Intelligent control method and control system for tunneling path of tunneling machine

Abstract

The invention discloses an intelligent control method and system for a tunneling path of a tunneling machine, and relates to the technical field of tunneling machine control, and the method comprises the steps: building a tunneling space risk distribution model; performing hierarchical surrounding collision detection on the tunneling space risk distribution model, and constructing a safe tunneling motion space model; performing tunneling path planning on the safe tunneling motion space model, and building a first tunneling path planning domain; performing overrun feedback optimization on the first tunneling path planning domain, and establishing a second tunneling path planning domain; joint optimization is carried out, and a tunneling path optimization result is determined; and performing tunneling path tracking control on the tunneling machine. The technical problems of poor safety and low efficiency of tunneling operation caused by insufficient risk prejudgment of the tunneling area and insufficient consideration of technical parameters and tunneling targets of the tunneling machine in path planning in the prior art are solved, accurate path tracking control of the tunneling machine is realized, and the tunneling efficiency is improved. And the technical effects of safety and high efficiency of tunneling operation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of roadheader control, and particularly to an intelligent control method and control system for the driving path of a roadheader. Background Art

[0002] In mining operations, as a key piece of equipment, the operation efficiency and safety of a roadheader directly affect the mining progress and benefits. The geological conditions in mining areas are complex and changeable. The risk prediction of the area to be driven by the roadheader is often inaccurate and incomplete, making it difficult to detect potential dangers in advance, resulting in frequent safety hazards such as rock collapses and gas leaks during the driving process. At the same time, in terms of path planning, traditional methods mostly rely on experience or simple algorithms, neither fully considering the technical parameter limitations of the roadheader itself nor closely fitting the complex mining objectives, resulting in problems such as low driving efficiency and resource waste.

[0003] The existing technology has the technical problems of insufficient risk prediction for the driving area and the path planning not fully considering the technical parameters of the roadheader and the driving objectives, resulting in poor safety and low efficiency of the driving operation. Summary of the Invention

[0004] This application provides an intelligent control method and control system for the driving path of a roadheader, aiming to solve the technical problems in the existing technology that there is insufficient risk prediction for the driving area and the path planning does not fully consider the technical parameters of the roadheader and the driving objectives, resulting in poor safety and low efficiency of the driving operation.

[0005] In view of the above problems, this application provides an intelligent control method and control system for the driving path of a roadheader.

[0006] In the first aspect of this application, an intelligent control method for the driving path of a roadheader is provided. The method includes: Performing risk prediction on the area to be driven by the roadheader to establish a risk distribution model for the driving space; performing hierarchical bounding volume collision detection on the risk distribution model of the driving space according to the roadheader to construct a safe driving motion space model; performing driving path planning on the safe driving motion space model according to the driving objectives of the area to be driven to build a first driving path planning domain; performing overrun feedback optimization on the first driving path planning domain according to the technical parameter constraints of the roadheader to establish a second driving path planning domain; performing joint optimization on the second driving path planning domain according to a predetermined driving evaluation factor to determine the driving path optimization result; and performing driving path tracking control on the roadheader according to the driving path optimization result.

[0007] In a possible implementation manner, a roadheader model is built according to the roadheader; based on the roadheader model, simulation collision detection is performed on the risk distribution model of the tunneling space according to multi-level bounding boxes, and a first safe tunneling space model is built, where the multi-level bounding boxes include a plurality of box-shaped colliders and a plurality of mesh colliders; point cleaning is performed on the risk distribution model of the tunneling space according to a point risk threshold, and a second safe tunneling space model is built; the intersection of the first safe tunneling space model and the second safe tunneling space model is obtained to generate the safe tunneling motion space model.

[0008] In a possible implementation manner, based on the roadheader model, simulation collision detection is respectively performed on the risk distribution model of the tunneling space according to the multi-level bounding boxes, and a first simulation collision space model is built; random combinations are made according to the multi-level bounding boxes to obtain a multi-level surrounding architecture; based on the roadheader model, simulation collision detection is performed on the risk distribution model of the tunneling space according to the multi-level surrounding architecture, and a second simulation collision space model is built; union recognition is performed on the first simulation collision space model and the second simulation collision space model to generate a simulation collision integration space model; the difference set between the simulation collision integration space model and the risk distribution model of the tunneling space is used as the first safe tunneling space model.

[0009] In a possible implementation manner, a first bounding box is extracted according to the multi-level bounding boxes; the roadheader model is abstractly surrounded according to the first bounding box to obtain a first surrounded roadheader model; multiple simulation collisions are performed on the risk distribution model of the tunneling space according to the first surrounded roadheader model to obtain multiple simulation collision spaces; the multiple simulation collision spaces are fused to obtain a first surrounded simulation collision space, and the first surrounded simulation collision space is added to the first simulation collision space model.

[0010] In a possible implementation manner, overrun evaluation is performed on each tunneling path in the first tunneling path planning domain according to the technical parameter constraints to obtain a plurality of path overrun coefficients; the first tunneling path planning domain is classified according to the plurality of path overrun coefficients to obtain a first cluster of tunneling paths with a path overrun coefficient less than a predetermined path overrun coefficient and a second cluster of tunneling paths with a path overrun coefficient greater than or equal to the predetermined path overrun coefficient; adaptive adjustment is performed on the second cluster of tunneling paths according to the technical parameter constraints to obtain a third cluster of tunneling paths; the first cluster of tunneling paths and the third cluster of tunneling paths are merged to generate the second tunneling path planning domain.

[0011] In a possible implementation manner, according to the predetermined tunneling evaluation factor, a tunneling evaluation expectation is set, and the predetermined tunneling evaluation factor includes tunneling efficiency and tunneling safety; according to the tunneling evaluation expectation, an optimization analysis is performed on the second tunneling path planning domain to establish a third tunneling path planning domain; weight distribution is performed according to the predetermined tunneling evaluation factor to establish a tunneling goodness evaluation function; according to the tunneling goodness evaluation function, a maximization optimization of the tunneling goodness of the third tunneling path planning domain is performed to generate the tunneling path optimization result.

[0012] In a possible implementation manner, according to the second tunneling path planning domain, the nth tunneling path is extracted, where n is a positive integer; according to the nth tunneling path, feature prediction is performed to determine the nth tunneling prediction result, and the nth tunneling prediction result includes the nth predicted tunneling efficiency coefficient and the nth predicted tunneling safety coefficient; it is determined whether the nth tunneling prediction result meets the tunneling evaluation expectation; if the nth tunneling prediction result meets the tunneling evaluation expectation, the nth tunneling path is added to the third tunneling path planning domain; if the nth tunneling prediction result does not meet the tunneling evaluation expectation, the nth tunneling path is eliminated.

[0013] In a possible implementation manner, the above-ground and underground parameters of the area to be tunneled are collected to obtain a tunneling space data set; three-dimensional reconstruction is performed according to the tunneling space data set to obtain a three-dimensional model of the tunneling space; risk prediction is performed on each point in the area to be tunneled according to the tunneling space data set to obtain the risk prediction coefficient of each point; the risk prediction coefficient of each point is rendered to the three-dimensional model of the tunneling space to generate the risk distribution model of the tunneling space.

[0014] In a possible implementation manner, feature recognition is performed on each point according to the tunneling space data set to obtain the feature data of each point; a point risk prediction network is trained based on a recurrent neural network; the feature data of each point is input into the point risk prediction network to generate the risk prediction coefficient of each point.

[0015] In the second aspect of the present application, an intelligent control system for the tunneling path of a tunneling machine is provided, and the system includes: A risk prediction module for predicting risks in the area to be tunnelled by a tunnelling machine and establishing a risk distribution model for the tunnelling space; a safe tunnelling motion space model construction module for performing hierarchical bounding box collision detection on the tunnelling space risk distribution model according to the tunnelling machine to construct a safe tunnelling motion space model; a tunnelling path planning module for planning a tunnelling path for the safe tunnelling motion space model according to the tunnelling target of the area to be tunnelled to build a first tunnelling path planning domain; an overrun feedback optimization module for performing overrun feedback optimization on the first tunnelling path planning domain according to the technical parameter constraints of the tunnelling machine to establish a second tunnelling path planning domain; a tunnelling path optimization result determination module for jointly optimizing the second tunnelling path planning domain according to a predetermined tunnelling evaluation factor to determine a tunnelling path optimization result; a tunnelling path tracking control module for performing tunnelling path tracking control on the tunnelling machine according to the tunnelling path optimization result.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: Predict risks in the area to be tunnelled by a tunnelling machine and establish a risk distribution model for the tunnelling space; perform hierarchical bounding box collision detection on the tunnelling space risk distribution model to construct a safe tunnelling motion space model; plan a tunnelling path for the safe tunnelling motion space model to build a first tunnelling path planning domain; perform overrun feedback optimization on the first tunnelling path planning domain to establish a second tunnelling path planning domain; jointly optimize the second tunnelling path planning domain to determine a tunnelling path optimization result; perform tunnelling path tracking control on the tunnelling machine. The technical effect of achieving precise path tracking control of the tunnelling machine and improving the safety and efficiency of the tunnelling operation is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0018] Figure 1 It is a schematic flow chart of an intelligent control method for the tunnelling path of a tunnelling machine provided by an embodiment of this application; Figure 2 It is a schematic structural diagram of an intelligent control system for the tunnelling path of a tunnelling machine provided by an embodiment of this application.

[0019] Description of reference numerals: risk prediction module 10, safety tunneling motion space model construction module 20, tunneling path planning module 30, over-limit feedback optimization module 40, tunneling path optimization result determination module 50, tunneling path tracking control module 60. Specific embodiments

[0020] This application provides an intelligent control method and control system for the tunneling path of a roadheader, aiming to solve the technical problems in the prior art that the risk prediction of the tunneling area is insufficient and the path planning does not fully consider the technical parameters of the roadheader and the tunneling target, resulting in poor safety and low efficiency of the tunneling operation.

[0021] The following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the protection scope of this application.

[0022] Embodiment 1, as Figure 1 shown, this application provides an intelligent control method for the tunneling path of a roadheader, and the method includes: Step S100: Predict the risk of the area to be tunnelled by the roadheader and establish a risk distribution model of the tunnelling space.

[0023] Specifically, first collect the above-ground and underground parameters of the area to be tunnelled, such as geological conditions, rock properties, surrounding environment and other information, to form a tunnelling space data set. Then, use these data for 3D reconstruction to generate an intuitive 3D model of the tunnelling space. Then, according to the tunnelling space data set, obtain the feature data of each point by feature recognition of each point, and train a point risk prediction network based on a recurrent neural network. Input the feature data of each point into the network to generate the risk prediction coefficient of each point. Finally, render the risk prediction coefficient of each point onto the 3D model of the tunnelling space, thereby establishing a risk distribution model of the tunnelling space that can clearly show the risk degree of each position in the area to be tunnelled.

[0024] Step S200: Perform hierarchical bounding volume collision detection on the risk distribution model of the tunnelling space according to the roadheader to construct a safety tunnelling motion space model.

[0025] Specifically, first, an excavation machine model needs to be constructed according to the actual shape, size, and operating characteristics of the excavation machine. Based on this model, a multi-level bounding box, namely multiple box colliders and multiple mesh colliders, is used to perform simulated collision detection on the risk distribution model of the excavation space. First, each bounding box is used to abstractly enclose the excavation machine model to obtain the corresponding enclosed excavation machine model, and then it is used to simulate collisions with the risk distribution model multiple times. The multiple simulated collision spaces are fused to construct the first simulated collision space model. After that, the multi-level bounding box is randomly combined to form a multi-level enclosure structure, and the risk distribution model is simulated for collision again to construct the second simulated collision space model. The union of these two simulated collision space models is identified to generate a simulated collision integrated space model, and the difference set between this model and the risk distribution model of the excavation space is the first safe excavation space model. At the same time, based on the point risk threshold, the risk distribution model of the excavation space is subjected to point cleaning to remove points with excessively high risks, and the second safe excavation space model is constructed. Finally, the intersection of the first safe excavation space model and the second safe excavation space model is obtained to construct a safe excavation motion space model that can ensure the safe operation of the excavation machine, providing a safe space range basis for subsequent excavation path planning.

[0026] Step S300: Perform excavation path planning on the safe excavation motion space model according to the excavation target of the area to be excavated, and build the first excavation path planning domain.

[0027] Specifically, the first excavation path planning domain is built by combining the excavation target of the area to be excavated with the safe excavation motion space model, using an improved A* algorithm. First, the excavation target is clarified, such as key information like the specified excavation end coordinates and the expected roadway shape. The safe excavation motion space model is gridded, dividing the space into small grid cells, and each grid contains attribute information such as the risk level and whether it is passable. Starting from the starting position of the excavation machine, the A* algorithm is used to search in the grid. The A* algorithm calculates the actual cost g(n) from each grid node to the starting point and the estimated cost h(n) to the target point, that is, f(n)=g(n)+h(n), and sorts and searches the nodes in ascending order of the f(n) value. During the search process, considering the risk factors in the safe excavation motion space model, for grids with a higher risk level, the movement cost is appropriately increased to make the algorithm avoid high-risk areas as much as possible. At the same time, the path is constrained by technical parameters such as the turning radius and the minimum straight excavation distance of the excavation machine. If the excavation machine's turn at a certain node does not meet the turning radius requirement, or the continuous straight excavation distance is too short, then this path branch is discarded. As the search progresses, the algorithm will generate a series of paths from the starting point to the target point that meet safety and technical requirements, and these paths together form the first excavation path planning domain, providing a basis for subsequent optimization.

[0028] Step S400: Optimize the first tunneling path planning domain through over-limit feedback according to the technical parameter constraints of the tunneling machine, and establish a second tunneling path planning domain.

[0029] Specifically, according to the technical parameters of the tunneling machine, such as the maximum excavation speed, cutting power, equipment size limit, maximum climbing angle, etc., a detailed over-limit evaluation is carried out for each tunneling path within the first tunneling path planning domain. Calculate the degree to which each parameter of each path deviates from the normal range during the execution process to obtain multiple path over-limit coefficients. These coefficients intuitively reflect the technical risks and equipment load conditions that each path may face. According to the calculated path over-limit coefficients, the paths within the first tunneling path planning domain are classified. The paths with path over-limit coefficients less than the predetermined path over-limit coefficient are classified into the first cluster of tunneling paths. Such paths can basically meet the technical parameter requirements of the tunneling machine and have relatively low risks; while the paths with path over-limit coefficients greater than or equal to the predetermined path over-limit coefficient are classified into the second cluster of tunneling paths. These paths have some parameter over-limit situations and need to be adjusted. For the second cluster of tunneling paths, adaptive adjustment is carried out according to the technical parameter constraints of the tunneling machine. For example, if the excavation speed of a certain path is too fast resulting in over-limit cutting power, then the excavation speed is appropriately reduced; if the turning angle of the path is too large and exceeds the equipment allowable range, then the path of the turning part is re-planned. Through adjustment and optimization measures, a third cluster of tunneling paths that meet the technical parameter requirements is obtained. Finally, the first cluster of tunneling paths and the optimized third cluster of tunneling paths are merged to generate a second tunneling path planning domain. The paths in the second tunneling path planning domain not only consider the requirements of the tunneling target and safety space but also meet the technical parameter constraints of the tunneling machine, further improving the accuracy and feasibility of the tunneling path planning.

[0030] Step S500: Jointly optimize the second tunneling path planning domain according to the predetermined tunneling evaluation factors to determine the tunneling path optimization result.

[0031] Specifically, for the second excavation path planning domain, joint optimization work is carried out according to the predetermined excavation evaluation factors to determine the final excavation path optimization results. The predetermined excavation evaluation factors cover two aspects: excavation efficiency and excavation safety. First, according to the actual engineering needs, the excavation evaluation expectations corresponding to these factors are set, such as clarifying the minimum excavation efficiency standard expected to be achieved and the highest acceptable risk level. The nth excavation path (n is a positive integer) is extracted from the second excavation path planning domain in turn, and a comprehensive feature prediction is performed on it. By analyzing factors such as the length of the path, the excavation speed, and the risk level of the crossing area, the nth predicted excavation efficiency coefficient and the nth predicted excavation safety coefficient are determined to quantify the performance of the path in excavation efficiency and safety. Subsequently, these prediction results are compared and judged with the excavation evaluation expectations. If the nth excavation prediction result meets the set excavation evaluation expectations, it means that the path has met the basic requirements in terms of efficiency and safety, and it is added to the third excavation path planning domain; if not, the path is directly eliminated. After the screening is completed, according to the importance of excavation efficiency and excavation safety in the entire excavation operation, weights are reasonably assigned to them, and then the excavation quality evaluation function is constructed. This function can comprehensively consider these two key factors and quantitatively evaluate the overall quality of the path. Finally, the optimization algorithm is used to maximize the excavation quality of the path in the third excavation path planning domain. In the optimization process, the parameters of the path, such as excavation speed, turning method, etc., are continuously adjusted to maximize the excavation quality of the path. The excavation path finally generated is the excavation path optimization result. This result not only meets the engineering excavation goals, but also takes into account the balance between excavation efficiency and safety.

[0032] Step S600: performing tunneling path tracking control on the tunnel boring machine according to the tunneling path optimization result.

[0033] Specifically, the optimized tunneling path is converted into control instructions recognizable by the tunneling machine, including parameters such as tunneling direction, speed, cutting force, etc. During the operation of the tunneling machine, various sensors installed on the machine body, such as laser sensors, inertial navigation sensors, displacement sensors, etc., are used to obtain real-time operation status information of the tunneling machine, such as its position, attitude, speed, etc. These actual operation statuses are compared and analyzed with the theoretical parameters of the optimized path. Once a deviation is found, the magnitude and direction of the deviation are calculated. For different types of deviations, corresponding adjustment strategies are adopted. If there is a deviation in the tunneling direction, the steering mechanism of the tunneling machine is adjusted to gradually return it to the optimized path; if the tunneling speed does not match the plan, the power of the drive system is adjusted to make the speed reach an appropriate value. At the same time, considering that geological conditions and environmental factors may change during the tunneling process, the control system continuously monitors the dynamic changes of these factors and optimizes the control instructions in real time according to the changes. In addition, to ensure the stability and reliability of the tunneling path tracking control, a feedback control mechanism is introduced to monitor and evaluate the adjusted operation status again, forming a closed-loop control process. If the deviation cannot be effectively eliminated after multiple adjustments, an alarm is automatically issued to prompt the operator for manual intervention. Through such tunneling path tracking control, it is possible to ensure that the tunneling machine efficiently and stably completes the tunneling operation along the optimal path, reduce construction risks, and improve project quality.

[0034] In a possible implementation manner, step S100 further includes: Step S110: Collect the aboveground and underground parameters of the area to be tunneled to obtain a tunneling space data set.

[0035] Step S120: Perform three-dimensional reconstruction based on the tunneling space data set to obtain a three-dimensional model of the tunneling space.

[0036] Step S130: Perform risk prediction on each point in the area to be tunneled according to the tunneling space data set to obtain the risk prediction coefficients of each point.

[0037] Step S140: Render the risk prediction coefficients of each point to the three-dimensional model of the tunneling space to generate the risk distribution model of the tunneling space.

[0038] Specifically, in the early stage of tunneling work, it is necessary to comprehensively collect the aboveground and underground parameters of the area to be tunneled. For aboveground parameters, technologies such as satellite remote sensing and unmanned aerial vehicle mapping can be used to obtain information such as topography, surface building distribution, and vegetation cover; for underground parameters, data such as geological structure, rock characteristics, groundwater level, and soil mechanical properties are collected through geological drilling and geophysical exploration. Integrating these rich data, a tunneling space data set is obtained.

[0039] Carry out 3D reconstruction based on the obtained tunneling space dataset. Use professional 3D modeling software, such as 3ds Max, Maya, etc., to construct a 3D solid model of the area to be tunneled according to the terrain, landform and underground structure information in the dataset. In this process, the software will convert the data into intuitive 3D graphics, accurately presenting the actual spatial form of the area to be tunneled, and finally obtaining a 3D model of the tunneling space.

[0040] Use a recurrent neural network to carry out risk prediction for each point in the area to be tunneled based on the tunneling space dataset, and then obtain the risk prediction coefficients for each point. First, preprocess the tunneling space dataset, which contains various parameters above and below the ground, such as geological structure, rock properties, groundwater level, surrounding environment, etc. Clean the data to remove noise and missing values, and then perform normalization processing to make the data within a suitable numerical range. Then, organize the processed data according to the spatial order and time series (if there is time-related data) of the points to construct a format suitable for input to the recurrent neural network. Construct a recurrent neural network model, select a classic recurrent neural network structure, determine hyperparameters such as the number of layers and neurons of the model, and initialize it. Use the preprocessed dataset to train the recurrent neural network model. During the training process, the model will learn the features and patterns in the data, as well as the potential relationship between each parameter and the risk. Continuously adjust the weights and biases of the model through the backpropagation algorithm to minimize the error between the predicted value and the actual risk value. After training, input the data of each point in the area to be tunneled into the trained recurrent neural network model for prediction. The model will output the risk prediction results for each point according to the learned patterns. Post-process these results, and through normalization, convert them into risk prediction coefficients between 0 and 1. The larger the coefficient, the higher the risk of the point. Finally, obtain the risk prediction coefficients for each point in the area to be tunneled.

[0041] Render the risk prediction coefficients of each point onto the 3D model of the tunneling space. Use visualization technology to assign different colors or textures to each point in the model according to the magnitude of the risk prediction coefficient. For example, high-risk points are represented by red, and low-risk points are represented by green. In this way, the abstract risk data is intuitively displayed in the 3D model, thereby generating a risk distribution model of the tunneling space.

[0042] In a possible implementation manner, step S130 further includes: Step S131: Perform feature recognition on each point according to the tunneling space dataset to obtain feature data for each point.

[0043] Step S132: Train a point risk prediction network based on a recurrent neural network.

[0044] Step S133: Input the feature data of each point into the point risk prediction network to generate the risk prediction coefficients of each point.

[0045] Specifically, the tunneling space dataset contains rich and diverse information, including geological, topographical, groundwater level and other aspects of data. To obtain the feature data of each point, first for the geological data, geostatistical methods are used to analyze the changes in formation lithology at each point, and identify the specific manifestations of different rock types, fault trends and fold characteristics at the points. For example, through the mineral composition analysis of rock samples, the characteristic data such as rock hardness and compressive strength at the points are determined. For the topographical data, using the digital elevation model (DEM) combined with geographic information system (GIS) technology, features such as slope, aspect and terrain undulation are extracted at each point to reflect the complexity of the terrain where the point is located. In terms of groundwater level data processing, with the help of hydrogeological simulation software, the dynamic change laws of groundwater levels at different points are analyzed to obtain characteristic data such as water level depth and water level fluctuation range. Combining the analysis results of these different types of data, comprehensive and targeted feature data for each point are sorted out.

[0046] Based on the recurrent neural network, the point risk prediction network is trained. First, the previously obtained tunneling space dataset is preprocessed, including data cleaning to remove outliers and noise data, and then normalization processing is carried out to make the data in different dimensions within a unified numerical range, which is beneficial to model training. Subsequently, a recurrent neural network model is constructed. After determining the key hyperparameters such as the number of network layers and the number of neurons, the preprocessed dataset is used to train the model. During the training process, by continuously adjusting the weights and biases of the network, the error between the predicted risk value and the actual risk value is minimized using the backpropagation algorithm, and the prediction accuracy of the model is gradually improved until the training converges, completing the training of the point risk prediction network.

[0047] Input the obtained feature data of each point into the trained point risk prediction network. The network analyzes and calculates the input point feature data based on the mapping relationship between the features and risks learned during the training process, and finally outputs the risk prediction results corresponding to each point. These results are appropriately post-processed and normalized to generate the risk prediction coefficients of each point, providing an accurate quantitative basis for subsequent risk assessment and tunneling path planning.

[0048] In a possible implementation manner, step S200 further includes: Step S210: Model the tunneling machine to construct a tunneling machine model.

[0049] Step S220: Based on the roadheader model, perform simulated collision detection on the tunneling space risk distribution model according to the multi-level bounding box, and construct the first safe tunneling space model. The multi-level bounding box includes a plurality of box-shaped colliders and a plurality of grid colliders.

[0050] Step S230: Perform point cleaning on the tunneling space risk distribution model according to the point risk threshold, and construct the second safe tunneling space model.

[0051] Step S240: Obtain the intersection of the first safe tunneling space model and the second safe tunneling space model to generate the safe tunneling motion space model.

[0052] Specifically, based on parameters such as the actual size, shape, structure of the roadheader and the movement range of each component, use modeling software to perform precise modeling to construct a roadheader model that can truly reflect the characteristics of the roadheader. This model is the basis for subsequent simulated collision detection, and its accuracy directly affects the construction effect of the safe space.

[0053] According to the outer contour, structural characteristics of the roadheader model and the positions of key components, arrange a plurality of box-shaped colliders and grid colliders to form a multi-level bounding box. For the relatively regular main part of the roadheader shape, use box-shaped colliders for preliminary and large-scale enclosure to quickly define the general area where collisions may occur; for parts with complex shapes and flexible movements such as the cutting head and boom of the roadheader, use grid colliders for precise fitting to capture more subtle collision situations. Subsequently, place the roadheader model in the three-dimensional space represented by the tunneling space risk distribution model, and simulate various possible motion postures and paths of the roadheader. Use collision detection algorithms such as the Separating Axis Theorem (SAT) to perform real-time collision detection on the high-risk areas in the multi-level bounding box and the tunneling space risk distribution model. In each simulated motion step, check whether the box-shaped colliders and grid colliders intersect with the risk area. If an intersection is detected, mark that position as a potential dangerous area. After completing the collision detection of all simulated motions, traverse the entire three-dimensional space, and screen out the parts of the space that are not marked as dangerous areas. Then, use spatial topology analysis methods to remove those isolated, narrow and meaningless space fragments for actual passage, perform connectivity analysis and optimization on the screened safe space. Finally, integrate the optimized safe space to construct the first safe tunneling space model, which clearly defines the space range where the roadheader can operate safely on the premise of avoiding collisions with high-risk areas.

[0054] Determine the risk threshold for points in advance according to multiple factors such as geological conditions, roadheader performance, and construction safety standards. Judge each point in the constructed risk distribution model of the excavation space to obtain its corresponding risk coefficient. Compare these risk coefficients with the risk threshold for points one by one. When the risk coefficient of a certain point is greater than or equal to the risk threshold for points, determine that point as a high-risk point. Subsequently, accurately remove all the determined high-risk points from the risk distribution model of the excavation space. After removing the high-risk points, re-integrate and construct according to the spatial range composed of the remaining low-risk points, and finally form the second safe excavation space model.

[0055] Convert the data of the two models into a unified spatial coordinate system to ensure the consistency of position information. Select a suitable spatial analysis algorithm, such as the Boolean intersection algorithm, traverse the grid cells of the two models, and find the grid cells that exist in both models, and mark these cells as the intersection part. During the calculation process, precise processing should be carried out on the boundary part to avoid gaps or overlapping errors. For the obtained intersection data, perform topological inspection and optimization, remove isolated small areas and unreasonable tiny spaces, and ensure the connectivity and rationality of the intersection space. Finally, based on the optimized intersection data, use 3D modeling software to reconstruct the safe excavation motion space model, providing a reliable spatial basis for subsequent excavation path planning and control.

[0056] In a possible implementation manner, step S220 further includes: Step S221: Based on the roadheader model, respectively perform simulated collision detection on the risk distribution model of the excavation space according to the multi-level bounding box, and construct the first simulated collision space model.

[0057] Step S222: Randomly combine according to the multi-level bounding box to obtain a multi-level bounding structure.

[0058] Step S223: Based on the roadheader model, perform simulated collision detection on the risk distribution model of the excavation space according to the multi-level bounding structure, and construct the second simulated collision space model.

[0059] Step S224: Perform union recognition on the first simulated collision space model and the second simulated collision space model to generate a simulated collision integrated space model.

[0060] Step S225: Use the difference set between the simulated collision integrated space model and the risk distribution model of the excavation space as the first safe excavation space model.

[0061] Specifically, relying on the constructed roadheader model, a multi-level bounding box composed of multiple box colliders and multiple mesh colliders is used and reasonably arranged around the roadheader model, so that the multi-level bounding box can comprehensively and accurately represent the range that the roadheader may occupy in space. Subsequently, the simulated collision detection work is carried out, simulating various possible postures and positions of the roadheader in the tunneling space. In each simulated state, it is checked whether there is an overlap between the multi-level bounding box and the risk area in the risk distribution model of the tunneling space. Once an overlap is found, it indicates that there is a possibility of collision between the roadheader and the risk area in this posture and position. After completing the detection of all simulated states, the spatial parts where the multi-level bounding box and the risk area do not overlap are marked. Finally, these marked spatial parts are integrated and processed, removing the possible isolated small pieces and unreasonable gaps, so as to construct the first simulated collision space model.

[0062] Facing the multi-level bounding box composed of multiple box colliders and mesh colliders, randomly select some of the colliders for combination operations. Since each box collider and mesh collider has a unique size, shape and preset position in space, the combination methods are extremely diverse. For example, select several of different sizes from multiple box colliders and match them with specific mesh colliders. During the selection process, a random algorithm is used to determine the types and quantities of colliders participating in the combination each time. After multiple random selections and combinations, a series of different multi-level surrounding architectures are obtained. Due to the random combination of colliders, these architectures present different surrounding forms and coverage ranges, and can simulate a variety of different roadheader collision detection scenarios.

[0063] Relying on the constructed roadheader model, the multi-level surrounding architecture obtained by randomly combining the multi-level bounding box is applied to the simulated collision detection work of the risk distribution model of the tunneling space, and then the second simulated collision space model is constructed. During the simulation, the roadheader is simulated to move in the tunneling space with various possible postures and positions. In each simulated state, it is checked whether there is an overlap between the multi-level surrounding architecture and the risk area in the risk distribution model of the tunneling space. If an overlap is detected, it indicates that there is a risk of collision between the roadheader and the risk area in this specific state. When the detection of all simulated states is completed, the spatial parts where the multi-level surrounding architecture and the risk area do not overlap are marked. Subsequently, these marked spatial areas are integrated and optimized, removing the isolated, tiny and meaningless spatial fragments, and finally the second simulated collision space model is constructed.

[0064] Convert the data of the two models into unified three-dimensional grid data. For each grid cell, assign it a unique identifier and spatial coordinate information. Then, create an empty set to store the union result. For each grid cell in the first simulated collision space model, check if it already exists in the result set. If not, add it. After that, perform the same operation on the second simulated collision space model. Traverse each grid cell in it. If the cell is not in the result set, add it. During the addition process, pay special attention to handling boundary cases, such as the overlapping parts between grid cells. For overlapping grid cells, only keep one copy to avoid duplicate calculations. Finally, based on the set storing the union result, reconstruct the space model to form the simulated collision integration space model.

[0065] Represent the models in the form of three-dimensional grids. First, perform unified index marking on the grids of the two models. For each grid cell in the simulated collision integration space model, obtain its spatial coordinates. Using a spatial index structure, such as an octree, quickly search in the tunneling space risk distribution model to check if there is a risk grid cell with the same spatial coordinates. If no coincident risk grid cell is found, mark the grid cell of the simulated collision integration space model as belonging to the difference set part and record its relevant attribute information. If a coincident risk grid cell is found, discard the cell. After traversing all the grid cells of the simulated collision integration space model, recombine all the grid cells marked as belonging to the difference set part to construct the first safe tunneling space model.

[0066] In a possible implementation manner, step S221 further includes: Step S2211: Extract the first bounding box according to the multi-level bounding box.

[0067] Step S2212: Abstractly enclose the tunneling machine model according to the first bounding box to obtain the first enclosed tunneling machine model.

[0068] Step S2213: Perform multiple simulated collisions on the tunneling space risk distribution model according to the first enclosed tunneling machine model to obtain multiple simulated collision spaces.

[0069] Step S2214: Fuse the multiple simulated collision spaces to obtain the first enclosed simulated collision space, and add the first enclosed simulated collision space to the first simulated collision space model.

[0070] Specifically, faced with a multi-level bounding box composed of multiple box-shaped colliders and mesh colliders, the first bounding box is extracted from it. Work is carried out according to pre-set extraction rules. For example, based on the appearance characteristics of the tunnel boring machine, the focus is on box-shaped colliders that are larger in size, regular in shape, and can cover the main parts of the tunnel boring machine. The degree of fit between different bounding boxes and the key components of the tunnel boring machine is also considered, and those colliders that can fit tightly to the prone-to-collision areas of the tunnel boring machine are given priority. At the same time, combined with the analysis of the risk distribution model of the tunneling space, a bounding box with better coverage of common high-risk areas is selected. By evaluating the multi-level bounding boxes one by one, the one that best meets the requirements is selected from the many colliders and determined as the first bounding box.

[0071] Determine the position and posture of the first bounding box in space so that it can cover the main structure and key parts of the tunnel boring machine model to the greatest extent, such as the fuselage, cutting head and main moving parts of the tunnel boring machine. In the encirclement process, ignore the complex details of the tunnel boring machine model and regard it as a whole contained in the first bounding box. Through the spatial transformation algorithm, unify the coordinate system of the first bounding box with the coordinate system of the tunnel boring machine model to ensure the accurate correspondence between the two in spatial position. Then, combine the contour information of the first bounding box with the tunnel boring machine model to generate the first encircled tunnel boring machine model. Although this model discards some details of the tunnel boring machine, it retains the key space occupancy information required for simulated collision detection. In the subsequent simulated collision link, it can efficiently and accurately represent the tunnel boring machine and the tunneling space risk distribution model to interact, providing a more concise and effective data basis for simulated collision detection.

[0072] Based on the first bounding roadheader model, conduct multiple simulations and collisions at all points of the tunneling space risk distribution model to obtain multiple simulated collision spaces. With the help of simulation software, such as ANSYS, ADAMS, etc., import the finely constructed first bounding roadheader model and the complete tunneling space risk distribution model into the software simulation environment to ensure the unity of the spatial coordinate systems of the two. Through programming or the parameter setting function built into the software, generate multiple sets of different motion parameters, including the starting position, traveling direction, motion speed, rotation angle, and attitude changes of the first bounding roadheader model in the tunneling space. Use a loop instruction to start multiple simulation processes. Each time a simulation is performed, set the initial state of the first bounding roadheader model according to the current parameters, and make it move along the preset trajectory in the virtual tunneling space. During the movement process, use the collision detection method based on spatial grid division to perform real-time collision detection on each point (regardless of the risk level) in the first bounding roadheader model and the tunneling space risk distribution model. Once a collision is detected, immediately record the collision information; if there is no collision, mark the spatial area where the current first bounding roadheader model is located as a safe area. After each simulation is completed, organize and store the data of the safe area obtained from this simulation to form a simulated collision space. Repeat this process. As multiple sets of motion parameters are applied in sequence, complete multiple simulations and accumulate multiple different simulated collision spaces.

[0073] After multiple simulated collision spaces have been obtained, use the union operation in Boolean operations to process the multiple simulated collision spaces. This algorithm will traverse the boundaries and internal regions of each simulated collision space, identify the overlapping and adjacent parts, and merge these parts into a continuous and complete space. During the fusion process, precisely process the boundaries, eliminate the gaps and overlaps caused by simulation errors, and ensure the continuity and accuracy of the space. At the same time, for possible isolated small regions, if their area is too small and has little impact on the overall space, choose to discard them to simplify the space structure. After the fusion operation, the obtained first bounding simulated collision space comprehensively and accurately reflects the safe movement range of the roadheader based on the first bounding box when avoiding collisions with risk areas. Finally, add this first bounding simulated collision space to the first simulated collision space model being constructed.

[0074] In a possible implementation manner, step S400 further includes: Step S410: Conduct an overlimit evaluation on each tunneling path within the first tunneling path planning domain according to the technical parameter constraints to obtain multiple path overlimit coefficients.

[0075] Step S420: Classify the first tunneling path planning domain according to the multiple path overlimit coefficients to obtain a first cluster of tunneling paths with a path overlimit coefficient less than a predetermined value, and a second cluster of tunneling paths with a path overlimit coefficient greater than or equal to the predetermined value.

[0076] Step S430: Perform adaptive adjustment on the second cluster of the tunneling path according to the technical parameter constraints to obtain the third cluster of the tunneling path.

[0077] Step S440: Merge the first cluster of the tunneling path and the third cluster of the tunneling path to generate the second tunneling path planning domain.

[0078] Specifically, for the technical parameter constraints, they are subdivided into different types, such as equipment performance parameters, geological condition parameters, safety standard parameters, etc., and weight values are set for each parameter to reflect its relative importance in the overall evaluation. For each tunneling path, a series of sampling points are selected at equal intervals along the path. At each sampling point, the actual technical parameter values are collected, such as tunneling speed, slope, distance from surrounding obstacles, etc. For each technical parameter, the deviation rate between the actual value and the constraint value is calculated, that is, (actual value - constraint value) / constraint value (when the actual value is greater than the constraint value) or (constraint value - actual value) / constraint value (when the actual value is less than the constraint value). Then, the deviation rate of each parameter is multiplied by its corresponding weight value, and the product results of all parameters are accumulated to obtain the local overrun coefficient of this sampling point. The local overrun coefficients of all sampling points on the path are weighted averaged (the weights can be set according to the sampling point spacing or the importance of the area where they are located), and finally the path overrun coefficient of this tunneling path is obtained. This algorithm is sequentially executed for all tunneling paths within the first tunneling path planning domain, thereby obtaining multiple path overrun coefficients.

[0079] Perform a classification operation on the first tunneling path planning domain according to the obtained multiple path overrun coefficients. Set a predetermined path overrun coefficient as the classification threshold. Classify the tunneling paths with path overrun coefficients less than this threshold into one category to form the first cluster of tunneling paths. This cluster of paths is relatively more in line with the technical parameter constraints and has a lower overrun degree. And classify the tunneling paths with path overrun coefficients greater than or equal to the predetermined path overrun coefficient into another category to form the second cluster of tunneling paths. The overrun situation of this cluster of paths is more obvious.

[0080] Perform adaptive adjustment on the second cluster of the tunneling path. Using an optimization algorithm, combined with the technical parameter constraints and the actual engineering requirements, adjust each path in the second cluster. For example, change parameters such as the path direction, slope, curvature, etc., so that it is as close as possible to the optimal state under the premise of meeting the technical parameter constraints. After the adaptive adjustment, the third cluster of the tunneling path is obtained. After the adjustment of this cluster of paths, its overrun situation has been improved and it is more in line with the technical requirements.

[0081] Merge the first cluster of tunneling paths and the third cluster of tunneling paths, integrate all the tunneling paths included in the two clusters together to form a new path set, and this set is the second tunneling path planning domain. This domain combines the originally compliant paths and the paths that meet the technical parameter constraints after adjustment, providing a more reasonable and optimized selection range for further determining the optimal tunneling path subsequently.

[0082] In a possible implementation manner, step S500 further includes: Step S510: Set a tunneling evaluation expectation according to the predetermined tunneling evaluation factors, where the predetermined tunneling evaluation factors include tunneling efficiency and tunneling safety.

[0083] Step S520: Perform an optimization analysis on the second tunneling path planning domain according to the tunneling evaluation expectation to establish a third tunneling path planning domain.

[0084] Step S530: Perform weight allocation according to the predetermined tunneling evaluation factors to establish a tunneling goodness evaluation function.

[0085] Step S540: Perform an optimization to maximize the tunneling goodness on the third tunneling path planning domain according to the tunneling goodness evaluation function to generate the tunneling path optimization result.

[0086] Specifically, based on the core impacts of the two major predetermined tunneling evaluation factors of tunneling efficiency and tunneling safety, the tunneling evaluation expectation needs to be set accordingly. For tunneling efficiency, combined with the project progress requirements, equipment performance and past construction experience, set an expected tunneling speed or the tunneling volume target per unit time. For tunneling safety, consider multiple aspects such as geological conditions, equipment operation stability, and personnel safety guarantee, and formulate specific safety expectation indicators such as the maximum allowable risk probability and the minimum safety distance.

[0087] Extract the nth tunneling path in sequence by number from the second tunneling path planning domain, where n is a positive integer, so as to ensure that each path is evaluated one by one. Then, for the extracted nth tunneling path, use information from multiple aspects such as historical data, geological conditions, and equipment performance, and apply a prediction model to conduct feature prediction, so as to determine the nth tunneling prediction result. This result includes the nth predicted tunneling efficiency coefficient reflecting the expected tunneling speed and workload, etc., and the nth predicted tunneling safety coefficient considering factors such as the geological stability around the path and the distance from dangerous areas. After that, compare the nth tunneling prediction result with the pre-set tunneling evaluation expectation to judge whether it meets the expected tunneling efficiency and safety standards. If it meets the tunneling evaluation expectation, it means that the path performs well in terms of efficiency and safety, and add it to the third tunneling path planning domain being constructed; if it does not meet, it indicates that there are efficiency or safety hazards in the path, and it will be eliminated. By repeating the above operations for all paths in the second tunneling path planning domain, a third tunneling path planning domain composed of paths that meet the tunneling evaluation expectation is finally formed.

[0088] Perform weight allocation according to the predetermined tunneling evaluation factors, and then establish a tunneling goodness evaluation function. The predetermined tunneling evaluation factors are mainly tunneling efficiency and tunneling safety. Considering the different importance of these two factors in different engineering scenarios, the analytic hierarchy process is used to determine their weights. Assume that through reasonable evaluation, the weight of tunneling efficiency is determined to be W1, and the weight of tunneling safety is determined to be W2, and W1 + W2 = 1 is satisfied. For tunneling efficiency, it is measured by calculating the ratio of the expected completion time t of the path to the expected tunneling time t0, denoted as E = , the larger this value, the higher the tunneling efficiency. For tunneling safety, comprehensively consider factors such as the geological risk around the path and the distance from obstacles, and a safety coefficient S can be obtained by establishing a risk assessment model. Its value range is between 0 and 1, and the larger the value, the higher the safety. Based on the above analysis, establish a tunneling goodness evaluation function. The tunneling goodness evaluation function is: F = W1×E + W2×S, where F represents the tunneling goodness, W1 represents the weight of tunneling efficiency, W2 represents the weight of tunneling safety, W1 and W2 together constitute the importance distribution of the overall evaluation, E represents the tunneling efficiency, and its calculation method is the ratio of the expected completion time t of the path to the expected tunneling time t0, t0 is the expected tunneling time, t is the expected completion time, and S represents the tunneling safety coefficient, with the value range set between 0 and 1. This function comprehensively considers the two key factors of tunneling efficiency and tunneling safety and their corresponding weights, and can quantify the goodness of each path.

[0089] With the established tunneling suitability evaluation function, the tunneling suitability of the third tunneling path planning domain is optimized to maximize. In this process, an optimization algorithm is used to calculate the tunneling suitability value of each path in the third tunneling path planning domain. The algorithm continuously adjusts the relevant parameters of the path (such as path direction, slope, etc.) to find the path that maximizes the value of the tunneling suitability evaluation function. After multiple rounds of iteration and optimization, the tunneling path optimization result is finally generated, which represents the most ideal tunneling path plan considering tunneling efficiency and tunneling safety and maximizing the suitability.

[0090] In a possible implementation manner, step S520 further includes: Step S521: Extract the nth tunneling path according to the second tunneling path planning domain, where n is a positive integer.

[0091] Step S522: Perform feature prediction according to the nth tunneling path to determine the nth tunneling prediction result, where the nth tunneling prediction result includes the nth predicted tunneling efficiency coefficient and the nth predicted tunneling safety coefficient.

[0092] Step S523: Determine whether the nth tunneling prediction result meets the tunneling evaluation expectation.

[0093] Step S524: If the nth tunneling prediction result meets the tunneling evaluation expectation, add the nth tunneling path to the third tunneling path planning domain.

[0094] Step S525: If the nth tunneling prediction result does not meet the tunneling evaluation expectation, eliminate the nth tunneling path.

[0095] Specifically, the nth tunneling path is extracted from the second tunneling path planning domain, where n is a positive integer. In order to evaluate and screen these paths one by one to further determine a better path plan, the paths are processed in the order of the numbers. First, clarify the numbering rule, assign a unique positive integer number to each path in the second tunneling path planning domain, starting from 1 and increasing sequentially. Then, according to the current number n to be processed, the corresponding nth tunneling path is accurately extracted from the path set through the indexing mechanism.

[0096] Use the random forest algorithm to perform feature prediction based on the nth tunneling path, determine the nth tunneling prediction result, collect a large amount of historical tunneling data, which contains the feature information of many paths, such as path length, slope, geological type, surrounding environment conditions, etc., and record the corresponding actual tunneling efficiency and safety indicators at the same time. For tunneling efficiency, it is measured by the tunneling length per unit time; for safety, it can be comprehensively evaluated based on the accident occurrence probability, risk level, etc. Divide the collected data into a training set and a test set, and use the training set data to construct a random forest model. The random forest consists of multiple decision trees. When constructing each decision tree, a part of the samples and features are randomly selected from the training data, which can increase the diversity and generalization ability of the model. For predicting the nth predicted tunneling efficiency coefficient, input the features of the nth tunneling path into the trained random forest model. Each decision tree in the model will give a predicted tunneling efficiency value according to its own rules. Finally, take the average of the prediction results of all decision trees to obtain the predicted tunneling efficiency of the nth tunneling path, and then compare it with the historical optimal tunneling efficiency and perform normalization processing to obtain the nth predicted tunneling efficiency coefficient. Similarly, when predicting the nth predicted tunneling safety coefficient, also input the features of the nth tunneling path into the model. Each decision tree outputs the safety prediction result respectively. After averaging, the predicted safety index is obtained, and then through operations such as normalization, the nth predicted tunneling safety coefficient is obtained. Finally, these two coefficients together constitute the nth tunneling prediction result.

[0097] Compare the nth tunneling prediction result with the preset tunneling evaluation expectation, which covers the expected tunneling efficiency and tunneling safety standards. Judge whether the nth predicted tunneling efficiency coefficient reaches the expected efficiency level, and whether the nth predicted tunneling safety coefficient meets the expected safety requirements.

[0098] If the nth tunneling prediction result meets the tunneling evaluation expectation, add the nth tunneling path to the third tunneling path planning domain being constructed. This means that this path meets the expectations in terms of both tunneling efficiency and safety and is a relatively good choice.

[0099] Conversely, if the nth tunneling prediction result does not meet the tunneling evaluation expectation, eliminate the nth tunneling path because it cannot reach the established standard and may pose potential problems in terms of efficiency or safety for subsequent tunneling work. By repeating the above operations for each path in the second tunneling path planning domain, a third tunneling path planning domain composed of paths that meet the tunneling evaluation expectation is finally formed, providing a more valuable set of candidate paths for subsequent path optimization.

[0100] Embodiment 2, based on the same inventive concept as the intelligent control method for the tunneling path of the tunneling machine in the foregoing embodiment, such as Figure 2As shown, the present application provides an intelligent control system for the tunneling path of a roadheader. The system and method embodiments in the present application are based on the same inventive concept. Among them, the system includes: A risk prediction module 10, configured to perform risk prediction on the area to be tunneled by the roadheader and establish a risk distribution model for the tunneling space.

[0101] A safe tunneling motion space model construction module 20, configured to perform hierarchical bounding volume collision detection on the risk distribution model of the tunneling space according to the roadheader, and construct a safe tunneling motion space model.

[0102] A tunneling path planning module 30, configured to perform tunneling path planning on the safe tunneling motion space model according to the tunneling target of the area to be tunneled, and build a first tunneling path planning domain.

[0103] An overlimit feedback optimization module 40, configured to perform overlimit feedback optimization on the first tunneling path planning domain according to the technical parameter constraints of the roadheader, and establish a second tunneling path planning domain.

[0104] A tunneling path optimization result determination module 50, configured to perform joint optimization on the second tunneling path planning domain according to a predetermined tunneling evaluation factor, and determine the tunneling path optimization result.

[0105] A tunneling path tracking control module 60, configured to perform tunneling path tracking control on the roadheader according to the tunneling path optimization result.

[0106] Furthermore, the system is also used to implement the following functions: Model the roadheader to construct a roadheader model; based on the roadheader model, perform simulated collision detection on the risk distribution model of the tunneling space according to multiple-level bounding boxes, and construct a first safe tunneling space model, where the multiple-level bounding boxes include multiple box-shaped colliders and multiple grid colliders; perform point cleaning on the risk distribution model of the tunneling space according to a point risk threshold, and construct a second safe tunneling space model; find the intersection of the first safe tunneling space model and the second safe tunneling space model to generate the safe tunneling motion space model.

[0107] Furthermore, the system is also used to implement the following functions: Based on the roadheader model, according to the multi-level bounding box, perform simulated collision detection on the risk distribution model of the tunneling space respectively to construct a first simulated collision space model; perform random combination according to the multi-level bounding box to obtain a multi-level bounding structure; based on the roadheader model, perform simulated collision detection on the risk distribution model of the tunneling space according to the multi-level bounding structure to construct a second simulated collision space model; perform union recognition on the first simulated collision space model and the second simulated collision space model to generate a simulated collision integrated space model; use the difference set between the simulated collision integrated space model and the risk distribution model of the tunneling space as the first safe tunneling space model.

[0108] Further, the system is also used to implement the following functions: Extract a first bounding box according to the multi-level bounding box; perform abstract bounding on the roadheader model according to the first bounding box to obtain a first bounding roadheader model; perform multiple simulated collisions on the risk distribution model of the tunneling space according to the first bounding roadheader model to obtain multiple simulated collision spaces; fuse the multiple simulated collision spaces to obtain a first bounding simulated collision space, and add the first bounding simulated collision space to the first simulated collision space model.

[0109] Further, the system is also used to implement the following functions: Perform overrun evaluation on each tunneling path in the first tunneling path planning domain according to the technical parameter constraints to obtain multiple path overrun coefficients; classify the first tunneling path planning domain according to the multiple path overrun coefficients to obtain a first cluster of tunneling paths with a path overrun coefficient less than a predetermined value, and a second cluster of tunneling paths with a path overrun coefficient greater than or equal to the predetermined value; perform adaptive adjustment on the second cluster of tunneling paths according to the technical parameter constraints to obtain a third cluster of tunneling paths; merge the first cluster of tunneling paths and the third cluster of tunneling paths to generate the second tunneling path planning domain.

[0110] Further, the system is also used to implement the following functions: Set a tunneling evaluation expectation according to the predetermined tunneling evaluation factor, where the predetermined tunneling evaluation factor includes tunneling efficiency and tunneling safety; perform optimization analysis on the second tunneling path planning domain according to the tunneling evaluation expectation to establish a third tunneling path planning domain; perform weight assignment according to the predetermined tunneling evaluation factor to establish a tunneling fitness evaluation function; perform maximization optimization of the tunneling fitness on the third tunneling path planning domain according to the tunneling fitness evaluation function to generate the tunneling path optimization result.

[0111] Further, the system is also used to implement the following functions: Extract the nth tunneling path according to the second tunneling path planning domain, where n is a positive integer; perform feature prediction based on the nth tunneling path to determine the nth tunneling prediction result, and the nth tunneling prediction result includes the nth predicted tunneling efficiency coefficient and the nth predicted tunneling safety coefficient; determine whether the nth tunneling prediction result meets the tunneling evaluation expectation; if the nth tunneling prediction result meets the tunneling evaluation expectation, add the nth tunneling path to the third tunneling path planning domain; if the nth tunneling prediction result does not meet the tunneling evaluation expectation, eliminate the nth tunneling path.

[0112] Further, the system is also used to implement the following functions: Collect the above-ground and underground parameters of the area to be tunneled to obtain a tunneling space data set; perform three-dimensional reconstruction based on the tunneling space data set to obtain a three-dimensional model of the tunneling space; perform risk prediction on each point in the area to be tunneled according to the tunneling space data set to obtain the risk prediction coefficient of each point; render the risk prediction coefficient of each point to the three-dimensional model of the tunneling space to generate the risk distribution model of the tunneling space.

[0113] Further, the system is also used to implement the following functions: Perform feature recognition on each point according to the tunneling space data set to obtain the feature data of each point; train a point risk prediction network based on a recurrent neural network; input the feature data of each point into the point risk prediction network to generate the risk prediction coefficient of each point.

[0114] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0115] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0116] This specification and the drawings are only exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. An intelligent control method for the driving path of a roadheader, characterized in that The method includes: Performing risk prediction on the area to be tunnelled by the tunnelling machine and establishing a risk distribution model for the tunnelling space; Performing hierarchical bounding box collision detection on the risk distribution model of the tunnelling space according to the tunnelling machine, and constructing a safe tunnelling motion space model; Performing tunnelling path planning on the safe tunnelling motion space model according to the tunnelling target of the area to be tunnelled, and building a first tunnelling path planning domain; Performing overrun feedback optimization on the first tunnelling path planning domain according to the technical parameter constraints of the tunnelling machine, and establishing a second tunnelling path planning domain; Performing joint optimization on the second tunnelling path planning domain according to a predetermined tunnelling evaluation factor to determine the tunnelling path optimization result; Performing tunnelling path tracking control on the tunnelling machine according to the tunnelling path optimization result.

2. The intelligent control method for the tunneling path of a roadheader according to claim 1, characterized in that, Performing hierarchical bounding box collision detection on the risk distribution model of the tunnelling space according to the tunnelling machine, and constructing a safe tunnelling motion space model, including: Performing modeling according to the tunnelling machine to construct a tunnelling machine model; Based on the tunnelling machine model, performing simulated collision detection on the risk distribution model of the tunnelling space according to multiple-level bounding boxes, and constructing a first safe tunnelling space model, where the multiple-level bounding boxes include a plurality of box-shaped colliders and a plurality of grid colliders; Performing point cleaning on the risk distribution model of the tunnelling space according to a point risk threshold to construct a second safe tunnelling space model; Finding the intersection of the first safe tunnelling space model and the second safe tunnelling space model to generate the safe tunnelling motion space model.

3. The intelligent control method for the tunneling path of a roadheader according to claim 2, characterized in that, Based on the tunnelling machine model, performing simulated collision detection on the risk distribution model of the tunnelling space according to multiple-level bounding boxes, and constructing a first safe tunnelling space model, including: Based on the tunnelling machine model, performing simulated collision detection on the risk distribution model of the tunnelling space respectively according to the multiple-level bounding boxes to construct a first simulated collision space model; Performing random combination according to the multiple-level bounding boxes to obtain a multi-level bounding structure; Based on the tunnelling machine model, performing simulated collision detection on the risk distribution model of the tunnelling space according to the multi-level bounding structure to construct a second simulated collision space model; Performing union recognition on the first simulated collision space model and the second simulated collision space model to generate a simulated collision integrated space model; Taking the difference set between the simulated collision integrated space model and the risk distribution model of the tunnelling space as the first safe tunnelling space model.

4. The intelligent control method for the tunneling path of a roadheader according to claim 3, characterized in that, Based on the tunnelling machine model, performing simulated collision detection on the risk distribution model of the tunnelling space respectively according to the multiple-level bounding boxes, and constructing a first simulated collision space model, including: Extracting a first bounding box according to the multiple-level bounding boxes; Performing abstract bounding on the tunnelling machine model according to the first bounding box to obtain a first bounded tunnelling machine model; Performing multiple simulated collisions on the risk distribution model of the tunnelling space according to the first bounded tunnelling machine model to obtain a plurality of simulated collision spaces; Fusing the plurality of simulated collision spaces to obtain a first bounded simulated collision space, and adding the first bounded simulated collision space to the first simulated collision space model.

5. The intelligent control method for the tunneling path of a roadheader according to claim 1, characterized in that, Perform over-limit feedback optimization on the first tunneling path planning domain according to the technical parameter constraints of the tunneling machine, and establish a second tunneling path planning domain, including: Perform over-limit evaluation on each tunneling path within the first tunneling path planning domain according to the technical parameter constraints to obtain multiple path over-limit coefficients; Classify the first tunneling path planning domain according to the multiple path over-limit coefficients to obtain a first cluster of tunneling paths with a path over-limit coefficient less than a predetermined value, and a second cluster of tunneling paths with a path over-limit coefficient greater than or equal to the predetermined value; Perform adaptive adjustment on the second cluster of tunneling paths according to the technical parameter constraints to obtain a third cluster of tunneling paths; Merge the first cluster of tunneling paths and the third cluster of tunneling paths to generate the second tunneling path planning domain.

6. The intelligent control method for the tunneling path of a roadheader according to claim 1, wherein Perform joint optimization on the second tunneling path planning domain according to a predetermined tunneling evaluation factor to determine the tunneling path optimization result, including: Set a tunneling evaluation expectation according to the predetermined tunneling evaluation factor, where the predetermined tunneling evaluation factor includes tunneling efficiency and tunneling safety; Perform optimization analysis on the second tunneling path planning domain according to the tunneling evaluation expectation to establish a third tunneling path planning domain; Perform weight allocation according to the predetermined tunneling evaluation factor to establish a tunneling fitness evaluation function; Perform optimization to maximize the tunneling fitness on the third tunneling path planning domain according to the tunneling fitness evaluation function to generate the tunneling path optimization result.

7. The intelligent control method for the tunneling path of a roadheader according to claim 6, characterized in that, Perform optimization analysis on the second tunneling path planning domain according to the tunneling evaluation expectation to establish a third tunneling path planning domain, including: Extract the nth tunneling path from the second tunneling path planning domain, where n is a positive integer; Perform feature prediction according to the nth tunneling path to determine the nth tunneling prediction result, where the nth tunneling prediction result includes the nth predicted tunneling efficiency coefficient and the nth predicted tunneling safety coefficient; Judge whether the nth tunneling prediction result meets the tunneling evaluation expectation; If the nth tunneling prediction result meets the tunneling evaluation expectation, add the nth tunneling path to the third tunneling path planning domain; If the nth tunneling prediction result does not meet the tunneling evaluation expectation, eliminate the nth tunneling path.

8. The intelligent control method for the tunneling path of a roadheader according to claim 1, characterized in that, Perform risk prediction on the area to be tunneled by the tunneling machine to establish a tunneling space risk distribution model, including: Collect the above-ground and underground parameters of the area to be tunneled to obtain a tunneling space data set; Perform three-dimensional reconstruction according to the tunneling space data set to obtain a three-dimensional model of the tunneling space; Perform risk prediction on each point in the area to be tunneled according to the tunneling space data set to obtain the risk prediction coefficient of each point; Render the risk prediction coefficients of each point to the three-dimensional model of the tunneling space to generate the tunneling space risk distribution model.

9. The intelligent control method for the tunneling path of a roadheader according to claim 8, characterized in that, Perform risk prediction on each point in the area to be tunneled according to the tunneling space data set to obtain the risk prediction coefficient of each point, including: Perform feature recognition on each point according to the tunneling space data set to obtain the feature data of each point; Train a point risk prediction network based on a recurrent neural network; Input the feature data of each point into the point risk prediction network to generate the risk prediction coefficients of each point.

10. An intelligent control system for the tunneling path of a roadheader, characterized in that, The system is used to implement the intelligent control method for the tunneling path of a roadheader according to any one of claims 1-9. The system includes: A risk prediction module for predicting the risk of the area to be tunnelled by the roadheader and establishing a risk distribution model for the tunnelling space; A safe tunnelling motion space model construction module for performing hierarchical bounding volume collision detection on the risk distribution model of the tunnelling space according to the roadheader to construct a safe tunnelling motion space model; A tunnelling path planning module for planning the tunnelling path of the safe tunnelling motion space model according to the tunnelling target of the area to be tunnelled and building a first tunnelling path planning domain; An overrun feedback optimization module for performing overrun feedback optimization on the first tunnelling path planning domain according to the technical parameter constraints of the roadheader to establish a second tunnelling path planning domain; A tunnelling path optimization result determination module for jointly optimizing the second tunnelling path planning domain according to a predetermined tunnelling evaluation factor to determine the tunnelling path optimization result; A tunnelling path tracking control module for performing tunnelling path tracking control on the roadheader according to the tunnelling path optimization result.

Images