representing mobile stations The two antennas A and B are related to the observation satellite. and The double-difference ambiguity and carrier phase observation noise. Antennas A and B represent any two antennas of the mobile station.

[0049] In one specific embodiment, due to the significant vibration and impact during electric shovel operation, even with meticulous preprocessing of the data from various sensors, the impact of vibration remains substantial under different working conditions. Therefore, this embodiment of the invention, based on the inherent motion characteristics of the electric shovel, designs corresponding constraint functions considering the shovel's walking, slewing, and digging processes, incorporating these as optimization factors into the optimization framework to improve the stability of the electric shovel's posture under high-dynamic environments. Specifically, the factor node construction unit, when constructing motion constraint factors, is used to: construct motion constraint factors including slewing motion factors, walking motion factors, and digging motion factors. , is represented as: ; in, This represents the set of motion constraint factors within the sliding window. This represents one of the motion constraint factors. This represents the covariance matrix corresponding to the motion constraint factor; This represents the motion constraint residuals constructed based on different motion modes; S, W, and D represent the sets of time series during the rotation, walking, and digging processes, respectively. , and These represent the rotational motion factor, the walking motion factor, and the digging motion factor, respectively. , and Let represent the covariance matrices corresponding to the rotational motion factor, walking motion factor, and mining motion factor at time k, respectively.

[0050] During the slewing process, the vibration of the electric shovel is relatively small. The movement of both the shovel body and the bucket is considered as a fixed-point circular arc motion around the slewing center. The distances from the cab antenna (A2) and bucket antenna (A4) to the slewing center are determined by mechanical parameters. Considering the slight deformation caused by centrifugal force during slewing, a correction amount positively correlated with the slewing angle is introduced. By establishing constraints on horizontal position invariance and yaw angle variation, the deviation of the carrier's horizontal position change between adjacent time points, as well as the deviation between the carrier's yaw angle change and the encoder or gyroscope integral measurement value, are used to characterize these deviations.

[0051] Connecting the current state with the previous state, constraining its horizontal position to remain essentially unchanged, and introducing observations of yaw angle changes, its rotational motion factor is also considered. Defined as: ; in, and These respectively indicate the positions of the cab and the bucket. At any given time, in the e-system direction and The position of direction, and When the cab position and bucket position are at time k+1, under the e-series... direction and The position of direction; This represents the change in yaw angle measured by the IMU from time k to time k+1 during the turn. This represents the change in heading angle observed through dual-antenna coordinated orientation.

[0052] Furthermore, during the movement, it is assumed that the electric shovel moves at a constant speed in a fixed direction, with slow changes in posture and continuous, small-amplitude changes in height. When the shovel chassis moves, posture stability constraints and vertical motion smoothness constraints are established. These constraint functions characterize the deviation between the expected and actual values ​​of the carrier's height change between adjacent time points, as well as the deviation between the carrier's posture change and the expected small changes. Therefore, the walking motion factor... Defined as: ; in, and Let K and K+1 represent the vertical positions of the cab and bucket in the e-frame at time k and time k+1, respectively. This indicates the expected vertical velocity during walking; This represents converting a quaternion into a Lie algebra vector. Indicates the expected change in attitude; and These represent the attitudes of the b-frame relative to the e-frame at times k and k+1, respectively.

[0053] Furthermore, during excavation operations with an electric shovel, the coordinates of the bucket and vehicle body in the carrier coordinate system (b-frame) are calculated using the shovel's forward kinematics model. These coordinates are then transformed to the geocentric rectangular coordinate system (e-frame), and the values ​​of their characteristic points should match the elevation values ​​of the corresponding horizontal positions in the pre-acquired digital elevation model or terrain data. Therefore, the excavation motion factor... Defined as: ; in, , and These represent the three-dimensional coordinates of three feature points—the bucket tip, left side, and right side—in the e-system, respectively. and These represent the bucket tip in the e-series. direction and The position of direction; and These respectively indicate the left side of the bucket in the E series. direction and The position of direction; and These respectively indicate the right side of the bucket in the E series. direction and The position of direction; Indicates the vertical height of the bucket tip in the e-series; This function represents a terrain elevation lookup function. This represents the normal vector of the bottom surface of the bucket. By rotation matrix from Switch to Tie; The surface normal vector at the contact point can be calculated using a digital elevation model (DEM).

[0054] In one specific embodiment, the pose determination module includes: The global optimization unit is used to construct the global optimization objective function of the factor graph optimization model, and solves the global optimization objective function by the maximum a posteriori probability to obtain the optimal state estimate of the global state vector to be estimated. The marginalization unit is used to transform the optimal state estimate and its corresponding observation constraints at the oldest time step in the sliding window into prior information after the global optimization is completed, and add it to the global optimization objective function at the next time step, so as to keep the optimization window size fixed and not lose historical information.

[0055] In this embodiment of the invention, optimal state estimation is achieved through maximum a posteriori (MAP) based on the aforementioned factors and the state nodes within the sliding window. It is assumed that all measurements are independent and that the noise of each measurement follows a zero-mean Gaussian distribution. Therefore, the MAP problem is solved by minimizing the sum of residuals associated with each specific measurement. The size of the sliding window is pre-set, and the ensemble system does not perform joint optimization until the sliding window is completed. The CERES solver (nonlinear least squares solver engine) is used for tightly coupled integration of the proposed state optimization; old IMU states and corresponding features should be marginalized within the sliding window. Finally, the optimal pose estimation result at the current time step and the updated prior constraint information are output.

[0056] The global optimization objective function is expressed as: ; in, It is prior information about the system state, which comes from the marginalization process in the sliding window; Represents the residual factor for each measurement; Represents a robust loss function; This represents the square of the Mahalanobis distance and its corresponding covariance matrix. ; This represents all consecutive IMU observation times within the sliding window. This represents all available GNSS observations. This represents the set of baseline factors constructed from all available GNSS observations. This represents the set of motion constraint factors within the sliding window.

[0057] Furthermore, to limit the computational complexity within the sliding window without losing historical information, this embodiment of the invention also introduces a marginalization process. For the inertia factor, the motion information from the previous moment is converted into a priori constraints for the current moment through the Schur complement, ensuring that subsequent pose estimation can still inherit the motion patterns of the old state. For the GNSS factor, when a visible satellite from the oldest moment is observed in consecutive moments, the corresponding double-difference ambiguity is maintained and further refined through the current observation in the sliding window and the latest priori constraints after marginalization. This requires long-term updates and accumulation of double-difference observations to accurately estimate and continuously improve the carrier phase double-difference ambiguity. However, when the satellite signal observed in the oldest moment is blocked in other moments, or when the corresponding ambiguity is reinitialized due to cycle slips, the double-difference ambiguity is unobservable and is thus marginalized along with the oldest time-varying state. For the motion constraint factor, when a single-state constraint is marginalized, its constraint information is passed to the subsequent state through the Schur complement, and the logic of the double-state constraint is consistent with that of the IMU factor.

[0058] In summary, the system provided in this invention addresses the problems of strong vibration, strong impact, and easy obstruction of satellite signals in open-pit mine electric shovel operations by proposing a tightly coupled positioning system. First, based on a diffusion model, the IMU observation sequence to be denoised is cleaned, and a three-stage GNSS threshold system is established for the electric shovel's unique operating conditions (walking, turning, digging) to obtain high-quality observations. Second, a factor graph optimization model is constructed, including inertia factors, GNSS factors, dual-antenna baseline factors, and electric shovel-specific motion constraint factors. Finally, the optimal electric shovel pose is solved through sliding window optimization. The main innovations of this invention are: 1. Utilizing a diffusion model to learn the IMU noise distribution of the electric shovel under strong vibration conditions to achieve targeted denoising. 2. Constructing motion constraint factors (turning center constraint, walking elevation smoothing, digging terrain matching) that conform to the mechanical characteristics of the electric shovel, effectively suppressing drift when GNSS signal lock-off occurs.

[0059] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A high-precision pose positioning system for an open-pit mine electric shovel based on factor graph optimization, characterized in that, include: The data acquisition module is used to acquire the original GNSS observation sequence and the original IMU observation sequence pre-installed on the electric shovel by the antenna and IMU, and to perform spatiotemporal alignment on the original GNSS observation sequence and the original IMU observation sequence to obtain the GNSS observation sequence to be corrected and the IMU observation sequence to be denoised. The denoising module is used to denoise the IMU observation sequence to be denoised using a pre-trained denoising diffusion model, so as to obtain the denoised IMU observation sequence. The working condition identification module is used to identify the current working condition of the electric shovel based on the GNSS observation sequence to be corrected and the denoised IMU observation sequence; The GNSS data processing module is used to correct the GNSS observation sequence to be corrected according to the current working condition of the electric shovel, and obtain the corrected GNSS observation sequence. The model building module is used to construct a multi-source fusion factor graph optimization model based on the denoised IMU observation sequence and the corrected GNSS observation sequence. The pose determination module is used to solve the factor graph optimization model to obtain the electric shovel pose.

2. The high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization according to claim 1, characterized in that, The operating condition identification module includes: The statistical feature calculation unit is used to calculate the horizontal velocity, rate of change of heading angle, and variance of vertical acceleration of the electric shovel at the current moment based on the denoised IMU observation sequence. The walking condition determination unit is used to determine the current working condition of the electric shovel as the walking condition when it is determined that the horizontal speed of the electric shovel at the current moment is greater than the preset horizontal speed threshold, the rate of change of the heading angle at the current moment is less than the preset rate of change of the heading angle threshold, and the variance of the vertical acceleration of the electric shovel at the current moment is less than the preset vertical acceleration variance threshold. The slewing condition determination unit is used to determine the current working condition of the electric shovel as a slewing condition when the current horizontal speed of the electric shovel is less than or equal to a preset horizontal speed threshold and the current heading angle change rate is greater than or equal to a preset heading angle change rate threshold. The excavation condition determination unit is used to determine the current working condition of the electric shovel as excavation condition when the signal-to-noise ratio of the electric shovel at the current moment in the GNSS observation sequence to be corrected continuously decreases or is interrupted.

3. The high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization according to claim 2, characterized in that, The GNSS observation sequence to be corrected includes a pseudorange sequence to be corrected, and the GNSS data processing module includes: The pseudorange correction unit for the walking mode is used to calculate the pseudorange mean and pseudorange standard deviation of a first preset number of pseudoranges in the pseudorange sequence to be corrected when the current working mode of the electric shovel is the walking mode. Based on the pseudorange mean and pseudorange standard deviation, it determines whether the pseudorange is an abnormal pseudorange. If the pseudorange is an abnormal pseudorange, it replaces the pseudorange with the previous pseudorange. After obtaining the pseudorange sequence after preliminary correction, it fits the pseudorange sequence after preliminary correction through a quadratic polynomial fitting model to obtain the corrected pseudorange as the corrected GNSS observation sequence. The slewing condition pseudorange correction unit is used to perform pseudorange smoothing correction on any pseudorange in the pseudorange sequence to be corrected when the current working condition of the electric shovel is slewing condition, based on the second preset number of pseudoranges in its neighborhood and the corresponding carrier phase, to obtain the corrected pseudorange sequence as the corrected GNSS observation sequence. The excavation-condition pseudorange correction unit is used to interpolate the pseudorange sequence to be corrected when the current working condition of the electric shovel is excavation, to obtain the interpolated pseudorange sequence, and to calculate the position change of the electric shovel through the denoised IMU observation sequence. Based on the preliminary corrected pseudorange sequence and the position change of the electric shovel, the corrected pseudorange sequence is calculated as the corrected GNSS observation sequence.

4. The high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization according to claim 1, characterized in that, The model building module includes: The state definition unit is used to define the global state vector to be estimated in the factor graph optimization model. The factor node construction unit is used to construct inertial factor, GNSS factor, dual-antenna baseline factor, and motion constraint factor based on the denoised IMU observation sequence, the corrected GNSS observation sequence, and the sliding window. The model building unit is used to combine the global state vector to be estimated, the inertia factor, the GNSS factor, the dual-antenna baseline factor, and the motion constraint factor to obtain the factor graph optimization model.

5. The high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization according to claim 4, characterized in that, The global state vector to be estimated Represented as: ； ； in, Indicates the size of the sliding window. This represents the state at time k. , and Let b represent the position, velocity, and attitude of the b-frame relative to the e-frame at time k, respectively. The b-frame is the carrier coordinate system, and the e-frame is the geocentric Cartesian coordinate system. and These represent the deviation vectors of the accelerometer and gyroscope in the IMU at time k, respectively. This represents the inter-system deviation vector of GNSS systems at time k. This represents the tropospheric wet delay error component in the zenith direction of the mobile station at time k. and These represent inter-station differential and inter-satellite differential, respectively. This represents the integer ambiguity after double difference. This indicates the number of consecutive ambiguities within the sliding window. and These represent the rover and the base station, respectively, with IF indicating an ionospheric-free combination. This represents the double-difference ambiguity between the rover and the base station after the ionosphere is removed.

6. The high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization according to claim 5, characterized in that, The factor node construction unit is used to: Based on the denoised IMU observation sequence, the time interval is recursively calculated using the following formula. IMU pre-integration term , and : ； in, , The time interval between adjacent observations in the denoised IMU observation sequence; , and Time interval The IMU pre-integration term within, , and Time interval The IMU pre-integration term within, It is a rotation matrix representation of quaternions; and These are the specific force sequence and angular velocity sequence from the denoised IMU observation sequence, respectively. Representing quaternion multiplication, Indicates the accelerometer in the IMU The deviation vector at time; According to time interval IMU pre-integration term , and The inertia factor is calculated using the following formula. : ； in, This represents the set of denoised IMU observations within the sliding window; This represents the state change from time k to time k+1 obtained by IMU pre-integration in system b; yes The rotation matrix representation; , and These represent the position, velocity, and attitude of frame b relative to frame e at time k+1, respectively. This represents the gravity vector in the e-frame. ; This represents the rotation vector for extracting quaternions. and These represent the deviation vectors of the accelerometer and gyroscope in the IMU at time k+1, respectively.

7. The high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization according to claim 5, characterized in that, The factor node construction unit is used to: GNSS factors, including pseudorange factor, carrier phase factor, and tropospheric-inter-system deviation factor, are constructed based on the corrected GNSS observation sequence. , is represented as: ； Where G represents the corrected GNSS observation sequence, and g represents a corrected GNSS observation value in the corrected GNSS observation sequence. This represents the GNSS residual term constructed from the corrected GNSS observation sequence. P, L, and T represent the time series and satellite sequence sets of pseudorange, carrier phase, and tropospheric-system bias, respectively, and i and j represent the indices of the observed satellites. and Let i and j represent observation satellites. Represents a robust loss function. This represents the covariance matrix corresponding to the GNSS factor; , and These represent the pseudorange factor, carrier phase factor, and tropospheric-system deviation factor, respectively. , and These represent the covariance matrices corresponding to the pseudorange factor, carrier phase factor, and tropospheric-system deviation factor at time k, respectively. ； in, This indicates that the satellite is being observed. The single difference between the rover and the base station predicted by INS is shown below. This indicates the observation satellites predicted by INS. With mobile station The distance between them; Indicates observation satellite With the base station The distance between them; This indicates the location of the mobile station as predicted by INS; Indicates observation satellite Location; Indicates the base station Location; Indicates observation of satellites The single difference between the rover and the base station predicted by INS is shown below. This represents the double-difference tropospheric residual; , represents the double-difference pseudo-range after the combination without ionosphere; Indicates information about observation satellites and The geometric distance between the mobile station and the base station; This indicates the noise of the double-difference pseudorange observation; ; in, Indicates wavelength; Indicates the double-difference ambiguity after ionosphere-free combination; , indicating the double-difference carrier phase after ionospheric combination; Indicates the wavelength after IF combination; This indicates double-difference carrier phase observation noise; ； in, , and These represent the inter-system bias parameters for BDS, Galileo, and GLONASS, respectively. It is the tropospheric wet delay error component in the zenith direction of the mobile station; k and k+1 represent time k and time k+1, respectively.

8. The high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization according to claim 5, characterized in that, The factor node construction unit is used to: Dual-antenna baseline factors were constructed based on the corrected GNSS observation sequence. , is represented as: ； in, and They represent observation satellites. and The single differential signal between mobile station antennas A and B; This represents the double-difference tropospheric residual between mobile station antennas A and B; Indicates the wavelength after IF combination; This represents the double-difference ambiguity between mobile station antennas A and B after they are combined without an ionosphere; Indicates observation of satellites and The double-difference carrier phase between mobile station antennas A and B; , and They represent observation satellites. and The unit observation vector at the receiver. This is the baseline vector between the two antennas A and B of the mobile station; Indicates wavelength; and Representing mobile stations The two antennas A and B are related to the observation satellite. and The double-difference ambiguity and carrier phase observation noise.

9. The high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization according to claim 5, characterized in that, The factor node construction unit is used to: Construct motion constraint factors including rotational motion factors, walking motion factors, and mining motion factors. , is represented as: ； in, This represents the set of motion constraint factors within the sliding window. This represents one of the motion constraint factors. This represents the covariance matrix corresponding to the motion constraint factor; This represents the motion constraint residuals constructed based on different motion modes; S, W, and D represent the sets of time series during the rotation, walking, and digging processes, respectively. , and These represent the rotational motion factor, the walking motion factor, and the digging motion factor, respectively. , and Let represent the covariance matrices corresponding to the rotational motion factor, walking motion factor, and mining motion factor at time k, respectively; ； in, and These respectively indicate the positions of the cab and the bucket. At any given time, in the e-system direction and The position of direction, and When the cab position and bucket position are at time k+1, under the e-series... direction and The position of direction; This represents the change in yaw angle measured by the IMU from time k to time k+1 during the turn. This represents the change in heading angle observed through coordinated orientation using two antennas; ； in, and Let K and K+1 represent the vertical positions of the cab and bucket in the e-frame at time k and time k+1, respectively. This indicates the expected vertical velocity during walking; This represents converting a quaternion into a Lie algebra vector. Indicates the expected change in attitude; and Let b represent the attitude of the b-frame relative to the e-frame at time k and k+1, respectively. ； in, , and These represent the three-dimensional coordinates of three feature points—the bucket tip, left side, and right side—in the e-system, respectively. and These respectively indicate the bucket tip in the e-series. direction and The position of direction; and These respectively indicate the left side of the bucket in the E series. direction and The position of direction; and These respectively indicate the right side of the bucket in the E series. direction and The position of direction; Indicates the vertical height of the bucket tip in the e-series; This function represents a terrain elevation lookup function. This represents the normal vector of the bottom surface of the bucket. By rotation matrix from Switch to Tie; This represents the normal vector of the terrain surface at the contact point.

10. The high-precision pose positioning system for open-pit mine electric shovels based on factor graph optimization according to claim 1, characterized in that, The pose determination module includes: The global optimization unit is used to construct the global optimization objective function of the factor graph optimization model, and solve the global optimization objective function by the maximum a posteriori probability to obtain the optimal state estimate of the global state vector to be estimated. The marginalization unit is used to transform the optimal state estimate and its corresponding observation constraints at the oldest time step in the sliding window into prior information after the global optimization is completed, and add it to the global optimization objective function at the next time step.

Images