A monorail crane real-time map construction method and system for a downhole roadway environment

Abstract

The present application belongs to the field of automatic driving and robot navigation technology, and particularly relates to a real-time map construction method and system for a monorail crane in a downhole roadway environment, which method comprises: acquiring vehicle-mounted sensor data and correcting motion distortion using an inertial measurement unit to generate standardized laser ranging data; constructing a discretized spherical sensitivity field, identifying a longitudinal degeneration vector through nonlinear mapping optimization, and generating a degeneration risk coefficient; mapping point cloud data to the longitudinal degeneration vector direction to generate a real-time density waveform, and calculating the best waveform offset by aligning with a reference density waveform; constructing a target function containing a longitudinal waveform locking constraint term weighted by the degeneration risk coefficient, solving the globally optimal pose, and completing map construction. The present application solves the problem of longitudinal positioning drift in a weak geometric structure roadway and achieves high-precision map construction.

Description

Technical Field

[0001] This invention relates to the fields of autonomous driving and robot navigation technology. More specifically, this invention relates to a method and system for constructing real-time maps of monorail cranes for underground tunnel environments. Background Technology

[0002] The monorail crane in coal mines is a core piece of equipment in the underground auxiliary transportation system. With the advancement of smart mine construction, the need for unmanned driving and high-precision autonomous navigation of monorail cranes using onboard sensors has become urgent in underground environments with no GNSS signals and unstable network communication. Currently, Simultaneous Localization and Mapping (SLAM) technology based on lidar is the mainstream solution to this problem. Its core algorithm mainly solves the vehicle's pose change by matching geometric features in the environment, that is, minimizing the residual distance between the current scanned point cloud and the feature points of the local map.

[0003] However, underground coal mine roadways are a typical degraded environment, characterized by a long corridor structure and high translational invariance along the roadway's direction. LiDAR data is mostly parallel to the roadway's sidewalls and track, providing strong lateral and vertical constraints, but severely lacking geometric constraints in the longitudinal direction (the direction of travel). This leads to degradation of the pose optimization equations in the longitudinal dimension. While existing technologies typically incorporate pre-integrated results from Inertial Measurement Units (IMUs) as prior constraints to mitigate degradation, in long-distance, geometrically weak straight roadway scenarios, the IMU's own bias and noise accumulate over time, making it impossible to effectively correct accumulated errors through geometric matching. This results in severe longitudinal positioning drift, causing the constructed map to stretch, deform, or break along its length, failing to meet practical application requirements. Summary of the Invention

[0004] To address the technical problem of longitudinal positioning drift caused by the degradation of geometric features in underground roadways with weak geometric structures, resulting in the stretching, deformation, or breakage of the constructed map along its length, this invention provides solutions in the following aspects.

[0005] In a first aspect, the present invention provides a method for constructing a real-time map of a monorail crane in an underground tunnel environment, comprising: generating standardized laser ranging data based on onboard sensor data; generating a set of discrete direction vectors on a unit sphere, and obtaining geometric sensitivity for each discrete direction vector in the set; obtaining a longitudinal degradation vector and a degradation risk coefficient based on the geometric sensitivity of all discrete direction vectors; mapping the standardized laser ranging data onto a straight line determined by the longitudinal degradation vector to generate a real-time density waveform; extracting a reference density waveform from a local map; obtaining signal correlation based on the real-time density waveform and the reference density waveform, and obtaining an optimal waveform offset by maximizing the signal correlation; obtaining an objective function containing the degradation risk coefficient, the longitudinal degradation vector, and the optimal waveform offset, solving for the global optimal pose of the monorail crane by minimizing the objective function, and completing map construction based on the global optimal pose.

[0006] This invention generates a set of discrete direction vectors on a unit sphere and calculates geometric sensitivity to identify longitudinal degradation vectors in underground tunnel environments and generate degradation risk coefficients. Then, using these longitudinal degradation vectors, standardized laser ranging data is reduced in dimensionality to generate real-time density waveforms distributed along the tunnel direction. The optimal waveform offset is calculated by matching these waveforms with reference density waveforms in a local map. Finally, a nonlinear least-squares objective function is constructed, including a longitudinal waveform locking constraint term weighted by the degradation risk coefficient. This introduces implicit geometric constraints related to the periodic distribution of tunnel facilities when solving for the monorail crane's pose, effectively suppressing longitudinal positioning drift caused by geometric feature degradation in long-distance, weakly geometric tunnel environments. This improves the accuracy and closed-loop consistency of underground tunnel environment map construction.

[0007] Preferably, the step of generating standardized laser ranging data based on vehicle-mounted sensor data includes: aligning the acquisition time of each frame of data from the vehicle-mounted sensor's 3D LiDAR to the time axis of the inertial measurement unit using hardware triggering; calculating the pose change of the 3D LiDAR relative to the start time of the frame at each sampling time during the scanning process using the angular velocity and acceleration measured by the inertial measurement unit; and based on the pose change, inversely mapping all laser points acquired at different times within a frame back to the coordinate system at the start time of the frame through coordinate transformation to generate standardized laser ranging data.

[0008] This invention utilizes high-frequency angular velocity and acceleration data collected by an inertial measurement unit rigidly connected to a 3D lidar to calculate the pose change of the 3D lidar at each sampling moment relative to the starting moment during the scanning of a frame. Based on this, the laser points collected at different moments are mapped back to the coordinate system of the starting moment. This eliminates the layering or distortion of the laser point cloud caused by the mechanical shaking of the monorail crane during startup or when passing through the track joint due to the flexible suspension method. This ensures that the standardized laser ranging data used for map construction can truly reflect the geometry of the underground tunnel.

[0009] Preferably, the geometric sensitivity satisfies the expression: ;in, Indicates the first Geometric sensitivity of discrete direction vectors; It represents the total number of laser points in the current frame of normalized laser ranging data; It is the first The geometric normal vector of each laser point; It is the first A discrete direction vector; Represents the dot product operation of vectors; It is a nonlinear mapping kernel function; Represents the absolute value symbol.

[0010] This invention evaluates the contribution of geometric features to movement constraints by calculating the geometric sensitivity of discrete direction vectors and using the mapping components of the geometric normal vector of the laser point on the discrete direction vectors. It also introduces a nonlinear mapping kernel function to process the mapping results. While evaluating the geometric constraint capability, it suppresses the interference of dust and minor noise in the underground environment and amplifies the constraint weight of effective geometric features. As a result, it can more robustly identify the longitudinal degradation vector that is easily lost in the positioning of monorail cranes in complex underground roadway environments.

[0011] Preferably, the degradation risk coefficient satisfies the expression: ;in, It is the degradation risk coefficient; It is the minimum geometric sensitivity corresponding to the direction of the longitudinal degradation vector; It is the maximum geometric sensitivity among all discrete direction vectors; It is a minimum constant to prevent the denominator from being zero.

[0012] This invention generates a degradation risk coefficient by comparing the ratio of the minimum geometric sensitivity corresponding to the longitudinal degradation vector direction to the maximum geometric sensitivity in all directions. This enables the system to assess the degree of geometric degradation of the current tunnel environment, allowing it to adaptively adjust the weight of the longitudinal waveform locking constraint in pose optimization based on the richness of environmental features. Specifically, it relies on conventional matching when features are abundant, but forces waveform correction when the corridor degradation effect is severe, ensuring that the monorail crane can maintain positioning stability in tunnel sections with different geometric features.

[0013] Preferably, the step of mapping standardized laser ranging data onto a straight line determined by a longitudinal degradation vector to generate a real-time density waveform includes: mapping standardized laser ranging data onto a straight line determined by a longitudinal degradation vector; discretizing the straight line into a series of small spatial intervals; statistically summing the weighted reflection intensities of laser points falling into each interval; and generating a real-time density waveform distributed along the tunnel direction.

[0014] This invention maps standardized laser ranging data onto a straight line determined by a longitudinal degradation vector and discretizes the line into tiny spatial intervals. The weighted sum of the reflection intensities of laser points falling into each interval is statistically analyzed to generate a real-time density waveform distributed along the tunnel direction. This method can compress sparse and difficult-to-match features of facilities such as anchor bolt rows, metal arches, or cable hooks in three-dimensional space into a one-dimensional signal with a strong periodic rhythm, thereby extracting highly recognizable features for longitudinal positioning in long, straight underground tunnels with simple geometric features.

[0015] Preferably, the signal correlation satisfies the expression: ;in, Assuming the candidate waveform offset is The function value of signal correlation at time; It is a spatial location variable along the direction of the alleyway; It is the lower limit position of the waveform analysis window; This is the upper limit position of the waveform analysis window; It is a translation of the candidate waveform offset. The subsequent real-time density waveform; This is the reference density waveform.

[0016] Preferably, the objective function satisfies the expression: ;in, It is the objective function; It is the position and orientation of the monorail crane; It is the residual distance between the laser point and the surface; It is the covariance matrix of laser measurements; It is the pre-integrated residual of the inertial measurement unit; It is the covariance matrix of the inertial measurement unit measurements; It is the degradation risk coefficient; It is the position vector of the monorail crane at the current moment that needs to be solved; It is the predicted position vector of the monorail crane from the previous moment; It is the longitudinal degenerate vector; It is the waveform offset; Represents the dot product operation of vectors; It is the matrix transpose; It is the vector magnitude; It is the standard deviation of the longitudinal waveform matching measurement.

[0017] Preferably, the step of completing map construction based on the global optimal pose includes: converting the standardized laser ranging data of the current frame to the global coordinate system based on the global optimal pose; and fusing the converted data into a global point cloud map or a probabilistic grid map to complete map construction.

[0018] Preferably, generating a set of discrete direction vectors on a unit sphere includes: generating multiple discrete direction vectors using the Fibonacci lattice algorithm, such that the discrete direction vectors are isotropic and uniformly distributed in space, forming a set of discrete direction vectors.

[0019] Secondly, the present invention provides a real-time map construction system for monorail cranes in underground tunnel environments, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned method for constructing real-time maps of monorail cranes in underground tunnel environments is implemented.

[0020] By adopting the above technical solution, a computer program is generated from the above-mentioned method for constructing real-time maps of monorail cranes for underground roadway environments, and stored in a memory for loading and execution by a processor. This allows for the creation of terminal devices based on the memory and processor, making them convenient to use.

[0021] The beneficial effects of this invention are as follows:

[0022] This invention identifies the longitudinal degradation vector in the tunnel environment and generates a degradation risk coefficient by constructing a discrete spherical sensitivity field. Then, it maps standardized laser ranging data to the direction of the longitudinal degradation vector to generate a real-time density waveform along the tunnel direction. The optimal waveform offset is calculated using a signal correlation function and introduced into a nonlinear least squares objective function that includes the pose state of the monorail crane. This allows for the excavation of implicit periodic structural constraints composed of anchor bolts or arches in long straight tunnels lacking obvious geometric features. This solves the problem of longitudinal positioning drift of the monorail crane in weak geometric environment due to geometric feature degradation and ensures the global consistency of map construction.

[0023] This invention utilizes high-frequency angular velocity and acceleration data collected by an inertial measurement unit rigidly connected to a 3D lidar for motion distortion correction. By integrating and calculating subtle pose changes within the radar scanning cycle, and mapping the laser points collected at each moment back to the coordinate system of the initial moment, it effectively eliminates the layering or distortion of the laser point cloud caused by irregular swaying and bumping during the operation of the monorail crane due to the flexible chain suspension method. This ensures that the standardized laser ranging data used for map construction can truly and accurately reflect the geometric shape of the underground tunnel environment, and improves the matching accuracy of the front-end odometer.

[0024] This invention generates a degradation risk coefficient by calculating the ratio of minimum geometric sensitivity to maximum geometric sensitivity, and introduces it as a dynamic weight adjuster into the longitudinal waveform locking constraint term of the nonlinear least squares objective function. This enables adaptive perception and fusion strategy adjustment of the degree of geometric degradation in the tunnel environment. Specifically, it mainly relies on laser geometric matching in feature-rich sections, while forcibly adopting waveform locking correction in degraded sections of the corridor. This allows the monorail crane to adapt to the complex and ever-changing tunnel environment underground, achieving robust positioning and mapping of the entire tunnel section while ensuring lateral and vertical positioning accuracy. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a method for constructing a real-time map of a monorail crane for underground tunnel environments according to the present invention;

[0026] Figure 2 This is a schematic diagram illustrating the construction of a discretized spherical sensitivity field and the identification of the longitudinal degradation vector;

[0027] Figure 3 This is a schematic diagram illustrating the matching of real-time density waveforms and reference density waveforms along the direction of the tunnel.

[0028] Figure 4 This is a schematic diagram illustrating the longitudinal positioning drift comparison in a long-distance tunnel environment. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0031] This invention discloses a method for constructing real-time maps of monorail cranes in underground tunnel environments, referring to... Figure 1 This includes steps S1-S4:

[0032] S1. Generate standardized laser ranging data based on vehicle-mounted sensor data.

[0033] It should be noted that the monorail crane operates using a flexible chain suspension system. During startup, braking, or crossing track joints, the monorail crane inevitably experiences irregular swaying and bumping. This mechanical swaying causes drastic changes in pose of the 3D LiDAR within the scanning cycle of a single frame, resulting in distortions such as layering, ghosting, or warping in the scanned environmental point cloud. If these distortions caused by the monorail crane's movement are not eliminated, subsequent map construction will be based on incorrect geometric data, severely impacting positioning accuracy and mapping quality. Therefore, this invention utilizes high-frequency inertial data to compensate for the motion of low-frequency LiDAR data, restoring the distorted point cloud to its true geometric form.

[0034] Specifically, a 3D LiDAR mounted on a monorail crane continuously scans the tunnel environment to acquire raw point cloud data. An inertial measurement unit (IMU) rigidly connected to the 3D LiDAR collects high-frequency angular velocity and acceleration data from the monorail crane. Using hardware triggering, the acquisition time of each frame of data from the 3D LiDAR is aligned to the time axis of the IMU. Using the angular velocity and acceleration measured by the IMU, the pose change of the 3D LiDAR at each sampling moment relative to the start of the frame is calculated through integration. Based on this pose change, all laser points acquired at different times within a frame are inversely mapped back to the coordinate system of the start of the frame through coordinate transformation, thus eliminating the influence of the monorail crane's motion on the scanning imaging and generating standardized laser ranging data.

[0035] S2. Generate a set of discrete direction vectors on the unit sphere. For each discrete direction vector in the set, obtain the geometric sensitivity. Based on the geometric sensitivity of all discrete direction vectors, obtain the longitudinal degradation vector and degradation risk coefficient.

[0036] It should be noted that the "degradation" in this invention refers to the fact that, due to the corridor-like structure of the underground tunnel, the geometric distribution of the point cloud observed by the 3D lidar in the direction of vehicle movement remains almost unchanged. This makes the influence of the pose increment in this direction on the point cloud residual approach zero, resulting in a decrease in the rank of the Jacobian matrix for pose solving in this direction, leading to unobservability. This invention collectively refers to phenomena such as insufficient geometric constraints, point cloud non-responsiveness to displacement changes, and ill-conditioned optimization equations in this direction as "geometric degradation."

[0037] The "longitudinal degradation vector" refers to the spatial direction that minimizes the geometric sensitivity of the point cloud, meaning that when a vehicle moves along this direction, the distribution of the point cloud in three-dimensional space hardly changes. This vector reflects the direction in which laser geometry cannot provide effective constraints in the current environment, and it is usually approximately consistent with the direction of the tunnel.

[0038] In complex underground environments, environmental noise such as dust and locally accumulated materials often does not conform to an ideal Gaussian distribution. This makes traditional matrix factorization methods based on global statistics susceptible to local disturbances and biases, making it difficult to accurately identify the longitudinal degradation vector of the tunnel. This invention seeks a direction in a discrete spherical space that minimizes sensitivity to environmental geometric constraints. This direction is precisely the direction indicated by the longitudinal degradation vector that is easily lost during monorail crane positioning, thus improving robustness while maintaining computational efficiency.

[0039] Specifically, a set of discrete direction vectors that are uniformly distributed is generated on the unit sphere. The Fibonacci lattice algorithm is used to generate A set of discrete direction vectors is given, making these vectors isotropic and uniformly distributed in space. For each laser point in the current frame of standardized laser ranging data, plane fitting or curvature calculation is performed using its neighborhood point set to obtain the geometric normal vector of each laser point. Each discrete direction vector in Calculate its geometric sensitivity, which is used to evaluate the monorail crane's movement along... The degree of change in geometric features observed by the 3D lidar when the direction is moved.

[0040] The geometric sensitivity of any discrete direction vector satisfies the expression:

[0041]

[0042] In the formula, It represents the total number of laser points in the current frame of normalized laser ranging data; It is the first The geometric normal vector of each laser point is obtained by analyzing the geometric features of the local point cloud. It is the first A discrete direction vector is generated by the Fibonacci lattice algorithm; Represents the dot product operation of vectors; It is a nonlinear enhancement function; Represents the absolute value symbol.

[0043] In the formula: using dot product operation Calculate the mapping component of the geometric normal vector of the laser point onto the discrete direction vector. This component physically represents the pair of laser points. The constraint contribution of directional movement; when the geometric normal vector of the laser point is perpendicular to the discrete direction vector, the dot product is zero, indicating that the laser point cannot be constrained to move in that direction; to suppress disturbances from abnormal points such as dust and local accumulations, this invention applies a nonlinear enhancement function to the dot product result. ,like This suppresses the contribution of large noise while keeping the true geometric constraints effective, resulting in a more stable estimate of geometric sensitivity.

[0044] It should be further explained that, regarding the selection of the power exponent in the nonlinear mapping kernel function, this embodiment preferably uses 1.5, and its reasonable value range is usually between 1.3 and 1.7. The reason for setting this range is that the dot product of the feature point normal vector and the probe direction reflects the geometric constraint ability of the point on the movement. If the exponent is set to 1, it is impossible to effectively distinguish between high-quality vertical constraints and low-quality wiping noise, resulting in accumulated noise overwhelming the key features. If the exponent is set too high, such as 2, although it can strongly suppress noise, it will also excessively weaken the non-vertical but still constraining tilted features. Therefore, the selection of the range of 1.3 to 1.7 aims to construct a soft threshold, which can nonlinearly compress the weights of low-confidence data to suppress noise, while retaining the effective contribution of tilted features, thereby obtaining the best robustness in unstructured environments.

[0045] Further explanation is needed regarding the number of discrete direction vectors. Its preferred value range is 60 to 100. If If the value is too small, the discrete direction vectors will be too sparsely distributed on the sphere, which may cause the true longitudinal degradation vector to be located between two discrete direction vectors and thus missed, reducing search accuracy; if While a large value improves coverage accuracy, it significantly increases the computational load of subsequent geometric sensitivity calculations, affecting the system's real-time performance. This embodiment selects 80 discrete direction vectors, striking a balance between computational efficiency and search accuracy while ensuring the capture of the main geometric features of the roadway.

[0046] Furthermore, the discrete direction vector with the lowest score is used as the initial reference discrete direction vector; a small search area is set around this reference discrete direction vector, which is set to cover the blind zone between adjacent discrete vectors; a longitudinal degradation vector is obtained by using the gradient descent method, which indicates the direction in which the monorail crane is most likely to drift in the current tunnel environment.

[0047] Furthermore, the degradation risk coefficient of the current environment is obtained, which satisfies the expression:

[0048]

[0049] In the formula, It is the degradation risk coefficient; It is the longitudinal degeneracy vector Minimum geometric sensitivity corresponding to the direction; It is the maximum geometric sensitivity among all discrete direction vectors; It is a minimum constant to prevent the denominator from being zero.

[0050] In the formula, the ratio of minimum geometric sensitivity to maximum geometric sensitivity is used to evaluate the geometric constraint capability of the current roadway environment on the vehicle's pose in each direction. When the vehicle enters a roadway section with long, straight, parallel sidewalls and lacking lateral or longitudinal geometric changes, the geometric observability of the point cloud in the longitudinal direction decreases sharply. It will approach 0, making the degradation risk coefficient... Approaching 1. In this scenario, laser matching cannot effectively limit longitudinal drift, and the system needs to enhance longitudinal displacement observation based on density waveforms to avoid the continuous accumulation of IMU drift; conversely, when the vehicle is in an area with rich structural changes, such as a lane with turns, dense arched equipment, obvious sidewall concavity or diverse geometric directions, the sensitivity in each direction is more balanced, and the system can effectively restore the geometric constraints in the longitudinal direction. At this time, the coefficient approaches 0.

[0051] For example, Figure 2 This is a schematic diagram of the construction of a discretized spherical sensitivity field and the identification of the longitudinal degradation vector. The red arrow indicates the direction of the longitudinal degradation vector, and the color intensity of the scattered points on the sphere represents the geometric sensitivity of each discrete directional vector.

[0052] S3. Map the standardized laser ranging data onto the straight line determined by the longitudinal degradation vector to generate a real-time density waveform; extract the reference density waveform from the local map; obtain the signal correlation based on the real-time density waveform and the reference density waveform, and obtain the optimal waveform offset by maximizing the signal correlation.

[0053] It should be noted that although underground roadways exhibit translational invariance in their geometric structure, they are equipped with a large number of facilities with fixed intervals, such as anchor bolt rows, metal arches, or cable hooks. These facilities exhibit a strong periodic rhythm in their distribution along the roadway direction. This invention utilizes this characteristic to compress standardized laser ranging data into a one-dimensional density waveform along the direction of the longitudinal degradation vector. By comparing the phase difference between the real-time density waveform and the reference density waveform, the longitudinal displacement of the monorail crane is calculated, thereby solving the longitudinal positioning drift problem in the roadway environment.

[0054] Specifically, the normalized laser ranging data of the current frame is mapped to a vector derived from the longitudinal degradation vector. On the determined straight line; the line is discretized into a series of small spatial intervals, and the weighted sum of the reflection intensity of laser points falling into each interval is calculated to generate a real-time density waveform distributed along the tunnel direction; taking the predicted position of the monorail crane at the previous moment as the center, a local point cloud is extracted from the constructed local map, and a reference density waveform is generated using the same method; signal correlation is constructed to search for the optimal waveform offset, and the signal correlation satisfies the expression:

[0055]

[0056] In the formula, Assuming the candidate waveform offset is Signal correlation at time; It is a spatial location variable along the direction of the alleyway; and These are the lower and upper limits of the waveform analysis window, respectively. It is a translation of the candidate waveform offset. The subsequent real-time density waveform; This is the reference density waveform.

[0057] Where: The degree of overlap between the translated real-time density waveform and the reference density waveform in the spatial domain is calculated; when the candidate waveform offset... When the peak of the real-time density waveform at the location of the corresponding anchor bolt or other facility is perfectly aligned with the peak of the reference density waveform, the product integral is... It will reach its maximum value at this time. This is the optimal waveform offset. This offset represents the adjustment made to align the currently observed periodicity of the environment with known geographical features. Figure 1 Therefore, the physical correction distance that the monorail crane should move in the longitudinal degradation vector direction.

[0058] It should be further added that the waveform analysis window is derived from the local map. The system obtains the predicted pose of the current frame based on the forward integration of the IMU and the optimization results of the previous frame. Using the predicted position as the center point, a fixed-length point cloud region is truncated in the local map along the longitudinal degradation vector direction. The point cloud of this region is mapped to the longitudinal degradation vector direction to form a reference density waveform. This truncated interval is the waveform analysis window, which is used to perform similarity matching with the real-time waveform generated in the current frame.

[0059] Regarding the discretization of the straight line into a series of tiny spatial intervals, the empirical value for the length of these intervals is set to 10 centimeters. The selection of this length parameter matches the physical dimensions and spacing characteristics of support facilities such as anchor bolts and metal arches in the underground roadway. If the spatial interval is set too small, the generated real-time density waveform is easily affected by the local sparsity of the laser points and random ranging noise, resulting in high-frequency jitter that has no physical meaning and affecting the stability of waveform matching. If the spatial interval is set too large, it will excessively smooth out the periodic structural peak features with positioning value, reducing the resolution and accuracy of longitudinal position locking.

[0060] For example, Figure 3 This is a schematic diagram of the matching between the real-time density waveform and the reference density waveform distributed along the direction of the tunnel. The red curve represents the real-time density waveform generated in the current frame, and the black dashed line represents the reference density waveform extracted from the local map. The phase alignment process between the two is used to calculate the optimal waveform offset.

[0061] S4. Obtain the objective function that includes the degradation risk coefficient, the longitudinal degradation vector, and the optimal waveform offset. Solve the global optimal pose of the monorail crane by minimizing the objective function, and complete the map construction based on the global optimal pose.

[0062] It should be noted that the positioning of a monorail crane is a multi-source information fusion process, which requires the simultaneous use of the high-precision lateral ranging capability of 3D lidar, the smooth calculation capability of inertial measurement unit in a short time, and the longitudinal constraint capability provided by periodic structural features. This invention uses the degradation risk coefficient as an adaptive weight to ensure lateral and vertical accuracy while forcibly constraining the longitudinal displacement to prevent drift, thereby calculating the global optimal pose, and registering the point cloud to the global coordinate system based on the pose to complete the real-time map construction.

[0063] Specifically, a system is constructed that includes the pose state of the monorail crane. objective function objective function Satisfying the expression:

[0064]

[0065] In the formula, It is the objective function; It is the laser point-to-surface distance residual, obtained by the point cloud registration algorithm; It is the covariance matrix of the laser measurement, determined by the sensor noise model; It is the pre-integrated residual of the inertial measurement unit, obtained by the dynamic integration of the inertial measurement unit; It is the covariance matrix of the inertial measurement unit measurements; It is the degradation risk coefficient; It is the position vector of the monorail crane at the current moment that needs to be solved; It is the predicted position vector of the monorail crane from the previous moment; It is the longitudinal degenerate vector; It is the waveform offset; Represents the dot product operation of vectors; It is the vector magnitude; It is the standard deviation of the longitudinal waveform matching measurement.

[0066] In the formula, By minimizing the spatial distance residual between feature points in the current frame laser point cloud and the corresponding geometric features of the local map, we ensure that the solved pose can keep the observation data and the environment model geometrically consistent. The relative motion increment between adjacent moments is calculated using angular velocity and acceleration data collected by the inertial measurement unit, and the pose change, velocity change and zero bias evolution of the monorail crane must conform to the laws of inertial dynamics. For the longitudinal waveform locking constraint term, where The calculation involves the actual displacement of the monorail crane in the longitudinal degradation vector direction. This calculation mandates that the displacement change of the monorail crane in the longitudinal degradation vector direction must approximate the ideal waveform offset derived from the periodic rhythm of the anchor bolts. Degradation risk coefficient Acting as a dynamic weight adjuster, when environmental features are rich The system is relatively small and mainly relies on conventional matching; however, when roadway degradation occurs... As the size increases, the system forcibly adopts the waveform locking correction result, thereby eliminating longitudinal drift and ensuring that the constructed map is not distorted; introducing... Keep the calculation stable and eliminate dimensions.

[0067] The optimal state vector is obtained by minimizing the objective function, thus obtaining the global optimal pose of the monorail crane. Based on this global optimal pose, the standardized laser ranging data of the current frame is transformed into the global coordinate system and fused into the global point cloud map or probabilistic grid map, thereby realizing real-time map construction of the monorail crane in the tunnel environment.

[0068] For example, Figure 4 This is a comparative schematic diagram of longitudinal positioning drift in a long-distance tunnel environment. The red curve represents the positioning error after using the method of this invention, and the blue curve represents the longitudinal drift caused by geometric degradation due to the accumulation of position along the tunnel direction in the traditional method.

[0069] This invention also discloses a real-time map construction system for monorail cranes in underground tunnel environments, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a real-time map construction method for monorail cranes in underground tunnel environments according to this invention.

[0070] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A real-time map construction method for a monorail hoist in a downhole gallery environment, characterized by, The method comprises the following steps: generating standardized laser ranging data according to vehicle-mounted sensor data; generating a set of discrete direction vectors on a unit sphere, obtaining a geometric sensitivity for each discrete direction vector in the set, obtaining a longitudinal degradation vector and a degradation risk coefficient based on the geometric sensitivity of all discrete direction vectors, and the geometric sensitivity satisfying: ； denotes the geometric sensitivity of the discrete direction vector; is the total number of laser points in the current frame normalized laser ranging data; denotes the geometric normal vector of the laser point; denotes the discrete direction vector of the laser point; denotes the dot product operation of vectors; is a nonlinear mapping kernel function; denotes the absolute value symbol; the degradation risk coefficient satisfying: ； is a degradation risk coefficient; is the minimum geometric sensitivity corresponding to the longitudinal degradation vector direction; is the maximum geometric sensitivity among all discrete direction vectors; is a minimum constant to prevent the denominator from being zero; mapping the standardized laser ranging data onto a straight line determined by the longitudinal degradation vector to generate a real-time density waveform, extracting a reference density waveform from a local map, and obtaining a signal correlation based on the real-time density waveform and the reference density waveform, and obtaining an optimal waveform offset by maximizing the signal correlation; obtaining a target function containing the degradation risk coefficient, the longitudinal degradation vector and the optimal waveform offset, solving a global optimal pose of the monorail crane by minimizing the target function, and completing map construction based on the global optimal pose.

2. The real-time map construction method for a monorail crane in a downhole gallery environment according to claim 1, wherein, The method comprises the following steps: aligning the collection time of each frame of data of the three-dimensional laser radar of the vehicle-mounted sensor to the time axis of the inertial measurement unit in a hardware-triggered manner, calculating the pose change amount of the three-dimensional laser radar at each sampling time relative to the starting time of a frame through integration based on the angular velocity and acceleration measured by the inertial measurement unit during scanning of the frame, and inversely mapping all laser points collected at different times within the frame into the coordinate system at the starting time of the frame through coordinate transformation based on the pose change amount to generate standardized laser ranging data.

3. The real-time map construction method for a monorail hoist facing a roadway environment downhole according to claim 1, characterized in that, The method comprises the following steps: mapping the standardized laser ranging data onto a straight line determined by the longitudinal degradation vector, and discretizing the straight line into a series of small spatial intervals to generate a real-time density waveform distributed along the strike of the roadway by statistically summing the reflection intensity weighted sum of laser points falling into each interval.

4. The real-time map construction method for a monorail hoist facing a roadway environment downhole according to claim 1, characterized in that, The signal correlation satisfies the expression: ； wherein, is a function value of signal correlation when a candidate waveform shift amount is ; is a spatial position variable along a tunnel strike; is a lower limit position of a waveform analysis window; is an upper limit position of a waveform analysis window; is a real-time density waveform which is shifted by a candidate waveform shift amount ; is a reference density waveform.

5. The real-time map construction method for a monorail hoist facing a roadway environment downhole according to claim 1, wherein, The target function satisfies the expression: ； wherein, is the target function; is the monorail crane pose state; is the laser point-plane distance residual; is the laser measurement covariance matrix; is the inertial measurement unit pre-integration residual; is the inertial measurement unit measurement covariance matrix; is the degeneracy risk coefficient; is the current time monorail crane position vector to be solved; is the last time predicted monorail crane position vector; is the longitudinal degeneracy vector; is the waveform offset; denotes the dot product operation of vectors; is the matrix transpose; is the vector norm; is the standard deviation of the longitudinal waveform matching measurement.

6. The real-time map construction method for a monorail hoist facing a roadway environment downhole according to claim 1, wherein, The method comprises the following steps: based on the global optimal pose, converting the standardized laser ranging data of the current frame to the global coordinate system, and fusing the converted data into the global point cloud map or the probability grid map to complete map construction.

7. The real-time map construction method for a monorail hoist facing a roadway environment downhole according to claim 1, wherein, The method comprises the following steps: a plurality of discrete direction vectors are generated by using the Fibonacci grid point algorithm, so that the discrete direction vectors are isotropic and uniformly distributed in space to form a set of discrete direction vectors.

8. A real-time mapping system for a monorail car in a downhole gallery environment, the system comprising: The method comprises the following steps: a processor and a memory, the memory storing computer program instructions which, when executed by the processor, implement the real-time map construction method for a monorail crane in a downhole roadway environment according to any one of claims 1-7.

Images