A narrow space door type heavy load carrying robot cooperative dynamic path planning method

Abstract

This invention discloses a collaborative dynamic path planning method for gantry heavy-duty handling robots in confined spaces, comprising: acquiring environmental space data, robot pose data, and load data for each gantry heavy-duty handling robot; acquiring a three-dimensional prior model and dividing it into ordinary areas and key restricted areas; constructing a robot boundary model; matching and verifying the robot boundary model with the key restricted areas to determine the robot's passability status in each key restricted area and generating key passage constraints; determining passage paths based on the key restricted areas and key passage constraints, generating the occupancy order and expected passage sequence for each key restricted area; performing conflict detection on the occupancy order and expected passage sequence of each key restricted area; determining the passage strategy for each robot through the key restricted areas with the objective of minimizing the number of key restricted areas; and outputting the path planning result for each robot.

Description

Technical Field

[0001] This invention relates to the field of collaborative dynamic path planning technology for gantry heavy-duty handling robots, and in particular to a collaborative dynamic path planning method for gantry heavy-duty handling robots in confined spaces. Background Technology

[0002] With the increasing demand for underground storage facilities, mine tunnels, mountain cave storage, and the transfer of large engineering materials, gantry cranes, due to their strong load-bearing capacity, adaptability to large and irregularly shaped goods, and ability to complete handling operations in narrow and confined spaces, are gradually being used in multi-robot collaborative transfer scenarios in complex environments such as underground storage facilities and mountain caves. To improve transportation turnover efficiency per unit time, multiple robots are typically required to enter the same work area simultaneously on construction sites and complete loading, unloading, passage, and handover in batches.

[0003] However, narrow passageways such as caves generally have structural characteristics such as limited passage width, limited turning radius at bends, single-machine passage in some areas, and discrete distribution of passing / loading / unloading points and waiting areas. In addition, there may be dynamic factors such as personnel activities, temporary stacking of goods or construction equipment inside the cave. This means that when multiple robots operate in cooperation, they must not only complete their own path selection and short-distance obstacle avoidance, but also face the bottleneck competition caused by shared bends, shared narrow passages, and shared passing / loading / unloading areas.

[0004] Existing multi-robot collaborative methods primarily rely on geometrically passable paths. Planning and control models typically focus on robot body contours and near-field obstacle avoidance, making it difficult to accurately characterize the actual passable boundaries of "narrow bottleneck sections" at the scheduling level. Furthermore, existing methods often lack dedicated modeling and constraint solving for the temporal relationships of shared bottleneck section occupancy. This results in each robot potentially having its own feasible path in the spatial dimension, but when actually entering key locations such as shared bends / narrow passages / gap zones, overlapping occupancy time windows can easily occur, triggering mutual exclusion constraint failures. This further leads to reliance on conservative waiting or post-event conflict resolution, resulting in decreased efficiency and even feasibility limitations or local deadlocks under complex coupling.

[0005] The purpose of this invention is to design a collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces, addressing the problems existing in the prior art. Summary of the Invention

[0006] In view of this, the purpose of this invention is to propose a collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces, which can solve the above-mentioned problems.

[0007] This invention provides a collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces, comprising: S1 acquires the environmental spatial data, robot pose data and load data of the local area where each gantry heavy-duty handling robot is located, preprocesses and aligns the various types of data, obtains the three-dimensional prior model of the current working environment, and divides it into ordinary areas and key restricted areas. S2 constructs a robot boundary model based on robot pose data and load data, matches and verifies the robot boundary model with key restricted areas, determines the robot's passability status in each key restricted area, and generates key passage constraints. Based on key restricted areas and key passage constraints, S3 determines the passage path of each robot along the key restricted areas by combining the current position and target position of each gantry heavy-duty handling robot, and generates the occupancy order and expected passage sequence of each key restricted area; S4 performs conflict detection on the occupancy order and expected passage sequence of each key restricted area. Under the premise of satisfying the key passage constraints, and with the goal of minimizing the number of key restricted areas, it determines the passage strategy of each robot through the key restricted areas and outputs the path planning results of each robot.

[0008] The beneficial effects of this invention are: First, by registering and constructing a 3D prior environment model, the geometric relationships between the robot's operating space, obstacles, and key confined areas are placed within a computable spatial framework, reducing planning deviations caused by environmental uncertainties. The continuous space is discretized and formed into an organizational structure usable for search / modeling (such as basic data for topological representation), providing a consistent spatial expression for subsequent key confined area sequence extraction, occupancy time-series calculation, and conflict optimization. This ensures that the planning process revolves around "bottleneck areas that affect safety" from the outset, rather than blindly searching the entire space, improving computational efficiency and interpretability.

[0009] Secondly, it unifies the dynamic risks such as geometric clearance and attitude / eccentric load stability into quantitative indicators of passage margin / equivalent attitude stability value, realizing the mapping from physical constraints to determinable passage modes. By using a hierarchical approach of "passage margin and safety margin", the risks of key restricted areas are refined from whether they can be passed to how they can be passed, thereby generating key passage constraint parameters that can be used for scheduling.

[0010] Third, spatial conflicts are transformed into time windows and occupancy sequences for key restricted areas, enabling multi-robot collaboration to move beyond geometric feasibility to temporal feasibility. By extracting only the sequence of key restricted areas from candidate paths and then calculating the expected entry / exit intervals segment by segment, the scope of unnecessary temporal conflict calculations is significantly reduced, improving overall scheduling efficiency. The occupancy sequence and passage time sequence of each key restricted area are output, providing directly usable constraint inputs for subsequent conflict detection and evolutionary optimization.

[0011] Fourth, by employing conflict detection and adaptive repair (advancing only the entry time without altering the selected key restricted area sequence), the system ensures that the solution satisfies the mutual exclusion constraint after repair, thus achieving a feasibility-first collaborative scheduling closed loop. Using Pareto non-dominated sorting and congestion distance to preserve non-dominated solutions simultaneously considers minimizing the passage through restricted sections (reducing collision risk) and global feasibility / congestion level (avoiding single-point greed leading to overall failure), increasing the probability of convergence to a high-quality feasible solution. Finally, by combining the sorting of solutions approximating the ideal solution, a compromise optimal solution is selected, resulting in directly executable path planning results for each robot, balancing safety and efficiency while reducing reliance on human experience parameters. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings required in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of the method in this embodiment.

[0014] Figure 2 This is a schematic diagram showing the placement of each sensor in this embodiment. Detailed Implementation

[0015] To facilitate understanding by those skilled in the art, the structure of the present invention will now be described in further detail with reference to the accompanying drawings. It should be understood that, unless otherwise specified, the order of the steps mentioned in this embodiment can be adjusted according to actual needs, and they can even be executed simultaneously or partially simultaneously.

[0016] like Figure 1 As shown, this embodiment of the invention provides a collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces, comprising: S1 acquires the environmental spatial data, robot pose data and load data of the local area where each gantry heavy-duty handling robot is located, preprocesses and aligns the various types of data, obtains the three-dimensional prior model of the current working environment, and divides it into ordinary areas and key restricted areas. S101 collects 3D point cloud data of the environment through a lidar deployed in the middle of the top beam of the robot, and collects infrared environmental image data through infrared cameras deployed on both sides of the top beam of the robot. S102 collects the robot's three-axis angular velocity and three-axis acceleration by IMU inertial measurement units deployed on the two pillars on both sides of the robot, and calculates the roll angle, pitch angle, heading change and attitude change rate based on the three-axis angular velocity and three-axis acceleration. S103 collects load data for each support wheel by embedding force sensors in each support wheel, and calculates the total load, lateral load center of gravity offset, and longitudinal load center of gravity offset based on the load data of each support wheel. The calculation formula is as follows: , , in, This indicates the offset of the center of gravity under lateral load. This indicates the offset of the longitudinal load center of gravity. These are the load data for each support wheel. Indicates the spacing between the left and right supports. Indicates the distance between the front and rear supports. Indicates the total load; S104 performs filtering, noise reduction, and outlier removal on the multi-source data acquired by each sensor, followed by data alignment and unified coordinate mapping for the robot. S105 registers the 3D prior model with the environmental spatial data. Based on the registered 3D prior model, areas that allow a single robot to pass through are designated as critical restricted areas, while areas that allow more than one robot to pass through are designated as ordinary restricted areas.

[0017] In this step, caves often possess natural long passageways and large internal spaces. Furthermore, the interior of caves is less affected by external temperature, humidity, and weather conditions, allowing them to accommodate larger-scale storage and turnover needs within limited surface space. Concentrating large quantities of goods in caves can create a relatively centralized and easily managed logistics system.

[0018] like Figure 2 As shown, portal frames typically form a closed support system for the load-bearing path through upper and lower / left and right frames (specific structures are existing technologies). This provides stronger structural rigidity and stress reliability when handling large-mass materials. In caves, LiDAR can accurately acquire information about the surrounding environment. The response characteristics of infrared cameras under certain lighting, dust, and humidity conditions complement those of lasers. The placement of cameras on both sides enhances the redundancy of observation of distances to the left and right side walls and local obstacles. Due to the coupling of heavy loads and narrow spaces, the robot's posture is a critical state affecting passage safety. Placing IMUs on the side supports can better reflect posture changes caused by forces / movements on both sides of the portal frame.

[0019] Wheel-end load is a direct physical quantity under heavy load conditions, and tunnel transport falls under this category. When a gantry heavy-duty transport robot walks, turns, or traverses uneven ground / narrow spaces, the dynamic response of the vehicle body and the load changes, especially in the following ways: due to the imbalance of left and right / front and rear support loads, the robot's posture tends to tilt and swing back. These posture changes cause the robot's equivalent occupied area in narrow spaces to change. At narrow passages and intersections, this change may determine whether it can "skim through." Therefore, it is necessary to use wheel load data to deduce the center of gravity shift and load magnitude, thereby mapping the "mechanical state" into the planning constraints.

[0020] In narrow tunnels, gantry heavy-duty transport robots often face multiple coupling risks: spatial geometric constraints (small clearance, changes in the size of corners / narrow passages / intersections), multi-robot interference constraints (collision risks caused by meeting, following, and cross-scheduling), and load and attitude constraints (swaying / off-center loading caused by heavy loads increases the equivalent space occupied by the robot). Therefore, it is necessary to pre-classify the restricted locations in the environment according to the number of robots allowed to pass, for subsequent path planning.

[0021] A three-dimensional prior environment model can be constructed by combining offline CAD / BIM models with existing 3D digital maps. The CAD / BIM model provides the geometric dimensions, spatial topology, and labeling information of facilities / loading areas for the underground structure; the existing 3D digital map provides historically surveyed three-dimensional surface / feature distribution and coordinate references. By unifying the coordinate system and aligning the semantics of these two types of prior data, a three-dimensional prior model containing both general and critical restricted areas is formed, serving as the basis for subsequent sensor point cloud registration and dynamic updates.

[0022] S2 constructs a robot boundary model based on robot pose data and load data, matches and verifies the robot boundary model with key restricted areas, determines the robot's passability status in each key restricted area, and generates key passage constraints. Based on the length, width, and height parameters of the gantry heavy-duty handling robot body, S201 establishes a basic robot boundary model. The basic robot boundary model is then corrected according to the robot's roll angle, pitch angle, yaw change, attitude change rate, lateral load center offset, and longitudinal load center offset to obtain the robot boundary model. S202 maps the robot boundary model to the corresponding key restricted areas, matches the robot boundary model with the effective passage space parameters of each key restricted area, and calculates the passage margin data. S2021 calculates the width margin based on the difference between the minimum net width of the critical restricted area and the lateral projection width occupied by the robot in the current posture. S2022 calculates the height margin based on the difference between the minimum net height of the critical restricted area and the vertical occupancy projection height of the robot in its current posture; S2023 calculates the turning margin based on the difference between the allowable curvature of the critical restricted area and the equivalent curvature under the constraint of the robot's current attitude change rate. S2024 calculates the robot's lateral and longitudinal stability based on the lateral load center of gravity offset and the longitudinal load center of gravity offset. The calculation formula is as follows: , , in, Indicates lateral stability. Indicates longitudinal stability. This indicates the offset of the center of gravity under lateral load. This indicates the offset of the longitudinal load center of gravity. Indicates the spacing between the left and right supports. Indicates the distance between the front and rear supports; S2025 minimizes both lateral and longitudinal stability, then introduces an attitude-related penalty term to calculate the equivalent attitude stability value. The calculation formula is as follows: , in, This represents the equivalent attitude stability value. Indicates the roll angle. Indicates pitch angle, Indicates the rate of change of roll angle. Indicates the rate of change of pitch angle. Indicates a penalty item. This represents a function that takes the minimum value. S2026 calculates the attitude stability margin based on the stability safety threshold and the equivalent attitude stability value.

[0023] In this step, the passage margin data is used to characterize whether the robot can safely pass through a certain critical restricted area under its current pose and load state, and to provide the size of the remaining safety margin. It typically includes components such as width margin, height margin, turning margin, and stability margin. For example, if the clearance at the narrow arched entrance of a cave is very small, and the robot's pitch angle is large, causing the vertical occupancy projection to increase, the calculated height margin may be less than or equal to 0. The system would then determine that this critical restricted area is impassable or can only be passed through in a very low-posture risk mode.

[0024] The equivalent attitude stability value is used to combine static off-center load stability and dynamic attitude risk into a comparable indicator. First, lateral stability and longitudinal stability are used to characterize the stability margin in the weakest lateral / longitudinal directions. Then, the minimum value is taken to reflect that "the most dangerous direction determines safety." Subsequently, a penalty term is used to deduct for situations where the attitude angle increases and the rate of attitude change intensifies, making the margin automatically more conservative in states of "approaching instability or rapidly deteriorating." Example: If, after a loaded load has just left a narrow section of the tunnel wall, the wheel-end load indicates that the lateral center of gravity offset is close to the limit, and the IMU calculation shows a large rate of change in the roll angle, then the equivalent attitude stability value will be significantly reduced.

[0025] Among them, the penalty item , in, Represents a dimensional coefficient. This represents the corresponding weighting coefficient. Indicates the roll angle safety threshold. Indicates the safe threshold for pitch angle. This indicates the safe threshold for the rate of change of roll angle. This indicates the safe threshold for the rate of change of pitch angle; The penalty item adopts the form of threshold triggering and secondary penalty, when When all quantities are within their respective safety thresholds, the penalty term is zero or small, thus ensuring that the stability margin is primarily dominated by static stability; when any quantity exceeds its corresponding safety threshold, the penalty term is applied... Only the portion exceeding the threshold is penalized, avoiding excessive conservatism caused by normal posture fluctuations within the threshold.

[0026] Meanwhile, the secondary penalty gradually applies the risk reduction for "just exceeding the threshold", while the state of "significantly exceeding the threshold or rapidly increasing the magnitude of exceeding the threshold" will trigger a stronger reduction effect, making the equivalent attitude stability value more sensitive and conservative when approaching the instability boundary.

[0027] S203 marks the corresponding passage mode for the robot in each critical restricted area based on the passage margin; S2031 If any passage margin data is ≤0, then mark the critical restricted area as an impassable section for the robot; S2032 If all passage margin data are greater than 0 and there is at least one passage margin data that does not exceed its corresponding safety margin, then mark the critical restricted area as the restricted passage section of the robot. S2033 If all passage margin data exceed their corresponding safety margin, then mark the critical restricted area as the passable section of the robot.

[0028] S204 generates corresponding key traffic constraint parameters based on the traffic pattern of the key restricted area. The key traffic constraint parameters include: maximum permissible speed, maximum permissible attitude angular velocity, maximum permissible curvature, whether passing is allowed, permissible traffic window, and whether entering the stop is allowed.

[0029] In this step, the "passage margin" is used to determine whether a certain area can be passed under the current state. Then, it is compared with the "safety margin" to distinguish between "passable" and "sufficient safety margin", thus obtaining three modes: impassable, restricted passage, and passable.

[0030] Example: Near a tunnel junction, if the width margin of the critical restricted area is positive but the stability margin is lower than the safety threshold, the robot will be marked as a "restricted passage section". This will generate critical passage constraints, such as reducing the maximum allowable speed, limiting the maximum allowable attitude angular velocity, and possibly prohibiting oncoming traffic. Conversely, if all three types of margins in the area are significantly abundant, passable constraint parameters can be generated to improve throughput efficiency.

[0031] Based on key restricted areas and key passage constraints, S3 determines the passage path of each robot along the key restricted areas by combining the current position and target position of each gantry heavy-duty handling robot, and generates the occupancy order and expected passage sequence of each key restricted area; S301 combines the passage patterns of the key restricted areas with the robot's starting point and ending point to filter the candidate spatial passage sequences of each robot, and extracts the key restricted area sequence based on the candidate spatial passage sequences; S3011 discretizes the registered 3D prior model into a topological graph, and maps each key restricted region to the edge of the topological graph after discretization. For each robot, S3012 maps its start and end points onto the topology map, filters key restricted areas marked as impassable sections, performs path search on the filtered topology map, and obtains the candidate spatial passage sequence for each robot. S3013 extracts the key restricted regions on each candidate path in the candidate space travel sequence of each robot, and obtains the key restricted region sequence corresponding to the candidate path.

[0032] In this step, during collaborative narrow passage, what truly determines whether safe passage is possible and the necessary temporal constraints are not ordinary corridor segments, but rather those critical restricted areas (such as narrow passages that allow only one vehicle to pass at a time / near the arch beam). Therefore, even if a candidate spatial path contains many ordinary edges on the topology graph, these ordinary edges typically do not trigger strong constraints such as "exclusive occupancy / no passing / stop ahead".

[0033] S302 calculates the passage time for each key restricted area along its key restricted area sequence from the starting point to the end point, forming the expected entry and exit time intervals for the robot in each key restricted area segment; S3021 For any critical restricted area, calculate the expected passage duration of the critical restricted area based on the critical passage constraint parameters of the critical restricted area and the robot's passage mode in the critical restricted area; S3022 updates the expected entry time of the next critical restricted area based on the transition conditions required for the connection of the topology segment during the topology connection process between adjacent critical restricted areas. Based on the updated expected entry time, it calculates the expected passage duration of the next critical restricted area, thereby obtaining the time interval between the expected entry time and the expected exit time of the robot in each critical restricted area.

[0034] S303 For each critical restricted area, collect all robot records that satisfy the passage of the critical restricted area, sort the collected robot records according to their expected entry time, and the sequence corresponding to the sorting result is the segment occupancy order of the critical restricted area. The set of expected entry / exit time intervals corresponding to the sorting sequence is the expected passage time sequence of the critical restricted area.

[0035] In this step, because multiple robots may experience time conflicts within a critical restricted area—even if their spatial paths differ, if two robots overlap in time within the same critical restricted area, they may experience failed encounters, congestion, or mechanical collisions—it is necessary to calculate the "entry / exit time" of each robot through the critical restricted area. Then, all robots within the same critical restricted area are sorted by their entry time to obtain the occupancy order and expected passage sequence. Example: At the same narrow archway, robot A is expected to enter and leave at [10s, 16s], while robot B is expected to enter and leave at [12s, 18s]. Due to the overlapping time periods, the narrow archway must allow A to pass first (occupancy order A→B), and a "passage window / necessary waiting time" must be generated for B, thus transforming the spatial conflict into an executable time constraint.

[0036] The traffic constraint parameters of key restricted areas (such as maximum permissible speed, maximum permissible attitude angular velocity, whether passing is allowed, and whether entering the front stop is allowed) will directly affect the robot's expected duration and transition time in that area.

[0037] In addition, the topological connection between adjacent critical restricted regions usually has "transition conditions" (for example, the attitude must be stable after leaving the critical region before entering the next critical region). Therefore, the time of entering the next critical region needs to be updated based on the departure time of the previous segment.

[0038] S4 performs conflict detection on the occupancy order and expected passage sequence of each key restricted area. Under the premise of satisfying the key passage constraints, and with the goal of minimizing the number of key restricted areas, it determines the passage strategy of each robot through the key restricted areas and outputs the path planning results of each robot.

[0039] S401 uses the path selection and entry time offset of all robots in a coordinated manner as individual coding objects. The decision variables include: the path selection variable of each robot corresponding to the candidate space passage sequence and the entry time offset of each robot in the key restricted area corresponding to its selected path. S402 decodes any individual in the population to obtain the time window in which each robot occupies the key restricted area, and obtains the sequence of key restricted areas selected by each robot after decoding. For each robot, the number of times the key restricted area marked as a restricted passage segment appears in its selected path is counted to form a multi-objective evaluation vector for Pareto comparison. In this step, each robot tends to choose candidate paths that appear less frequently in restricted access sections. However, different robots may still point to the same set of restricted bottlenecks on the graph. Even if a robot chooses a restricted section less frequently, as long as multiple robots still form overlapping time windows in the same restricted section, mutual exclusion / collision constraints will be triggered. Therefore, the conflict arises from the "time overlap of the same restricted section." For example, when robot A chooses a detour to avoid traversing a restricted section less, robot B or C may be forced to choose an alternative path closer to the same restricted section, resulting in overlapping time windows in another restricted section.

[0040] Therefore, we should first make the search biased towards avoiding restricted sections (minimizing the number of times restricted passage sections appear in the multi-objective evaluation vector), which is in line with the principle of reducing collision risk from the root.

[0041] S403 When a decoded individual violates the mutual exclusion constraint in any critical restricted area, adaptive repair is performed on the individual to restore the mutual exclusion feasibility without changing the selected path sequence determined by each robot in the decoded process. S4031 checks each key restricted area to see if two robots occupy overlapping time windows. If a conflict exists, it records the key restricted area of ​​the conflict and the corresponding pair of conflicting robots. Within the same key restricted area, it identifies the leading robot that entered earlier and marks the remaining conflicting parties as lagging robots that need to be advanced. S4032 performs entry time advancement only on the lagging robot in its conflict-critical restricted area, so that its occupation time window does not overlap with the occupation time window of the leading robot, and updates the exit time in the critical restricted area accordingly. S4033 calculates the propulsion amount at the current entry time and updates the entry time by sequentially cascading the propulsion amount according to the key restricted areas subsequently traversed by the lagging robot. S4034 performs conflict detection again on the repaired individual. If a conflict still exists, iterative updates continue until the individual satisfies the mutual exclusion constraint in all critical restricted areas.

[0042] In this step, even if a path with fewer restricted sections is selected, conflicts may still occur, thus requiring adaptive repair. The repair does not change the sequence of critical restricted sections for each robot; it only advances the entry time on the conflicting critical restricted sections, eliminating overlapping time windows and ensuring mutual exclusivity.

[0043] S404 performs Pareto non-dominated sorting on the multi-objective evaluation vectors of the repaired feasible individuals to obtain the non-dominated frontier level. Each component of the multi-objective evaluation vector represents the degree of minimization of the occurrence of restricted passage sections in the selected path of each robot. The non-dominated solution is stored in the external archive set. S405 selects the parent robot to generate offspring based on the non-dominated hierarchy and congestion distance, performs cross-variation and mutation of the path selection variables of each robot, cross-variation and mutation of the entry time offset of each robot, performs conflict detection and adaptive repair on the generated offspring, calculates the updated multi-objective evaluation vector, and updates the offspring to enter the next round of iteration. When the iteration reaches the preset upper limit, S406 calculates the relative proximity by sorting the solutions from the non-dominated front obtained by merging the parent and child solutions, and selects the solution with the largest relative proximity as the path planning output.

[0044] In this step, different solutions may result in some robots having fewer restricted passages, but sacrificing other robots to lead to greater overall congestion / tighter timing. The Pareto mechanism allows these compromise solutions to be preserved, rather than forcibly pursuing a single minimum. For example, there are two feasible solutions: Solution 1 reduces the number of restricted passages for A, but makes B more congested; Solution 2 is the opposite, but neither dominates the other, so both are preserved to wait for subsequent crossover mutations to find a better combination.

[0045] Parents are selected based on non-dominated hierarchy and congestion distance to generate offspring: crossover and mutation are performed on path selection variables, as well as on entry time offsets. Conflict detection and adaptive repair are then performed on each offspring to obtain new feasible individuals and an updated multi-objective evaluation vector. If the current solution leads to frequent conflicts at R1, the offspring may allow a robot to switch between candidate paths (reducing the frequency of restricted sections at the spatial level) while simultaneously increasing / decreasing the entry offset (staggering time points), thereby improving both feasibility and the evaluation vector.

[0046] When the iteration reaches its limit, the undominated front solution set obtained by merging the parent and child generations is sorted to approximate the ideal solution. The relative proximity is calculated, and the solution with the highest proximity is selected as the final path planning output. Among several compromise undominated solutions, a certain solution may not have the "fewest restricted sections" for every robot, but it is more balanced in terms of feasibility and congestion, and therefore has the highest relative proximity, and is output as the final cooperative passage strategy.

[0047] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0048] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0049] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0050] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0051] It should be noted that any reference signs placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The words first, second, and third, etc., do not indicate any order. These words can be interpreted as names.

[0052] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0053] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0054] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0055] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms should not be construed as necessarily referring to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

Claims

1. A collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces, characterized in that, include: S1 acquires the environmental spatial data, robot pose data and load data of the local area where each gantry heavy-duty handling robot is located, preprocesses and aligns the various types of data, obtains the three-dimensional prior model of the current working environment, and divides it into ordinary areas and key restricted areas. S2 constructs a robot boundary model based on robot pose data and load data, matches and verifies the robot boundary model with key restricted areas, determines the robot's passability status in each key restricted area, and generates key passage constraints. Based on key restricted areas and key passage constraints, S3 determines the passage path of each robot along the key restricted areas by combining the current position and target position of each gantry heavy-duty handling robot, and generates the occupancy order and expected passage sequence of each key restricted area; S4 performs conflict detection on the occupancy order and expected passage sequence of each key restricted area. Under the premise of satisfying the key passage constraints, and with the goal of minimizing the number of key restricted areas, it determines the passage strategy of each robot through the key restricted areas and outputs the path planning results of each robot.

2. The collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces according to claim 1, characterized in that, The process of acquiring environmental spatial data, robot pose data, and load data for the local area where each gantry heavy-duty handling robot is located, preprocessing and aligning the various types of data, obtaining a three-dimensional prior model of the current working environment, and dividing it into ordinary areas and critical restricted areas includes: S101 collects 3D point cloud data of the environment through a lidar deployed in the middle of the top beam of the robot, and collects infrared environmental image data through infrared cameras deployed on both sides of the top beam of the robot. S102 collects the robot's three-axis angular velocity and three-axis acceleration by IMU inertial measurement units deployed on the two pillars on both sides of the robot, and calculates the roll angle, pitch angle, heading change and attitude change rate based on the three-axis angular velocity and three-axis acceleration. S103 collects load data for each support wheel by embedding force sensors in each support wheel, and calculates the total load, lateral load center of gravity offset, and longitudinal load center of gravity offset based on the load data of each support wheel. The calculation formula is as follows: ， ， in, This indicates the offset of the center of gravity under lateral load. This indicates the offset of the longitudinal load center of gravity. These are the load data for each support wheel. Indicates the spacing between the left and right supports. Indicates the distance between the front and rear supports. Indicates the total load; S104 performs filtering, noise reduction, and outlier removal on the multi-source data acquired by each sensor, followed by data alignment and unified coordinate mapping for the robot. S105 registers the 3D prior model with the environmental spatial data. Based on the registered 3D prior model, areas that allow a single robot to pass through are designated as critical restricted areas, while areas that allow more than one robot to pass through are designated as ordinary restricted areas.

3. The collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces according to claim 1, characterized in that, The process of constructing a robot boundary model based on robot pose data and load data, matching and verifying the robot boundary model against key restricted areas, determining the robot's traversability status in each key restricted area, and generating key traversability constraints includes: Based on the length, width, and height parameters of the gantry heavy-duty handling robot body, S201 establishes a basic robot boundary model. The basic robot boundary model is then corrected according to the robot's roll angle, pitch angle, yaw change, attitude change rate, lateral load center offset, and longitudinal load center offset to obtain the robot boundary model. S202 maps the robot boundary model to the corresponding key restricted areas, matches the robot boundary model with the effective passage space parameters of each key restricted area, and calculates the passage margin data. S203 marks the corresponding passage mode for the robot in each critical restricted area based on the passage margin; S204 generates corresponding key traffic constraint parameters based on the traffic pattern of the key restricted area. The key traffic constraint parameters include: maximum permissible speed, maximum permissible attitude angular velocity, maximum permissible curvature, whether passing is allowed, permissible traffic window, and whether entering the stop is allowed.

4. The collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces according to claim 3, characterized in that, The process of mapping the robot boundary model to the corresponding key confined areas, matching the robot boundary model with the effective passage space parameters of each key confined area, and calculating the passage margin data includes: S2021 calculates the width margin based on the difference between the minimum net width of the critical restricted area and the lateral projection width occupied by the robot in the current posture. S2022 calculates the height margin based on the difference between the minimum net height of the critical restricted area and the vertical occupancy projection height of the robot in its current posture; S2023 calculates the turning margin based on the difference between the allowable curvature of the critical restricted area and the equivalent curvature under the constraint of the robot's current attitude change rate. S2024 calculates the robot's lateral and longitudinal stability based on the lateral load center of gravity offset and the longitudinal load center of gravity offset. The calculation formula is as follows: ， ， in, Indicates lateral stability. Indicates longitudinal stability. This indicates the offset of the center of gravity under lateral load. This indicates the offset of the longitudinal load center of gravity. Indicates the spacing between the left and right supports. Indicates the distance between the front and rear supports; S2025 minimizes both lateral and longitudinal stability, then introduces an attitude-related penalty term to calculate the equivalent attitude stability value. The calculation formula is as follows: ， in, This represents the equivalent attitude stability value. Indicates the roll angle. Indicates pitch angle, Indicates the rate of change of roll angle. Indicates the rate of change of pitch angle. Indicates a penalty item. This represents a function that takes the minimum value. S2026 calculates the attitude stability margin based on the stability safety threshold and the equivalent attitude stability value.

5. The collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces according to claim 3, characterized in that, The method for marking the robot's corresponding passage mode for each critical restricted area based on the passage margin includes: S2031 If any passage margin data is ≤0, then mark the critical restricted area as an impassable section for the robot; S2032 If all passage margin data are greater than 0 and there is at least one passage margin data that does not exceed its corresponding safety margin, then mark the critical restricted area as the restricted passage section of the robot. S2033 If all passage margin data exceed their corresponding safety margin, then mark the critical restricted area as the passable section of the robot.

6. The collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces according to claim 1, characterized in that, The process of determining the travel path of each robot along the key restricted area based on key restricted area and key passage constraint information, combined with the current and target positions of each gantry heavy-duty handling robot, and generating the occupancy order and expected passage sequence for each key restricted area includes: S301 combines the passage patterns of the key restricted areas with the robot's starting point and ending point to filter the candidate spatial passage sequences of each robot, and extracts the key restricted area sequence based on the candidate spatial passage sequences; S302 calculates the passage time for each key restricted area along its key restricted area sequence from the starting point to the end point, forming the expected entry and exit time intervals for the robot in each key restricted area segment; S303 For each critical restricted area, collect all robot records that satisfy the passage of the critical restricted area, sort the collected robot records according to their expected entry time, and the sequence corresponding to the sorting result is the segment occupancy order of the critical restricted area. The set of expected entry / exit time intervals corresponding to the sorting sequence is the expected passage time sequence of the critical restricted area.

7. The collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces according to claim 6, characterized in that, The process of combining the travel patterns of key restricted areas with the robot's start and end points to filter candidate spatial travel sequences for each robot, and extracting key restricted area sequences based on these candidate spatial travel sequences, includes: S3011 discretizes the registered 3D prior model into a topological graph, and maps each key restricted region to the edge of the topological graph after discretization. For each robot, S3012 maps its start and end points onto the topology map, filters key restricted areas marked as impassable sections, performs path search on the filtered topology map, and obtains the candidate spatial passage sequence for each robot. S3013 extracts the key restricted regions on each candidate path in the candidate space travel sequence of each robot, and obtains the key restricted region sequence corresponding to the candidate path.

8. The collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in confined spaces according to claim 6, characterized in that, The step of calculating the passage time for each robot along its key restricted area sequence from the starting point to the ending point, forming the expected entry and exit time intervals for the robot in each key restricted area segment, includes: S3021 For any critical restricted area, calculate the expected passage duration of the critical restricted area based on the critical passage constraint parameters of the critical restricted area and the robot's passage mode in the critical restricted area; S3022 updates the expected entry time of the next critical restricted area based on the transition conditions required for the connection of the topology segment during the topology connection process between adjacent critical restricted areas. Based on the updated expected entry time, it calculates the expected passage duration of the next critical restricted area, thereby obtaining the time interval between the expected entry time and the expected exit time of the robot in each critical restricted area.

9. The collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in a confined space according to claim 1, characterized in that, The process involves conflict detection based on the occupancy order and expected passage sequence of each critical restricted area. Under the condition of satisfying critical passage constraints, and with the objective of minimizing the number of critical restricted areas, the passage strategy for each robot through the critical restricted areas is determined, and the path planning results for each robot are output, including: S401 uses the path selection and entry time offset of all robots in a coordinated manner as individual coding objects. The decision variables in each individual include: the path selection variable of each robot corresponding to the candidate space passage sequence and the entry time offset of each robot in the key restricted area corresponding to its selected path. S402 decodes any individual in the population to obtain the time window in which each robot occupies the key restricted area, and obtains the sequence of key restricted areas selected by each robot after decoding. For each robot, the number of times the key restricted area marked as a restricted passage segment appears in its selected path is counted to form a multi-objective evaluation vector for Pareto comparison. S403 When a decoded individual violates the mutual exclusion constraint in any critical restricted area, adaptive repair is performed on the individual to restore the mutual exclusion feasibility without changing the selected path sequence determined by each robot in the decoded process. S404 performs Pareto non-dominated sorting on the multi-objective evaluation vectors of the repaired feasible individuals to obtain the non-dominated frontier level. Each component of the multi-objective evaluation vector represents the degree of minimization of the occurrence of restricted passage sections in the selected path of each robot. The non-dominated solution is stored in the external archive set. S405 selects the parent robot to generate offspring based on the non-dominated hierarchy and congestion distance, performs cross-variation and mutation of the path selection variables of each robot, cross-variation and mutation of the entry time offset of each robot, performs conflict detection and adaptive repair on the generated offspring, calculates the updated multi-objective evaluation vector, and updates the offspring to enter the next round of iteration. When the iteration reaches the preset upper limit, S406 calculates the relative proximity by sorting the solutions from the non-dominated front obtained by merging the parent and child solutions, and selects the solution with the largest relative proximity as the path planning output.

10. The collaborative dynamic path planning method for a gantry-type heavy-duty handling robot in a confined space according to claim 9, characterized in that, When a decoded individual violates the mutual exclusion constraint in any critical restricted area, adaptive repair is performed on that individual to restore the feasibility of mutual exclusion without changing the selected path sequence already determined by each robot. This includes: S4031 checks each key restricted area to see if two robots occupy overlapping time windows. If a conflict exists, it records the key restricted area of ​​the conflict and the corresponding pair of conflicting robots. Within the same key restricted area, it identifies the leading robot that entered earlier and marks the remaining conflicting parties as lagging robots that need to be advanced. S4032 performs entry time advancement only on the lagging robot in its conflict-critical restricted area, so that its occupation time window does not overlap with the occupation time window of the leading robot, and updates the exit time in the critical restricted area accordingly. S4033 calculates the propulsion amount at the current entry time and updates the entry time by sequentially cascading the propulsion amount according to the key restricted areas subsequently traversed by the lagging robot. S4034 performs conflict detection again on the repaired individual. If a conflict still exists, iterative updates continue until the individual satisfies the mutual exclusion constraint in all critical restricted areas.

Images