Mechanical and electrical installation task topology analysis and intelligent scheduling method based on BIM digital twinning

Abstract

The application discloses a mechanical and electrical installation task topology analysis and intelligent scheduling method based on BIM digital twinning, relates to the technical field of mechanical and electrical installation scheduling, and comprises the following steps: pre-compiling the component sweep interference into a static anti-collision constraint directed graph containing impedance edge weight; when optical anomalies occur, performing one-way denial visual release of the work machine bottom layer to determine the deadlock abnormal node; extracting a local wave and subgraph around the deadlock abnormal node and generating a replacement flow transfer instruction sequence; updating the graph edge impedance according to the execution delay, and outputting the machine posture micro-shake cumulative offset tensor and the safety envelope approximation limit alarm mark when the continuous micro-shake erosion remaining clearance distance is output; and according to the tensor, performing virtual anti-collision envelope expansion and writing back the polarization impedance weight. The method reduces the edge end three-dimensional anti-collision calculation burden, improves the determination reliability in a strong light dust scene, and realizes the early diversion and anti-collision control under the condition of continuous micro-shake.

Description

Technical Field

[0001] This invention relates to the field of electromechanical installation scheduling technology, specifically a method for topology analysis and intelligent scheduling of electromechanical installation tasks based on BIM digital twins. Background Technology

[0002] With the development of prefabricated electromechanical construction and digital construction technologies in complex projects such as intercity railway stations, components such as air ducts, water pipes, cable trays, electromechanical equipment, support components, and ALC lightweight partition walls are increasingly entering the construction site in prefabricated or semi-prefabricated forms. Installation is completed through processes such as hoisting, transportation, positioning, wall mounting, opening, assembly, and placement. However, on-site work often involves multiple disciplines working simultaneously, high requirements for the coordination of wall panel segmentation and numbering, difficulties in controlling the precision of pipe penetrations, and potential spatial and procedural conflicts between installation tasks.

[0003] Current engineering practices typically rely on BIM and related digital twin technologies to establish a model linking components, construction environment, and electromechanical systems. Before construction, intelligent component layout, collision detection, path planning, hoisting simulation, and risk checks are carried out. During construction, real-time on-site monitoring information is used to dynamically adjust task priorities, resource allocation, and schedule, thereby improving construction efficiency, finishing quality, and operational safety.

[0004] Chinese patent document CN115271269A proposes a BIM-based safety management method for the hoisting of large precast components. After establishing a BIM model of the hoisting site for large precast components, tower crane parameter information is acquired, a spatial coordinate system is built, and the transportation space is divided into spatial layers according to environmental factors. Based on this, objective functions for precast component transportation path planning are constructed according to environmental, economic, and safety factors. Then, a hoisting efficiency neural network model is built to calculate the hoisting efficiency score and safety score in the path planning objective function. Finally, the transportation offset of the precast component is obtained based on the optimal solution of the objective function. The data flow relationships of the input layer, preprocessing layer, offset layer, deep learning layer, and output layer are given, where the offset layer is used to calculate the impact of environmental factors in the obstruction space layer on the hoisting path. The core idea of ​​this technology is pre-construction BIM modeling, environmental layering, objective function solving, and construction offset monitoring.

[0005] While the existing technologies described above are suitable for hoisting large precast components, their application in deep-ground electromechanical installation still has certain limitations. Firstly, the analysis basis of existing solutions is the building information model, spatial layering, and objective function established before construction, primarily targeting predetermined paths. This cannot adequately reflect changes in component posture, access boundaries, and relationships with adjacent obstacles during electromechanical installation. Secondly, underground machine rooms, integrated utility tunnels, and equipment mezzanines typically have numerous pipelines, severe obstructions, unstable lighting, and strong interference from welding arc light and dust, making on-site perception prone to inaccuracies. Existing solutions focus on model solving and offset monitoring, making it difficult to effectively support task status and scheduling coordination under complex interference conditions. Thirdly, during the processes of heavy-duty equipment adhering to walls, turning, drilling, and positioning, the operational boundaries are not statically constant due to factors such as off-center loading, cantilever rebound, slight ground slopes, and space constraints. Existing solutions still rely primarily on predetermined paths and static models for decision-making, which can easily lead to local task obstruction or inaccurate judgment of safety clearances.

[0006] Therefore, how to reliably perform task topology analysis and secure scheduling during electromechanical installation processes with limited computing power and strong on-site interference is a technical problem that needs to be solved. Summary of the Invention

[0007] (a) Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a topology analysis and intelligent scheduling method for electromechanical installation tasks based on BIM digital twins. In cases of optical anomalies, the method uses the underlying waveform of the work equipment to unidirectionally reject visual release and identify deadlock anomaly nodes. It extracts local spillover subgraphs around the deadlock anomaly nodes and generates alternative flow instruction sequences. Based on execution hysteresis, it updates the graph edge impedance and outputs the cumulative offset tensor of the equipment's attitude micro-shaking and a safety envelope approaching the limit alarm flag when continuous micro-shaking erodes the remaining clearance. It then performs virtual anti-collision envelope expansion based on the tensor and writes back the polarization impedance weights. This method reduces the computational burden of 3D anti-collision at the edge, improves the reliability of judgment in strong light and dust scenarios, and enables early rerouting and anti-collision control under continuous micro-shaking conditions; thus solving the technical problems described in the background art.

[0009] (II) Technical Solution

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A topology analysis and intelligent scheduling method for electromechanical installation tasks based on BIM digital twins, executed by an edge gateway, includes: pre-storing a static anti-collision constraint directed graph obtained by component sweeping interference dimensionality reduction compilation; triggering visual suspension based on on-site optical data, and identifying the current target node as a deadlock abnormal node when the underlying waveform of the working equipment does not meet the real chain break criterion; extracting a local ripple subgraph around the deadlock abnormal node, overwriting the abnormal edges with impedance masks and generating a replacement flow instruction sequence; issuing the replacement flow instruction sequence and updating the graph edge impedance according to the execution hysteresis; and generating a cumulative offset tensor of the equipment attitude micro-shaking and a safety envelope approaching the limit alarm flag when continuous micro-shaking causes the safety envelope to approach the limit.

[0012] In response to the safety envelope approaching the limit alarm, a virtual anti-collision envelope expansion calculation is performed based on the cumulative offset tensor of the machine's attitude micro-shaking. The expansion result is converted into polarization impedance weights and written back to the edge weights of the static anti-collision constraint directed graph, triggering a rerouting or retreat.

[0013] Furthermore, the static anti-collision constraint directed graph is generated by the edge gateway based on the overlap results of the pre-stored component sweep space domain and the envelope of the adjacent obstacle; the edge gateway defines the segment nodes as lifting position, turning position, wall-hugging transition position, entry position and landing position, and generates directed edges and corresponding edge weight index items based on the continuous action relationship between the segment nodes.

[0014] Furthermore, when the glare excitation index exceeds a preset threshold, the edge gateway suspends the visual release judgment and writes it to the lock window start point; within the lock window, the edge gateway collects the underlying execution waveform of the operating equipment, and when the underlying execution waveform does not meet the true chain break judgment criteria, the current target node is identified as a deadlock abnormal node.

[0015] Furthermore, the edge gateway calls the adjacency list of the static anti-collision constraint directed graph around the deadlock abnormal node, extracts the local spillover subgraph according to the abnormal out-degree and skip-level boundary; and writes the directed edges pointing to the blocked bit segment into the abnormal edge set, and generates a replacement flow instruction sequence after overwriting the edge weights corresponding to the abnormal edge set with impedance mask.

[0016] Furthermore, the edge gateway will split the replacement flow instruction sequence into back segment, lift segment, turn segment and position segment, and update the graph edge impedance according to the start deviation, position deviation and holding deviation of the actual action relative to the planned action; when continuous micro-shaking causes the remaining net distance to decrease, the result of the segment is frozen and the cumulative offset tensor of the machine attitude micro-shaking and the safety envelope approaching the limit alarm mark are generated.

[0017] Furthermore, the edge gateway pre-stores an edge weight index for each directed edge; the edge weight index records at least the current edge identifier, the target edge weight address, and the set of adjacent edge identifiers that share the start or end point with the current edge, as well as the corresponding task number and graph version number, for subsequent writing of polarization impedance weights.

[0018] Furthermore, when generating the alternative flow instruction sequence, the edge gateway simultaneously writes the task number, current edge identifier, deadlock abnormal node identifier, abnormal edge set, alternative flow instruction sequence, and frozen status identifier; and writes the fields to the local task buffer and execution buffer for use by the execution controller and field terminals.

[0019] Furthermore, when the security envelope approaching limit alarm flag is set, the edge gateway outputs the current edge flag, the prohibited edge flag, the alternative flow instruction sequence, and the frozen status to the field terminal; and maintains the prohibited passage display status of the current edge to the field terminal until the security envelope approaching limit alarm flag is released.

[0020] Furthermore, the edge gateway sends an edge weight overwrite instruction frame through the fieldbus master station; the edge weight overwrite instruction frame includes at least the current edge identifier, the target edge weight address, the polarization impedance weight, and the timestamp; and when it receives an overwrite failure response, it maintains the current frozen state and stops the current graph re-solution.

[0021] Furthermore, the edge gateway connects the optical acquisition terminal, waveform acquisition terminal, execution controller, geometric mirror cache, and fieldbus master station; the optical acquisition terminal is used to provide field optical data, the waveform acquisition terminal is used to provide the underlying execution waveform of the work equipment, and the geometric mirror cache is used to provide the original envelope and the envelope of the adjacent obstacles corresponding to the current edge.

[0022] Furthermore, the width of the lock window is 400 milliseconds to 600 milliseconds; the edge gateway only receives the underlying execution waveform of the working machine within the lock window, and excludes the execution waveform outside the lock window from the current visual suspension cycle, and uses the start point of the lock window as the unified time reference for the waveform determination.

[0023] Furthermore, when extracting the local sweep subgraph, the edge gateway limits the upper limit of the out-degree of the immediate successor of the deadlock anomalous node to 8 to 12, and limits the maximum hop level of the candidate node relative to the deadlock anomalous node to two levels; nodes that exceed the boundary are not written into the local sweep subgraph.

[0024] Furthermore, if the estimated time for resolving the map exceeds the preset emergency response window, the edge gateway will stop resolving the map and convert the virtual anti-collision envelope expansion calculation result into a yield displacement command along the outer normal of the dangerous side, send it to the hydraulic servo drive to execute the yield action, and keep the current edge frozen.

[0025] (III) Beneficial Effects

[0026] This invention provides a method for topology analysis and intelligent scheduling of electromechanical installation tasks based on BIM digital twins, which has the following beneficial effects:

[0027] The component sweep interference is first compiled into a static anti-collision constrained directed graph. The initial setting time of the material and the secondary demolition penalty coefficient are written into the impedance edge weights. The edge gateway can call the graph edge constraints of the engineering consequence attributes without reviewing the 3D intersection operation during the construction phase. On-site optical data triggers vision suspension. The visual results are unidirectionally rejected by the underlying execution waveform of the work equipment. The apparent release affected by strong light, dust, and reflection no longer directly enters the scheduling chain. The deadlock anomaly node is stably output and the consistency between the anomaly identification result and the work status is improved. Local ripple subgraphs are extracted around the deadlock anomaly node. The anomaly edges are overwritten with impedance masks to generate an alternative flow instruction sequence, which limits the rearrangement range to the disturbed range and avoids the spread and recalculation of the entire graph due to a single point anomaly.

[0028] The replacement flow instruction sequence is broken down into executable actions. The graph edge impedance is updated according to the execution hysteresis. When continuous micro-shaking erodes the remaining net distance, the cumulative offset tensor of the machine attitude micro-shaking and the safety envelope approaching the limit alarm sign are thrown outward. The dangerous residual is quickly separated from the closed loop at this level. Virtual anti-collision envelope expansion calculation is performed according to the cumulative offset tensor of the machine attitude micro-shaking. The calculation result is converted into polarization impedance weights and written back to the edge weights of the static anti-collision constraint directed graph. The actual physical micro-deformation can reverse change the upper-level scheduling boundary. Finally, a closed loop is formed from interference compilation, anomaly rejection, local rearrangement, residual throwing to edge weight reverse writing, which solves the problems of incomprehensibility, indiscernibility and static boundary desensitization to continuous micro-shaking, and enhances the continuity of task scheduling and anti-collision control at the deep-earth heavy assembly site. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the overall architecture of the intelligent scheduling closed-loop system for electromechanical installation in an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram illustrating the process of compiling the swept space domain and generating the anti-collision constraint graph in an embodiment of the present invention;

[0031] Figure 3 This is a schematic diagram of the unidirectional hard interlock arbitration process under optical distortion conditions in an embodiment of the present invention;

[0032] Figure 4 This is a schematic diagram illustrating the generation of rerouting and alternative flow instructions based on local sweep subgraphs in an embodiment of the present invention.

[0033] Figure 5 This is a schematic diagram of the alternative flow execution, basic correction, and micro-shaking external projection process in an embodiment of the present invention;

[0034] Figure 6 This is a schematic diagram of envelope expansion and polarization impedance reverse overwriting driven by micro-shaking in an embodiment of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Please see Figures 1-6 This invention provides a method for topology analysis and intelligent scheduling of electromechanical installation tasks based on BIM digital twins, including: Step 1 is executed by a cloud-based geometry compilation server, and the results are received by an edge gateway. This step first extracts the three-dimensional space that the electromechanical components will scan during installation, and then compiles the interference relationship with existing entities into a downloadable two-dimensional impedance weight, so that subsequent steps no longer rely on the intersection of continuous curved surfaces and three-dimensional anti-collision rendering at the edge end, but directly call the already solidified static anti-collision base.

[0037] Step 1: Translate the three-dimensional assembly interference that is difficult to withstand from the edge end before construction into a two-dimensional impedance edge weight with irreversible material damage properties, and solidify the edge weight into the anti-collision constraint directed graph as the starting base for subsequent side channel arbitration and local rerouting.

[0038] The dangers of deep-ground heavy assembly sites come not only from the current volume occupied by the components, but also from the future space swept by the components during lifting, turning, perforation, wall contact, and positioning. If this future space is left for on-site calculation by the edge gateway, the edge gateway needs to simultaneously handle the intersection of continuous curved surfaces, changes in hoisting posture, and the crossing relationships of multiple components, which can easily lead to computational blockage when interference is concentrated. Therefore, this implementation method moves the most arduous geometric work forward to before formal construction. The cloud-based geometry compilation server reads the building information model, installation sequence list, hoisting tool movement trajectory, and material and process ledger, first establishing a swept space domain for each component to be installed, and then compiling the interference results into a two-dimensional anti-collision constrained directed graph. The key to this approach is not simply reducing the dimension, but rather incorporating the potential for irreversible damage from crossing into the impedance edge weights, so that subsequent steps receive not ordinary routing costs, but construction boundaries already carrying engineering consequences.

[0039] The surveying terminal first synchronizes the 3D shapes of the components to be installed, existing components, and the enclosure structure to the cloud-based geometry compilation server. The cloud-based geometry compilation server determines the lifting points, turning points, crossing points, and landing points based on the installation sequence table, and then generates a swept spatial domain using these action nodes as the skeleton. Subsequently, the cloud-based geometry compilation server calls the boundary representation solver or an equivalent entity geometry kernel to perform Boolean intersection calculations on the swept spatial domain and the existing entity boundaries to obtain the interference overlap volume. Then, the initial setting time and secondary demolition penalty from the material process ledger are injected into the same edge weight calculation chain to form impedance weights. Finally, the impedance weights are written into the anti-collision constraint directed graph and encapsulated into a JavaScript object representation JSON data packet via the Message Queue Telemetry Transmission Protocol (MQTT) and sent to the edge gateway. The edge gateway no longer holds the 3D entities, but only the nodes, edges, and impedance weights on the edges, thus enabling subsequent actions to unfold around the static graph base.

[0040] In a preferred embodiment, the construction objects include rectangular air ducts, circular water pipes, cable trays, hangers, equipment bases, and concrete enclosures. The rectangular air ducts retain the overall outer contour of the galvanized sheet body, flange thick edges, and outer insulation layer; the circular water pipes retain the combined outer diameter of the copper pipe, clamps, and insulation shell; and the concrete enclosure retains the boundaries of pre-drilled holes, no-drill zones, and post-cast strips. The cloud-based geometry compilation server does not directly perform collision analysis on the static shape, but rather drags the component's outer contour into a swept space domain along the installation action. For example, when hoisting a rectangular air duct through the passage between the cable tray and the sprinkler main in an underground pump room, the installer selects the lifting position, turning position, and lowering position in the modeling terminal. The system then generates the entire swept space domain of the rectangular air duct from lifting to wall contact, based on the relative position of the lifting point and the component's center of gravity. This swept space domain is then subjected to Boolean intersection with the existing cable tray, sprinkler main, and concrete beam side boundaries. Any overlapping portions remaining after the intersection are considered interference overlap volumes.

[0041] To ensure that the intersection results consistently correspond to on-site actions, the cloud-based geometry compilation server performs model cleanup before generating the swept space domain. This involves merging duplicate faces, repairing flipped normals, and closing open edges to prevent thin-walled components from being misidentified as empty shells. Furthermore, it unifies the component coordinate system, the lifting device coordinate system, and the floor reference coordinate system to prevent the same duct from being in place in the model but still undergoing a flipping process on-site. For long cable trays and ducts, the swept space domain retains the overall proportional relationship along the component's long axis. Protruding parts such as flanges, clamps, and lateral lifting lugs are not flattened, as these hard edges are often the first to come into contact with beam sides or existing pipelines on-site.

[0042] For components that need to pass through pre-reserved holes, if there are uneven edges, residual sealing material, or sleeve linings at the edges of the pre-reserved holes, these contours will be included in the intersection object to prevent deviations between the hole positions on the drawing and the hole positions on site from being prematurely smoothed out. Specifically:

[0043]

[0044] Where: interference ratio : Indicates the degree of overlap between the swept space domain of the component to be installed and the boundary of the existing entity, with a value of This is used to unify the interference quantities of components of different sizes into comparable dimensionless results; interference overlap volume. : Represents the overlap volume of entities obtained after Boolean intersection, with a value of It originates from the intersection of the swept spatial domain and existing entities; safety tolerance volume. : Represents the safety buffer volume reserved around the component to be installed along the construction direction, with a value of It is composed of the net distance envelope, the lifting device swing envelope, and the manual standing position envelope;

[0045] To avoid the common pitfall of focusing only on the current position while ignoring the process position, the swept spatial domain is generated using piecewise cubic splines or a hybrid circular-linear interpolation method. This method combines five types of actions—lifting, turning, wall contact, hole insertion, and positioning—into a continuous motion trajectory. For thin-walled components, flanges and stiffening edges are not simplified because these local protrusions are often the first to be scraped or bumped. For components that need to pass through pre-drilled holes or sleeves, the actual trimming state of the pre-drilled holes is preserved to prevent misjudging dangers as passable when the hole positions on the drawings do not match the actual hole positions on site. Therefore, the cloud-based geometry compilation server does not obtain an abstract path, but rather a geometric process volume that directly corresponds to the on-site actions.

[0046] Furthermore, the basic data collected by the edge gateway already includes the fact that the installation process occupies volume, and the obstructed edges that can be seen on site are the same as the physical blockages at the construction site; the rerouting on the subsequent diagram needs to be done in the process space rather than the instantaneous cross section, and the collision when the component turns will not be exposed in advance.

[0047] Only interference ratio This is still insufficient to support construction decisions because the engineering consequences of different materials being penetrated are not the same. Penetrating temporary empty areas has consequences that are not on the same order of magnitude as penetrating newly poured concrete sides, structural adhesive bonding areas, or waterproof sealing areas. This implementation method incorporates the initial setting time of materials in the material process log into the same edge weight calculation chain, making the impedance edge weight no longer just a geometric quantity, but a coupling quantity between geometry and material consequences. The logic behind this approach is that any area that will leave irreversible traces due to curing, sealing, or demolition should be magnified in advance in the static diagram base; otherwise, the edge will only be treated as a slightly more costly ordinary edge, failing to reflect the irreversible nature of engineering risks.

[0048] In a preferred embodiment, the material process log is established with entries based on material codes, including at least the material name, construction stage, initial setting time, patching method, modification method, and repair restrictions. If the area to be avoided corresponds to the side of a newly poured concrete beam, then the initial setting time of the material... Take the initial setting record from the process card for that pouring batch; if the area to be avoided corresponds to a structural adhesive-bonded hanger, then the initial setting time of the material... Take the initial setting record from the process card for the structural adhesive; if the area to be avoided corresponds to a fire-resistant or waterproof sealing layer, then the penalty coefficient for secondary demolition is... It is set to a higher level in the repair rule table because disassembly often involves not only the repositioning of electromechanical components, but also the redoing of the sealing layer.

[0049] In this process, the injection doesn't simply add the material name as a note to the edge; instead, it directly incorporates the material's technological consequences into the edge weight calculation, ensuring that the geometric and technological relationships close on the same edge. Specifically:

[0050]

[0051] Where: impedance weight : Represents the algebraic impedance value written on the edge of the directed graph of the anti-collision constraint, with a value of This is used to guide subsequent steps in preserving, weakening, or masking the relevant directed edges; interference overlap volume : Represents the overlap volume of entities obtained by Boolean intersection, with a value of Safety tolerance volume : Represents the safety buffer volume set around the component to be installed, with a value of It is used to normalize the dimensions of different components to the same comparison scale;

[0052] Natural constant : Represents an exponentially amplified base, with a fixed value, used to transform the differences in consequences caused by the initial setting time of materials into nonlinear amplification; initial setting time of materials : This represents the process time parameter from the completion of construction to the point where the material corresponding to the area to be avoided reaches an irreversible initial setting state. The value is... This information originates from the materials and processes log and is used to amplify the consequences of not being able to penetrate as close to the curing boundary as possible.

[0053] Secondary demolition penalty coefficient : Indicates the degree of additional damage caused by secondary drilling, glue application, dismantling, or repair in this area, with a value of [value missing]. It originates from a repair rule table jointly maintained by structural and mechanical / electrical engineering professionals, and is used to amplify the engineering costs that are difficult to restore after passing through.

[0054] If the adjacent obstacle corresponding to the current edge has a material code written in step one during mapping, then the initial setting time of the material corresponding to that material code is read. Penalty coefficient for secondary demolition The current edge is used in impedance calculations. If the adjacent obstacle corresponding to the current edge is not coded in the material code, the current edge will only use the ratio of volume difference to safety tolerance volume to form the basic impedance, without adding the material amplification term. When an installed electromechanical component is used as an adjacent obstacle, its corresponding material code from the drawing in step one will be used.

[0055] Taking a structural adhesive-bonded hanger passing through a duct as an example, after the installer selects the material code corresponding to the hanger on the terminal, the cloud-based geometry compilation server reads the initial setting time of the structural adhesive from the material process ledger. Then, read the secondary demolition penalty coefficient from the repair rule table. If the swept spatial domain rubs into the outer edge of the adhesive layer, then the impedance weight of the relevant edge at that location... The initial setting time of the material is significantly increased; if the same geometric interference occurs next to a removable temporary support, the initial setting time of the material is significantly increased. Penalty coefficient for secondary demolition The resulting amplification remains at a low level. For wall areas requiring secondary drilling for rewiring, the cloud-based geometry compilation server binds the material code of that area to the repair rule table, thus giving edges crossing that area a naturally higher impedance weight. In this way, the interference overlap volume of the same size Different levels of impedance weights will be compiled under different material backgrounds. .

[0056] Furthermore, the static diagram base naturally falls into two categories: accommodative congestion and irreversible penetration. Therefore, there's no need to further familiarize oneself with material processing techniques or apply impedance weighting. This allows us to determine whether relevant edges should be bypassed; after the material and process consequences are written into the edge diagram, the construction path is not only sensitive to geometric collisions, but also to the consequences of solidification, sealing, and demolition projects.

[0057] When impedance weight After the calculation is complete, the cloud-based geometry compilation server writes each directed edge into the anti-collision constraint directed graph, and then encapsulates it into a JSON data packet and sends it to the edge gateway. Optimized data fields include source node identifier, target node identifier, material code, and impedance weight. The edge gateway receives the map, which includes sweep category, retransmission flag, and CRC checksum field. Upon receiving the map, the edge gateway first verifies the map source based on the version number, then performs an integrity check based on the CRC checksum field. If the current message verification fails, the previous version of the fixed map is retained, and valid edge weights are not overwritten with incomplete messages. After the edge gateway completes the fixing process, it displays node relationships and prohibited edges to the construction terminal, rather than a three-dimensional entity.

[0058] Specifically: In the node tree on the handheld terminal, some connecting edges are marked as prohibited from crossing, while others remain open. The installation team adjusts the lifting point sequence and turning actions accordingly. Taking the scenario of installing cable trays before air ducts in an underground utility tunnel as an example, the edge gateway does not display a 3D animation, but rather a static directed relationship where the device moves from node A to node B, then around node C and lands at node D. The team leader can then understand which edges are blocked and which are retained by comparing the cable trays, beam sides, and reserved holes seen on site.

[0059] In parallel alternative paths, if the project side does not use a continuous surface solver but a discrete point cloud compilation path, the cloud-based geometry compilation server converts the component shape into a voxel grid, then organizes the space-occupying units in an octree manner, accumulates the overlap between the swept space domain of the component to be installed and the existing solid-occupying units, and finally outputs the impedance weights under the same terminology system. In this alternative approach, the change lies in determining the interference overlap volume. The geometric implementation method does not change the directed graph of anti-collision constraints or the initial setting time of the material. and secondary demolition penalty coefficient The call chain is such that the basis of rights is not locked by a single geometric kernel.

[0060] In the degradation path, if the cloud cluster becomes congested due to pre-construction compilation tasks, the system extracts the bounding box projections of the components to be installed and existing entities on the horizontal plane, first calculates the planar overlap area, and then writes the result as a minimum impedance edge. At this time, the vertical height information is hidden, and the edge gateway can still obtain a simplified anti-collision constraint directed graph, ensuring that the construction team will not blindly move around without a graph base. For work areas with short-term network disconnection, the edge gateway will also write the most recent valid graph to the local flash memory. After the link is restored, only the edge weight differences are synchronized, and the entire graph is not retransmitted.

[0061] Furthermore, the main path, alternative path, and degradation path all share the same output object: impedance weight. With the anti-collision constraint directed graph, the subsequent call interface is fixed; after the object is compressed into a data packet with clearly defined nodes, edges and fields, the edge side only performs receiving and fixing actions, avoiding the need for three-dimensional geometric calculations to return to the field equipment.

[0062] Step 2: When optical appearance is distorted, the physical execution waveform of the underlying equipment is used to implement one-way hard interlock arbitration for visual judgment, thereby clearing visual misjudgments and delivering a reliable deadlock anomaly node identifier to the subsequent local rerouting steps. It is executed collaboratively by the edge gateway, optical acquisition terminal, and waveform acquisition terminal.

[0063] Optical contamination at deep-ground electromechanical installation sites is characterized by its suddenness, localization, and strong reflectivity. Welding arc light, cutting sparks, dust re-spreading, and specular reflection from galvanized steel sheets can quickly transform the originally continuous edges of components in an image into large, bright, connected areas. Although the optical acquisition terminal continues to capture images, the open channels and displaced components seen on the screen are often just blank illusions obscured by the intense light. If the edge gateway directly accepts these visual results, the impedance weights solidified in step one... This would be mistakenly interpreted as the situation having been resolved, and subsequent hoisting orders would be issued prematurely. The first thing that would happen on site would not be a smooth path, but rather a situation where the components remained in place while subsequent components continued to approach, creating a pressure situation.

[0064] This implementation first acknowledges that the optical image has structural distortion in a strong light environment, and then uses the execution waveform located inside the hydraulic circuit, pneumatic circuit or servo circuit to make the final decision.

[0065] Because pressure curves, valve position status, servo current, and relay disconnect pulses are located on the internal energy chain of the equipment and do not share the same contamination path as arc light and dust, they are more suitable for determining whether unloading, clamping, positioning, or detachment has been truly completed. At this stage, the edge gateway invokes the source node identifier, target node identifier, and impedance weight from step one. It is only used to lock the local edge that is currently under construction, preventing unrelated camera images and unrelated machine waveforms from entering the arbitration chain.

[0066] The edge gateway first receives continuous frames from the optical acquisition terminal and extracts a projection window of the target component based on the target node identifier; then, the edge gateway calculates the glare abrupt change index for the projection window. Once the glare abrupt change index Exceeding the glare threshold Immediately suspend visual clearance judgment and simultaneously unlock the window width. A defined waveform sampling window for the machinery is established; within this window, the waveform acquisition terminal reads the vacuum pressure register, hydraulic valve position register, limit switch register, or servo drive status register from the Modbus TCP register transfer protocol, and sends the verified timing waveform back to the edge gateway; the edge gateway then calculates the amount of evidence of mechanical chain breakage. When the amount of evidence of mechanical chain breakage The evidence threshold was not met. At this time, the visual result is cleared, and the current target node identifier is written as the deadlock exception node identifier. When the evidence of mechanical chain breakage Reaching the evidence threshold When the visual suspension is lifted, the original impedance weights are maintained on the front side. Then proceed to the next action. The entire chain, from input to output, revolves around the graph object that has been solidified in step one, so no new node naming system will be introduced during the arbitration phase.

[0067] Preferably, the optical acquisition terminal adopts a combination structure of a global shutter industrial camera, a fixed-focus lens, and a quartz protective window. The outer shell is made of aluminum alloy sealed housing and is installed on the side wall of the hoisting channel or on the equipment column. The optical axis of the lens forms an intersecting angle with the main movement direction of the component, so that the main outline of the duct, cable tray, or equipment housing remains visible when entering the corresponding area of ​​the target node mark. For components with strong specular reflection, such as galvanized sheet ducts, stainless steel housings, and aluminum alloy cable trays, the edge gateway reads the material code in step one, calls the corresponding reflection template, and first corrects the outline continuity coefficient. Regenerate glare exponential index .

[0068] When an arc of the same intensity falls on a black-coated shell and a galvanized sheet surface, the degree of contour collapse in the image differs. If material coding is not introduced for interpretation, the suspension condition will be triggered indiscriminately. Specifically:

[0069]

[0070] Where: Glare abrupt change index : Represents the overall intensity of optical representation collapse within the current projection window, with a value of Used to determine whether to suspend visual clearance judgment; Laplace high-frequency energy : Represents the sum of highlight edge energy extracted by the Laplacian convolution kernel, with a value of Overexposure connectivity ratio : Represents the ratio of the area of ​​the connected region where the pixel value reaches the saturation limit to the area of ​​the target projection window, with a value of Profile continuity factor : Indicates the degree to which the contour of the target component template is continuously preserved in the current image, with values ​​ranging from 0 to 10. Used to compensate for situations where the brightness increases but the component outline remains intact; stability constant : Represents a positive constant to prevent the denominator from reaching zero, and takes the value of Used to maintain the glare exponential index The values ​​are stable;

[0071] Glare threshold During the pre-construction calibration phase, the edge gateway is written, preferably divided into three groups of thresholds according to the work area: welding area, cutting area, and handling area; window width. The preferred time is 400 to 600 ms, and more preferably 500 ms. This is because if the window is shorter than this range, the unloading valve has not yet started to move, and if the window is longer than this range, subsequent personnel movement and secondary reflections will mix the current arbitration with the next action.

[0072] To avoid parallel suspension of standards in step two, this implementation method uniformly uses the glare abrupt change index. As an execution criterion during the construction phase, the temporal variation of the Laplace high-frequency pixel variance is used to calibrate the glare threshold before construction. It is no longer used as a correlation with the glare abrupt change index. Parallel independent construction triggering conditions. Specifically, during the calibration phase, the edge gateway collects three types of images: high-intensity welding, cutting sparks, and normal handling. It calculates the temporal variation of the Laplacian high-frequency pixel variance and selects sample intervals where overexposure connectivity expansion and target component outline fragmentation occur simultaneously to determine the glare threshold. Once construction begins, only when the glare abrupt change index... Exceeding the glare threshold Visual clearance determination is initiated at any time.

[0073] The image preprocessing workflow of the edge gateway is preferably deployed on an industrial Linux or equivalent field operating system. First, distortion correction is performed. Then, a fixed viewport is cropped based on the target node identifier. Following this, Laplacian convolution, saturated connected component labeling, and template contour tracking are performed. The entire process outputs only the glare excitation index. No new geometric boundaries are generated in step two. During on-site implementation, if a worker is transferring a rectangular air duct in an underground pump room and a welder suddenly starts an arc, a bright white area will appear on the monitoring screen, and the edge line of the air duct will be erased from the screen. At this time, the edge gateway does not recognize that the channel has been cleared, but immediately freezes the visual clearance status of the edge according to the target node identifier, waiting for subsequent evidence from the equipment side.

[0074] Specifically, material coding is involved in the glare abrupt change index. The generation process resulted in different suspension thresholds for the specular and non-spectral reflection zones; visual suspension interrupted the misrelease chain the moment glare appeared, and the impedance weighting in step one... Therefore, the apparent blankness will not be mistaken for a resolution.

[0075] Upon visual suspension triggering, the edge gateway immediately sends the window lock start time to the waveform acquisition terminal and binds the corresponding device according to the current target node identifier. If the current device is a vacuum lifting device, the waveform acquisition terminal reads the vacuum pressure register and unloading valve position register; if the current device is a hydraulic clamp, it reads the clamp pressure register, return valve position register, and limit switch register; if the current device is a winch servo lifting device, it reads the servo driver current register and brake release register.

[0076] The register reads the message frame header and writes the source node identifier, destination node identifier, and timestamp. The frame tail writes the cyclic redundancy check (CRC) field. If two consecutive frames fail CRC checks, the edge gateway uses the valid value from the previous frame and simultaneously sets the communication anomaly flag, preventing missing waveforms from participating in the decision-making process. Specifically:

[0077]

[0078] Where: Evidence of mechanical chain breakage : Represents the actual unloading evidence formed by the pressure cliff and valve opening within the lock window width, with a value of Used to determine whether the equipment has physically disengaged; pressure curve : Indicates the time of the machine's execution circuit. The pressure value is taken as follows: ; Window lock start point : Indicates the start time when glare suspension is triggered, and the value is a positive timestamp; Lock window width : Indicates the duration of the effective sampling window for the machine waveform, with a value of Valve position function : Indicates time The valve position or release status is 0 or 1, where 1 indicates that the release channel is open and 0 indicates that the release channel is not open.

[0079] Evidence threshold The no-load calibration waveform, half-load calibration waveform, and full-load calibration waveform before construction are collectively written into the edge gateway. During arbitration, only when the evidence of mechanical link failure is sufficient... Not lower than the evidence threshold Only when the edge gateway is the visual suspension lifted; if the mechanical link is broken, the evidence is insufficient. Below the evidence threshold If the edge gateway fails to do so, it will veto the current visual status (already out of position, completed, or allowed to proceed) and reset the current target node identifier to a deadlock exception node identifier. .

[0080] For example, when moving a modular air conditioner casing on the equipment floor, the monitoring screen was illuminated by cutting sparks, and the edge of the casing suddenly disappeared, but the vacuum pressure curve of the vacuum lift remained unchanged. Still maintain adsorption state, unloading valve position function Keep closed, mechanical chain breakage evidence Unable to reach the evidence threshold The edge gateway then maintains the shell placeholder and outputs the corresponding target node identifier as a deadlock exception node identifier. .

[0081] Specifically, the basis for the ruling has shifted from apparent brightness to the internal energy chain of the equipment. Even if a visual false positive enters the suspended window, it cannot bypass physical unloading evidence to be allowed passage; deadlock anomaly node identification. Since the edges are directly derived from the current graph edges, object matching is not required again in subsequent step three.

[0082] In the main implementation path, the edge gateway identifies deadlock-prone nodes. Current target node identifier, communication anomaly identifier, and original impedance weight. Write the data back to the local task buffer, and then notify step three to extract the local subgraph using the same task number. Since step two does not rewrite the impedance weights... Only the access status of edges and abnormal node objects are rewritten, so the graph boundaries established in step one remain unchanged. Step three can directly block abnormal edges, extract locally related edges, and initiate local rerouting within the same naming system. For the construction team, the results of the actions seen on-site are clear: the hoisted component is not judged as having left its position by the system simply because the screen turns white for a moment, the crane will not continue to advance, and the path connected to the current target node identifier on the handheld terminal remains in a red prohibited state until the waveform adjudication releases the suspension or step three provides an alternative flow sequence.

[0083] In parallel expansion paths, if the construction area does not use vacuum lifting tools or hydraulic clamps, but instead employs winch servo lifting tools and manual support, the edge gateway will switch the window sampling object to the ultra-wideband microwave dwell waveform and servo driver current waveform of the worker's safety helmet. Specifically, after the glare is lifted, the edge gateway reads the dwell change of the safety helmet tag near the target node marker and simultaneously reads the current spike curve of the winch servo driver. Only when both the person is still in the danger zone and the servo is still in a taut state are the visual clearance reset and a deadlock abnormal node marker output. After this replacement, the suspension logic, window locking logic, and one-way veto logic in the main chain remain unchanged; only the pressure evidence is replaced with combined evidence of location residency and current.

[0084] In the degradation implementation path, if the wireless link is blocked by steel structure or suppressed by electromagnetic interference, the waveform acquisition terminal reverts to the hard-wired relay dry contact reading path. The edge gateway only receives the disconnect and reset pulses of the mechanical relay. (If within the lock window width...) If no disconnect pulse is received, the visual zeroing result will be used and a deadlock exception node identifier will be output. If a disconnection pulse is received, the passage status is restored by combining the target node identifier with the corresponding task stage. Although this degradation path reduces waveform details, it still retains the structure of first suspecting visual evidence and then looking for physical evidence of the equipment, and does not degenerate step two into a judgment that simply relies on the image.

[0085] Specifically, the main path, parallel paths, and degraded paths share the same output object, namely the deadlock exception node identifier. With the task side passage status, after the arbitration result is written back, the path status displayed on the construction terminal is consistent with the actual movement of the equipment. The team does not need to understand the waveform details to execute the stop according to the prohibited side given by the system.

[0086] Step 3: Based on the deadlock anomaly node identifier A local affected subgraph constrained by physical space limits is established within the anti-collision constrained directed graph, and the abnormal edge shielding and replacement flow instruction sequence is generated within this local affected subgraph to block the network-wide recalculation storm.

[0087] Step two has already cleared the visual false positives and compressed the truly unacceptable locations into deadlock anomaly node markers. However, if the edge gateway subsequently performs a full-graph rerouting of the entire anti-collision constraint directed graph, surrounding processes and hoisting positions within the pipe gallery will be simultaneously pulled into the search range. What was originally a single blocked node will eventually propagate throughout the entire graph, resulting in a chain of recalculations. Especially in areas where ducts, cable trays, sprinkler mains, and equipment bases are densely intersecting, a blocked node often has multiple subsequent actions attached to it. If the propagation boundary is not cut off first, any search will involve unrelated installation segments. Even if the edge gateway no longer processes 3D geometry, queue accumulation and path flipping will still occur at the graph theory level.

[0088] Specifically, the maximum number of hard contact relationships that physical components can form within a confined installation space is directly expressed as the out-degree boundary of the anti-collision constraint directed graph. This is because in narrow pipe galleries, equipment mezzanines, and shaft corners, the number of objects that can actually form hard obstruction relationships around a single assembly section is capped by the combined radii of walls, beams, installed components, and the slewing radius of the lifting equipment. After solidifying this capped amount as the out-degree boundary in step three, the edge gateway only needs to focus on the deadlock anomaly node identifier. By processing the local affected subgraph, the alternative flow instruction sequence is thus kept within the scope of implementation.

[0089] The edge gateway first reads the task number and deadlock exception node identifier from the task buffer written back in step two. With the original impedance weight Subsequently, deadlock anomaly nodes were identified in the anti-collision constraint directed graph of the local flash memory image. By analyzing the adjacency list and counting the number of its immediate successor edges, we can obtain the out-degree of abnormal conditions. The edge gateway then forwards the abnormal outgoing routes. With lower limit constant and upper limit constant Compare, when abnormal out-of-degree When the value falls within a preset closed interval, a local swept subgraph is extracted and further processing continues. When the out-degree is abnormal... Greater than the upper limit constant When this occurs, it indicates that the physical obstruction relationships around the current node have exceeded the preset local spillover range, and the edge gateway outputs a shutdown hold flag to step four; when an abnormal out-of-degree occurs... Less than the lower limit constant In such cases, instead of shutting down the machine directly, the actual abnormal output should be considered. As the outer width of the local ripple subgraph, and the upper limit of the jump layer. Fixed at 1; when abnormal out-degree When the data falls within a closed interval, a local swept sub-map is extracted according to the standard pattern.

[0090] Subsequently, the edge gateway identifies the deadlock-abnormal node. Directly related edges are written as masked edges, and then the local rerouting cost is calculated for the candidate edges in the local affected subgraph. Finally, the replacement flow instruction sequence is output.

[0091] Preferably, a node in the anti-collision constraint directed graph is not an abstract coordinate point, but a combination of assembly processes and spatial segments. Taking rectangular duct installation as an example, a node corresponds to a specific segment among the lifting position, rotation position, beam-attaching transition position, or placement position; an edge corresponds to the necessary action relationship from the current segment to the next segment. The edge gateway maintains a compressed row storage adjacency list of this graph on the industrial control operating system. After being awakened in step three, it first locates the deadlock abnormal node identifier in the hot zone index. Then, count the number of immediately following edges of the node to obtain the out-degree of the abnormal node. .

[0092]

[0093] In the formula: local ripple subgraph : Indicates the identifier of the deadlock anomaly node. The extracted local node set, which is a finite set of nodes, is used to limit subsequent mask overwriting and rerouting searches; deadlock exception node identifier. : Represents the abnormal node object output from step two, with values ​​being the identifiers of legal nodes in the directed graph with collision avoidance constraints; the set of adjacent nodes. : Indicates the identifier of the deadlock exception node. The set of nodes directly connected by directed edges takes the value of a finite set of nodes;

[0094] Abnormal out-degree : Indicates the deadlock exception node identifier The number of immediately following associated edges, taking a positive integer value; lower bound constant. : Indicates the lower bound of the allowed out-degree of local extraction, preferably 8; upper bound constant. : Indicates the upper bound of the allowed out-degree of local extraction, with a preferred value of 12; Current skip layer : Indicates candidate node deadlock exception node identifier The topological skip layer, with a value of a non-negative integer, is used to limit the outward expansion depth of the local affected subgraph; the upper limit of the skip layer. : Represents the maximum allowed topological skipping level in the local affected subgraph, preferably 2, and alternatively 3, to prevent local extraction from skipping the current construction segment; Candidate node : Represents a single node object in the set of adjacent nodes, with values ​​representing legal nodes in the directed graph with anti-collision constraints;

[0095] Lower limit constant With upper limit constant The settings are not based purely on software experience, but rather determined according to the spatial contact density obtained during the pre-construction drawing review. For the main work area, the number of individual assembly sections that simultaneously have direct hard obstruction relationships with walls, beam bottoms, installed air ducts, cable trays, sprinkler mains, hangers, and lifting equipment envelopes is capped by spatial geometric limits. If abnormal deviations occur... Still above the upper limit constant This indicates that the area around the node is no longer locally congested, but rather that there is a problem with the model mapping, component numbering, or the site space being overwhelmed by unplanned components. In this case, the edge gateway will directly output a shutdown hold flag to prevent the solver from continuing to expand.

[0096] During on-site implementation, in a section of an underground utility tunnel, a rectangular duct was preparing to move from the beam-mounted transition point to its final position when the gap between the cable tray and the main sprinkler pipe suddenly became impassable. Therefore, step two marked this beam-mounted transition point as a deadlock abnormal node. Step three, after reading the node, reveals that it has nine immediately following edges, indicating an abnormal out-degree. Since it falls within a closed interval, only a localized sweep sub-map of two skip layers is extracted outward from that node. The construction worker does not see the entire floor plan on the terminal. Figure 1 The screen only flashes, and a small cluster of nodes near the current duct placement channel is highlighted.

[0097] Specifically, the anomaly propagation boundary is determined by the entity dense packing limit rather than the software timeout parameter; therefore, the local affected submap closely resembles the actual situation; anomaly out-degree When the physical limit is exceeded, outward expansion automatically stops to prevent erroneous graph models from altering the entire graph. After the local affected subgraph is generated, the edge gateway does not delete the original edges, but instead marks the deadlock-related abnormal nodes. The relevant edges pointing to the blocked bit segments are overwritten with impedance masks.

[0098] Among them, the impedance weight solidified in step one Material and spatial consequences are still retained. If the original edge is deleted directly, steps four and five cannot trace the underlying risks of that edge. Therefore, step three adopts the method of retaining the original edge and increasing the cost: the identity of the edge remains unchanged, but its passage cost is increased to a range that the local solver will not select, while the mapping relationship between the edge and the component, process, and material codes is preserved.

[0099]

[0100] Where: cost of partial diversion : Indicates candidate path The total cost within the local ripple subgraph is taken as a value. Used to select alternative flow instruction sequences; candidate paths : Represents a system consisting of several directed edges The resulting local travel scheme takes the value of a finite edge sequence;

[0101] Candidate edges : Indicates candidate path A single directed edge in the graph is defined as a valid edge in the directed graph with collision avoidance constraints; edge impedance weight. : indicates a candidate edge The impedance weight solidified in step one is taken as a value. ;

[0102] Mask coefficient : indicates a candidate edge The mask amplifies the values, assigning finite positive values ​​to normal edges and high-resistance isolation values ​​to abnormal edges, thus excluding blocked edges from the local solver; the mask coefficients... A binary hierarchical method is used for assignment: when the current edge is not in the abnormal edge set, Set to 1; when the current edge belongs to the abnormal edge set, Take the preset barrier value Local density coefficient By statistical candidate edges The number of occupied entities within the current segment whose distance from the outer envelope of the current component is less than a preset net distance threshold is obtained; the occupied entities include installed air ducts, cable trays, sprinkler mains, hangers, temporary supports, and the slewing envelope of the lifting equipment.

[0103] Local density coefficient : indicates a candidate edge The degree of crowding of the components surrounding the location segment is set to a value. This originates from the occupancy count of adjacent entities in the local affected subgraph; local density coefficient. By statistical candidate edges The number of occupied entities within the current segment whose distance from the outer envelope of the current component is less than a preset net distance threshold is obtained; the occupied entities include installed air ducts, cable trays, sprinkler mains, hangers, temporary supports, and the slewing envelope of the lifting equipment.

[0104] Steering penalty : indicates a candidate edge The required additional load for lifting gear turning, component overturning, or wall-hugging actions is taken as follows: Steering penalty Based on candidate edges The determination is based on whether the corresponding action segment contains turning, rolling, and wall-hugging actions; candidate edges that only contain backward or straight-line lifting actions are given a lower turning penalty, while candidate edges that contain both turning and wall-hugging actions are given a higher turning penalty.

[0105] Mask coefficient The write operation is performed by the edge gateway, preferably by directly overwriting the edge weight cache in the hot zone index using memory-mapped pages, and then asynchronously flushing the modified result back to the local flash image; this avoids damaging the original image version and ensures that the task sees the increased cost. Local density coefficient The entity occupancy count from the current local affected subgraph includes installed ducts, cable trays, sprinkler mains, hangers, temporary supports, and the rotation envelope of hangers. Steering penalty. These are derived from the process motion library. Any edge that requires lifting before turning over, retraction before sticking to the wall, or where workers need to reset the guide rope positions is assigned a higher turning penalty. The edge gateway then invokes a masked bidirectional heuristic solver, preferably a masked bidirectional A* searcher, or alternatively a two-ended Dijkstra searcher, both of which use the above equation as a unified cost function.

[0106] During on-site implementation, in the aforementioned rectangular duct scenario, the edge leading directly to the landing position was assigned a high-resistance isolation value. The edge gateway then compared two alternative paths: one was to first retreat to the rotation position and then bypass the outside of the cable tray to enter the landing position, and the other was to continue along the beam and then pass under the main sprinkler pipe. Due to the local density coefficient of the segment where the latter path is located... Higher, and requires significant ductwork relocation, resulting in higher costs for localized diversions. The duct was raised; the solver ultimately selected the previous path. The construction worker observed on-site that the ductwork retreated one section before turning around along the outside of the cable tray, the queue not spreading towards the floors.

[0107] Specifically, the anomalous edges do not disappear; they are masked and isolated while preserving the original mapping relationship. The source of the anomalous edge risk can still be identified after step four; local density coefficients and steering penalty After all costs are considered, the cost no longer depends on the distance on the graph, but on the on-site actions. The masked bidirectional heuristic solver selects the local rerouting cost. After finding the smallest candidate path, the edge gateway does not output it directly as an abstract node chain, but expands it into a sequence of alternative flow instructions.

[0108] Preferably, an edge is expanded into one or more on-site action segments, including retreating, raising, turning, wall-hugging, positioning, and stopping for confirmation. Each action segment is bound to a target segment, execution sequence number, allowed lifting device posture, and reserved manual standing position boundary. The alternative flow instruction sequence is written to the execution buffer of step four through a shared memory queue. The message fields include task number, deadlock exception node identifier, exception edge mask list, rerouting node list, stop-and-hold identifier, and version number. If step four finds that the version number has changed during reading, the old queue is cleared first, and then the new version instruction is retrieved to avoid mixing old and new paths.

[0109] In the enhanced combined implementation, the edge gateway handles abnormal out-of-degree connections. The state where the local affected subgraph falls within the closed interval and has been successfully extracted is written into a hardware holding register. This holding register is only released after the complete replacement flow instruction sequence has been generated. In this way, the lifting equipment, guide rope, and worker positions remain in a safe posture until the solution is complete, preventing disjointed actions such as the equipment continuing to advance before the edges are calculated. In the parallel replacement implementation, if there is sufficient computing power on site, the outer boundary of the local affected subgraph is expanded except for the upper limit of the jump layer. In addition, physical radius constraints can be superimposed. The edge gateway pre-stores the centerline coordinates of the bit segments on the nodes and simultaneously checks the physical radius. In this case, skip-layer filtering is responsible for limiting topology spread, while physical radius filtering is responsible for limiting cross-regional erroneous connections. When both filtering conditions coexist, the deadlock abnormal node is still identified. Centered on.

[0110] In the degradation-and-hold implementation, if the edge gateway cannot complete the masked bidirectional heuristic solution due to temperature rise or task congestion, it directly reads the abnormal node jump dictionary from the flash memory. The dictionary is written by the cloud geometry compilation server before construction according to common bottlenecks, and the fields are divided into deadlock abnormal node identifiers. Replacement node chains and shutdown retention identifiers.

[0111] Specifically, step three does not incur local rerouting costs under this path. The calculation only outputs a fixed sequence of alternative flow instructions corresponding to the dictionary output. If the dictionary does not have a corresponding entry, only a stop-and-hold flag is output, and the terminal no longer refreshes multiple candidate paths, directly displaying a path that reverts to the previous sequence segment hold or stops and waits for manual review. Step three delivers an executable action sequence rather than abstract graph nodes to step four, and the execution side does not perform secondary translation. Enhanced paths use holding registers to seal tool drift during the solution process, parallel paths use physical radii to reinforce the jump boundary, and degraded paths maintain the integrity of the action chain even when they cannot be solved, preventing the system from degenerating into disordered waiting.

[0112] Step 4: Convert the alternative flow instruction sequence generated in Step 3 into field actions, and complete the impedance weighting during execution. The basic correction involves externalizing the continuous micro-shaking residuals that cannot be absorbed by the static base into the cumulative offset tensor of the machine attitude micro-shaking. Safety envelope approaching limit warning indicator It is jointly executed by the edge gateway, execution controller, attitude acquisition terminal, and execution feedback terminal.

[0113] Step three has provided an alternative sequence of instructions: retreat, then lift, then turn, then stick to the wall. However, the fact that the path on the diagram is valid does not mean that the actions on site have been reliably implemented. When heavy-duty assembly equipment executes the alternative sequence of instructions, it is affected by the combined effects of slight ground slope, boom rebound, hydraulic oil lag, guide rope pull, and component eccentricity. Therefore, the same instruction may manifest as a delayed start, overshoot during rotation, wall-sticking sway, and short-cycle micro-wobbling on site. If step four only allows these residuals to be absorbed by the smoothing algorithm, although the static diagram shows surface convergence, the actual safety clearance will be gradually eroded by cantilever fatigue and continuous swaying.

[0114] First, it is responsible for converting the alternative flow instruction sequence into an executable message, and adjusting the impedance weights based on execution hysteresis. Make basic corrections to enable the static graph base to learn the actual execution inertia of the field equipment. Next, focus specifically on the continuous micro-shaking that the static graph base cannot express. When the attitude acquisition terminal detects that the micro-shaking is unidirectionally accumulating and the envelope clearance is continuously narrowing, the edge gateway no longer continues to pursue internal smoothness, but actively removes the residual from the closed loop and suspends the output, allowing step five to take over.

[0115] The edge gateway first reads the alternative flow instruction sequence, abnormal edge mask, and task number from the execution buffer in step three, and then breaks down the alternative flow instruction sequence into retraction segments, lifting segments, rotation segments, wall-hugging segments, and positioning segments. Subsequently, the edge gateway sends each segment to the execution controller via the fieldbus, while simultaneously collecting execution feedback from the stroke encoder, hydraulic pressure sensor, clamping status switch, and chassis gyroscope. Finally, the edge gateway generates an execution hysteresis normalization value based on the hysteresis between the actual action and the planned action. And use it to correct the impedance weight of the current side. If the attitude acquisition terminal simultaneously detects that the integrals of pitch and roll are continuously accumulating, and the remaining net distance between the component's outer envelope and adjacent entities is narrowing to one side, the edge gateway will immediately freeze the impedance weighting. Further smoothing involves packaging dangerous residuals into a cumulative offset tensor of machine attitude micro-shaking. Safety envelope approaching limit warning indicator Output to step five.

[0116] Preferably, the execution entity adopts a combined structure of an assembly tool with a hydraulic telescopic arm, a clamping actuator, and an edge gateway. A three-axis gyroscope is mounted on the chassis of the assembly tool, an angle encoder is installed inside the telescopic arm joint, a clamping switch and a pressure sensor are installed inside the clamping actuator, and passive reflective markers or passive position tags are installed at the ends of the components. The edge gateway runs on an industrial Linux or equivalent operating system, uses a shared memory queue to receive the alternative flow instruction sequence output in step three, and then releases action messages to the execution controller via EtherCAT or an equivalent fieldbus. The action message preferably includes fields such as task number, current edge identifier, action segment number, target position segment, allowable arm length range, allowable rotation angle range, and hold duration. Each time the execution controller completes an action segment, it sends the stroke encoder value, pressure value, and clamping status back to the edge gateway. Based on this, the edge gateway determines three states: action has started, action has reached its position, and action has stabilized.

[0117] Taking the handling of a modular air conditioning unit segment at the equipment layer as an example, the alternative transfer instruction sequence given in step three requires the assembly tool to first retract one segment, then raise the clamping end, then rotate around the outside of the cable tray, and finally be placed against the wall. The edge gateway breaks this sequence into four action segments and sends them to the execution controller in sequence.

[0118] The process involves several steps: first, the machine retracts a short distance, moving the box section away from its previous obstructed position; then, the telescopic boom rises, and the bottom edge of the box section passes over the temporary support in front; subsequently, the boom rotates outward, and the box section bypasses the cable tray; finally, the clamping end slowly approaches the landing surface. After each action segment is completed, the edge gateway compares the planned start time with the actual start time, and the planned landing position with the actual landing position, and combines these differences into a normalized execution hysteresis value. Specifically:

[0119]

[0120] Where: Update impedance weight : Indicates the impedance value of the current edge written back to the static graph base after this execution, with a value of This is used to ensure that subsequent similar edges reflect the actual execution inertia of the field equipment; impedance weighting : This represents the current foundation impedance value that has been solidified in step one and reused from previous steps, and its value is [value missing]. ;

[0121] Smoothing coefficient : Indicates the proportion of historical impedance weight in this basic adjustment, with a value of Pre-store motion segments in the edge gateway according to their categories; when a motion segment includes lifting and turning linkages, a higher value is used to avoid a sudden delay from completely overwriting historical edge weights; smoothing coefficient Action segments are pre-stored in the parameter table of the edge gateway according to their categories. Action segments that include rotation and wall-hugging linkage have a higher historical retention rate, while action segments that only include straight backward or straight upward movement have a lower historical retention rate.

[0122] Execution delay normalization quantity : Represents the current execution hysteresis amount synthesized by startup hysteresis, arrival hysteresis, and hold hysteresis, with a value of This comes from the actual action data of the execution controller; execution hysteresis normalization. This is obtained by normalizing the differences between the planned start time and the actual start time, the planned position and the actual position, and the planned duration and the actual duration in the current action segment. Where:

[0123]

[0124] in, This represents the execution hysteresis normalization, used to convert the time, position, and hold deviation of the current action segment into a single hysteresis value. , , These represent the weighting coefficients for start-up hysteresis, arrival deviation, and hold-up hysteresis, respectively. These three are preset in the edge gateway parameter table according to action segment category and satisfy the following conditions: ; Indicates the actual start time of the action segment. Indicates the planned start time of the action sequence; Indicates the planned startup window length; This represents the vector indicating the actual end position of the action segment. This represents the planned end position vector of an action segment; Indicates the planned path length of the current action segment; Indicates the actual duration of retention. Indicates the planned duration. This indicates the planned window length. The aforementioned time parameters are obtained from the timestamps returned by the execution controller, and the position parameters are obtained jointly from the travel encoder and the component end position label.

[0125] In the enhanced combined implementation, the edge gateway is the smoothing coefficient. Overlaying dynamic adaptive forgetting gating. If the execution feedback indicates that the pause was caused by the worker readjusting the guide rope or debris underfoot, resulting in a delay in the machine's power chain, then the edge gateway maintains a high smoothness coefficient. A single pause is used as a bypass marker, and the impedance weights are not rewritten or updated. .

[0126] Specifically, after the replacement flow instruction sequence is broken down into executable segments, the execution controller can execute segment by segment; the execution hysteresis normalization... Enter the update impedance weight The static diagram base is modified to reflect the characteristics of the equipment.

[0127] When updating impedance weights After writing to the edge gateway, step four does not consider the current closed loop complete and continues to read the combined feedback from the chassis gyroscope, the boom-end angle encoder, and the component end position tag. Preferably, the gyroscope is mounted on a rigid mounting base on the assembly machine chassis, the angle encoder is placed at the telescopic boom's rotation and pitch axes, and the position tag is installed at the component's center of gravity line. The edge gateway observes whether the pitch and roll angle offsets accumulate unidirectionally, starting from the end of the motion segment and ending at the end of the hold segment.

[0128] If the offset can cancel each other out between adjacent cycles, the current closed loop is still considered to be able to absorb it; if the offset continues to accumulate on the same side, and the remaining clearance from the component's outer envelope to adjacent cable trays, pipelines, or walls continues to decrease, then the edge gateway stops absorbing the residual. Specifically:

[0129]

[0130] Where: cumulative offset tensor of machine attitude micro-shaking : Represents the critical residual load composed of pitch, roll, and approach frequencies within the holding segment, and its value is a three-dimensional column vector; pitch integral. : Represents the time integral of the absolute value of the pitch angle offset within the segment, with a value of Used to characterize the continuous nodding and yaw of a component along the front-to-back direction; roll integral. : Represents the time integral of the absolute value of the roll angle offset within the segment, with a value of , used to characterize the continuous lateral sway of a component in the left-right direction;

[0131] Approximation frequency integral : Represents the cumulative frequency integral that keeps the safety envelope boundary within the segment continuously approximated, with a value of ;

[0132] Start time : Indicates the start time of the action segment's completion and entry into the hold phase; the value is a valid timestamp; end time. : Indicates the time when the segment ends or freeze is triggered, and the value must be a valid timestamp that satisfies the following conditions. Pitch angle offset : Indicates time The pitch angle offset relative to the planned attitude, expressed in degrees; the roll angle offset. : Indicates time The roll angle is the offset relative to the planned attitude, and its value is in degrees.

[0133] Approximation function : Indicates time The safe envelope approximates the event strength, taking a non-negative value. A higher value is taken when the remaining net distance continuously decreases in the same direction, which is used to incorporate the fact that the distance is being eaten into the tensor.

[0134]

[0135] In the formula: Safety envelope approaching limit alarm indicator : Indicates whether the dangerous residual is thrown outward in step four to step five, with a value of 0 or 1, where 1 indicates entering the outward throwing suspension state; remaining net distance : Represents the minimum clear distance between the outer envelope of the current component and its neighboring entities, with a value of The remaining net distance is derived from the combined calculation of location tags, angle encoders, and pre-stored component envelopes; Take the minimum Euclidean distance from the current component's outer envelope sampling point to the nearest obstacle's outer envelope sampling point.

[0136] Clearance threshold : Represents the net distance threshold that triggers the outward throw suspension, with a value of Pre-existing edge gateways based on component type, covering material thickness, and clamping method; integration threshold. : Represents the pitch integral With roll integral The total trigger threshold, with a value of This is used to distinguish between a single small oscillation and a continuous, cumulative micro-shaking; the approximation function When the remaining net distance is in multiple adjacent sampling periods A valid approximation event is defined as an event that continuously decreases in the same direction, while an invalid approximation event is defined as an event that occurs when the remaining net distance increases or the decreasing direction reverses.

[0137] Approximation function Determined based on the trend of remaining net distance changes within adjacent sampling periods. When the remaining net distance... If the danger direction marker is not less than the preset net distance contraction threshold and remains consistent for two consecutive sampling periods, the current sampling period is recorded as a valid approximation event. Set to 1; when the remaining net distance increases, reverses direction, or changes by less than the net distance contraction threshold, Set to 0. Step four: Based on this, accumulate the approximation frequency integral within the holding segment. .

[0138]

[0139] in, This represents the current remaining clearance, used to determine whether the static safety clearance is being continuously eroded. It is obtained by sampling the current component envelope set. Set of sampling points with neighboring obstacle envelope Calculate the Euclidean distance point by point and take the minimum value; Indicates the current component envelope sampling point. This indicates sampling points within the envelope of nearby obstacles.

[0140] As a supplement, an alarm should only be triggered when the safety envelope approaches the limit indicator. Only then will the edge gateway update the impedance weights corresponding to the current action segment. Write back to the directed graph with static collision avoidance constraints; when When this happens, freeze the smoothing results of the current action segment that have not yet been committed, and only retain the updated impedance weights that have been committed for completed action segments. As a historical boundary weight, it also outputs the cumulative offset tensor of the machine's attitude micro-shaking. Direction markers Direction markers Safety envelope approaching limit warning indicator Provided for takeover in step five.

[0141] During on-site implementation, once the modular air conditioning unit section had bypassed the cable tray and was ready to be positioned against the wall, the machine appeared to have stopped major movements. However, due to uneven ground and the rebound of the telescopic boom, the chassis was still exhibiting slight swaying in the same direction. Workers standing to the side could see the edge of the unit section gradually approaching the wall, each approach being small but consistently in the same direction. At this point, the edge gateway measured the pitch integral. With roll integral The remaining net distance continues to increase. Gradually below the net distance threshold Safety envelope approaching limit warning indicator Set to 1. The edge gateway no longer attempts to update impedance weights alone. Instead of swallowing the oscillation, freeze the current level smoothness.

[0142] Specifically, the cumulative offset tensor of machine attitude micro-shaking It retains the directionality and cumulativeity of continuous micro-shaking; the safety envelope approaches the limit warning indicator. Clearly separate out the states that can no longer be digested internally.

[0143] When the safety envelope approaches the limit, an alarm is triggered. Once set to 1, the edge gateway stops writing and updating impedance weights to the current edge. And combine the execution context of the current action segment with the cumulative offset tensor of the machine's attitude micro-shaking. The payload is packaged together as an external probe payload. Preferably, the payload is expressed in JSON or C structure format, with fields including task number, current edge identifier, and deadlock exception node identifier. Pitch integral Roll integral Approximation frequency integral Remaining net distance Safety envelope approaching limit warning sign The edge gateway sends the payload to the takeover buffer corresponding to step five via shared memory or high-speed bus, and simultaneously sends a freeze message to the execution controller to maintain the current safe posture, so that the assembly tool maintains the existing clamping state, restricts rotation, and suspends further wall contact.

[0144] In the parallel extended implementation, in addition to writing the outward probe payload to the takeover cache in step five, the edge gateway can also push status messages to the cloud dashboard via COAP or an equivalent lightweight transmission protocol, and reset the on-site sign indicating that the edge is frozen and not allowed to stick to the wall to the AR terminal worn by the worker.

[0145] In the degradation-assurance implementation, if the wireless link is lost or the cloud is unreachable, the edge gateway pushes the probe payload into a local circular queue and keeps the frozen message valid on the execution controller side. The fields recorded in the circular queue are consistent with the fields of the normal outgoing payload, and are resent in timestamp order after the link is restored; if the micro-shake attenuates during the freeze period and the remaining clearance is... Return to net distance threshold In addition, the edge gateway does not directly restore the original action, but requires step five or the manual confirmation unit to explicitly unfreeze.

[0146] Specifically, the dangerous residual is completely stripped into the closed loop of this level, and step five can receive the complete context and tensor load; the degradation protection path also retains the dangerous residual and freeze action when the chain breaks. Therefore, step five can receive the complete probe load with the current edge context, net distance state and micro-shaking direction accumulation.

[0147] Step 5: Accumulate the offset tensor of the slight wobbling of the throwing device from Step 4. The inverse solution is used to calculate the virtual collision avoidance envelope expansion, and the resulting polarization impedance weights are then used. It is directly overwritten onto the directed graph of static anti-collision constraints, and entities are forced to reroute their paths before a collision. This is accomplished collaboratively by the reverse overwrite engine, geometry mirror cache, impedance overwrite register, and fieldbus master station within the edge gateway.

[0148] Steps one through four have reduced the three-dimensional entity to a static anti-collision constraint directed graph and solved the problems of optical misjudgment and local rerouting. However, the boundary of this static anti-collision constraint directed graph is still based on the premise that the component passes through according to the planned attitude.

[0149] For deep-ground heavy assembly sites, the truly dangerous moment is not when the component is completely out of control, but when the component is still moving along the correct path, but experiences continuous micro-swaying due to boom rebound, eccentric load pull, and support settlement, causing the originally sufficient static safety clearance to be gradually eroded. If the smoothing correction within step four is still relied upon at this point, the directed graph of the static anti-collision constraint will only treat these micro-swaying as short-term disturbances. By the time the edge end realizes the problem, the outer envelope of the entity has already approached the cable tray, beam side, or existing air duct.

[0150] Specifically, we first acknowledge that the continuous micro-shaking has exceeded the expressive capacity of the static anti-collision constraint directed graph. Then, we convert this continuous micro-shaking into an envelope expansion in virtual space. Next, we recalculate the impedance using the material penalty logic disclosed in step one, and write the new impedance directly back to the static anti-collision constraint directed graph. As a result, the higher-level scheduling brain no longer simply dispatches orders downwards, but its passage boundaries are reversed by the lower-level physical state, ultimately forming a cyber-physical backfeed closed loop.

[0151] The edge gateway first reads the task number, current edge identifier, and cumulative offset tensor of the device's attitude micro-shaking from the takeover cache in step four. Safety envelope approaching limit warning sign and updating impedance weights When the safety envelope approaches its limit, an alarm is triggered. When in the set state, the geometry mirror cache loads the original component envelope and the neighboring obstacle envelope corresponding to the current edge identifier; subsequently, the reverse overwrite engine accumulates the offset tensor of the machine attitude micro-shaking. pitch integral in Roll integral With approximation frequency integral Substitute the nonlinear dilation operator to generate the dilation envelope volume; then, based on the safety tolerance volume from step one... Initial setting time of materials and secondary demolition penalty coefficient The volume difference of expansion is converted into an increment of polarization impedance; finally, the fieldbus master station weights the polarization impedance. The weights are overwritten into the weight registers of the current edge and its adjacent edges, and the local rerouting engine in step three is immediately triggered to solve the problem again. If the solution is solved again after the emergency response window has expired, the inverse kinematics retreat path is switched to.

[0152] Preferably, a pre-built geometric mirror cache is provided within the edge gateway. The cache content comes from the component sweep space domain mirror generated during compilation before construction in step one, which includes the original envelope, neighboring obstacle envelope, slewing envelope, and segment number of the component to be installed. After step five is triggered, the cloud model is not accessed again; instead, the envelope fragment corresponding to the current edge identifier is read directly from the local geometric mirror cache. The reverse overwrite engine then accumulates the offset tensor by micro-shaking the machine's attitude. As the only dangerous input, the pitch integral is... This is interpreted as the increase in head and tail swing along the longitudinal direction of the component, and the integral of the roll. Interpreted as the lateral sway increment along the transverse direction of the component, and the approximation frequency integral. This can be interpreted as persistent evidence of repeated approach to the same dangerous direction.

[0153] All three elements enter the nonlinear expansion operator, which does not simply perform linear expansion, but rather applies directional expansion to the original component envelope, causing the envelope on the side already exhibiting unidirectional risk accumulation to expand more rapidly. Specifically:

[0154]

[0155] Where: the volume of the expansion envelope : Represents the cumulative offset tensor of machine attitude micro-shaking The virtual anti-collision envelope volume obtained after the action is taken as a value. Used to reflect the dangerous occupancy range of a component under continuous micro-swaying; original envelope volume. : Represents the basic envelope volume of the current component written to the geometry mirror cache during the compilation phase in step one, with a value of Pitch integral : Represents the cumulative pitch offset output from step four, with a value of Roll integral : Represents the cumulative roll direction offset output from step four, with a value of Approximation frequency integral : Represents the cumulative amount of approximation events output in step four, with a value of This is used to indicate the degree to which a dangerous direction is repeatedly approached;

[0156] Frequency Amplification : Represents the integral quantity of the approximation frequency The gain coefficient, taking a value of Pre-storage is preferred based on component length and clamping method; integration threshold : This represents the threshold value used in step four to distinguish between a one-time small oscillation and a continuous micro-oscillation, and its value is [value missing]. Stability constant : Represents a positive constant to prevent the exponent from becoming unstable due to an excessively small denominator; its value is [value missing]. This is used to ensure the continuity of the expansion operator;

[0157] Step 5: Call the accumulated offset tensor of the machine's attitude micro-shaking Instead of isotropically expanding the original envelope, the danger approach side is first determined based on the pitch and roll directions. Then, dilation is performed only on the envelope patch corresponding to the danger approach side; the envelope patch opposite to the danger approach side retains its original envelope. The resulting dilation volume difference is used to determine the polarization impedance weights of the current edge and its adjacent edges.

[0158] During on-site implementation, a rectangular duct was hoisted into the underground refrigeration room. The duct had already bypassed the cable tray along the alternative path, but when passing through the final narrow gap against the beam, the boom experienced a slight sway in the same direction due to eccentric loading. Workers could see that the duct flange edge moved closer to the beam side after each sway. Step 5: Read the cumulative offset tensor of the machine's attitude slight sway. Then, a rapid expansion is performed on the virtual envelope on the side where the duct flange is located, and the passage gap on the screen is immediately reduced, and the edge that could be passed through before is regarded as a high-risk edge.

[0159] Specifically, step five no longer treats continuous micro-wobbles as ordinary angular errors, but directly converts them into virtual space occupancy increments; because the expansion uses pitch integrals. Roll integral and the integral of the approximation frequency The combination of quantities, with continuous slight swaying, triggers dangerous tightening earlier than occasional simple pendulum movements.

[0160] Only the expansion envelope volume It cannot directly drive path rerouting yet because steps three and four still call the edge weights of the static collision avoidance constraint directed graph, rather than the 3D volume itself. Therefore, the reverse overwrite engine obtains the dilated envelope volume. Then, the volume difference of expansion is extracted and returned to the impedance system established in step one.

[0161] If the current danger zone is adjacent to newly poured concrete, structural adhesive joints, fireproof sealing layers, or areas requiring secondary drilling, then the initial setting time of the material already specified in step one should be considered. Penalty coefficient for secondary demolition This is invoked again here, highlighting that the envelope expansion caused by physical micro-shaking not only reflects spatial intrusion but also engineering consequences. Specifically:

[0162]

[0163] Where: Polarization impedance increment : Represents the additional impedance calculated from the volume difference due to expansion, and its value is... This is used to translate the spatial risk of continuous micro-shaking into algebraic quantities with identifiable edge weights in a directed graph; the expansion envelope volume. : Represents the virtual envelope volume output by the nonlinear dilation operator, with a value of Original envelope volume : Represents the basic envelope volume in the geometric mirror cache of step one, with a value of Safety tolerance volume : Represents the safety tolerance volume used in step one to normalize spatial risk, with a value of Natural constant : Indicates the base of the exponential amplification, which is a fixed value and is used to extend the nonlinear amplification of the material consequences of the step;

[0164] Initial setting time of materials : This represents the initial setting time parameter of the material in the material process log of step one, with a value of This is used to amplify the risks as they approach the irreversible curing region;

[0165] Secondary demolition penalty coefficient : Indicates the secondary demolition penalty coefficient in the step one repair rule table, with a value of This is used to increase the impedance when penetrating regions with high repair costs.

[0166]

[0167] Where: polarization impedance weight : This indicates the new edge weight to be overwritten in the static collision avoidance constrained directed graph in step five, with a value of This is used to directly change the cost basis of subsequent solutions in step three; and to update the impedance weights. This indicates that step four has already written back the basic impedance value based on the execution hysteresis, and the value is [value missing]. Polarization impedance increment : Represents the additional impedance formed by the combined effects of volume expansion and material consequences, with a value of ;

[0168] During on-site implementation, if the right side of the rectangular duct beam-attached passage is a structurally glued hanger base, and the left side is a removable temporary support, then even if the expansion volume difference caused by continuous slight shaking on both sides is similar, the polarization impedance increment on the right side will be... It will also increase significantly because the structural adhesive area corresponds to a longer initial setting time of the material. Penalty coefficient for secondary demolition Based on this, the edge gateway prioritizes classifying the more dangerous side as impassable.

[0169] Specifically, step five continues the terminology and consequence logic of step one. Continuous micro-oscillation is no longer just a geometric problem, but is directly compiled into polarization impedance weights with material consequences. Different material environments will result in different edge weight increases for the same envelope expansion, thus making the rerouting results more consistent with the construction consequences. In step one, when generating the static anti-collision constraint directed graph, an edge weight index is generated synchronously for each directed edge. The edge weight index records at least the current edge identifier, the target edge weight address, and the set of adjacent edge identifiers sharing the same start or end point as the current edge. In step five, after receiving the current edge identifier, the target edge weight address corresponding to the current edge identifier is first located based on the edge weight index. Then, based on the overlap increment between the expansion envelope of the component to which the current edge identifier belongs and the envelope of the adjacent obstacle, the set of adjacent edge identifiers whose weights need to be synchronously increased is determined. Subsequently, the polarization impedance weight is written into the target edge weight address, and the attenuated polarization impedance weight is written into the target edge weight address corresponding to the set of adjacent edge identifiers. The current edge identifier still uses the ordered combination of the source node identifier and the target node identifier generated in step one, without introducing a new edge numbering system.

[0170] When polarization impedance weight After formation, step five proceeds to the cross-dimensional write-back stage. Preferably, the fieldbus master station uses EtherCAT or an equivalent fieldbus to access the impedance write-back register, which is mapped one-to-one with the edge buffer of the static anti-collision constraint directed graph. The reverse write-back engine first locks the register address corresponding to the current edge identifier, and then applies the polarization impedance weights. Write the information to this address and send an immediate re-solution signal to the local rerouting engine in step three. The overwrite message preferably includes the task number, current edge identifier, register address, target weight, timestamp, and cyclic redundancy check field. If overwriting fails for two consecutive frames, the fieldbus master maintains the current frozen state and directly enters the inverse kinematic retreat path. At this point, the edge weights of the static anti-collision constraint directed graph have been rewritten in reverse by the physical residuals, and even if the original solver is used in step three, the raised dangerous edge will be automatically avoided.

[0171] In the enhanced combination implementation, step five not only overwrites the current edge, but also raises the impedance by two or three layers outward according to the adjacent jump layer, forming an impedance ripple broadcast.

[0172]

[0173] in, Indicates the first Ripple impedance weighting of adjacent edges; This indicates the polarization impedance weight of the current edge; This represents the ripple attenuation coefficient, and its value satisfies... Before construction, the parameters are written into the edge gateway parameter table according to the component category and equipment level; Indicates the first The number of hops between adjacent edges relative to the current edge. The edge gateway only supports this. Adjacent edges that do not exceed the preset broadcast jump level limit will be broadcast in a ripple manner, while adjacent edges that exceed the jump level limit will retain their original weights.

[0174] The specific method is as follows: Adjacent edges sharing a common endpoint with the current edge are assigned a weakened broadcast weight, causing the coordinated machine group around the danger zone to contract its action boundary in advance, preventing multiple components from simultaneously approaching the same dangerous gap. This broadcast process is completed using a pre-stored skip-layer table and does not require full-map diffusion. The visible result on-site is that not only does the current duct stop passing over the beam, but adjacent cable trays awaiting installation and subsequent small pipe sections are also briefly restricted. Workers in the surrounding area see multiple paths turning red simultaneously, but this only occurs near the danger zone and does not affect the entire floor's work area.

[0175] In parallel alternative implementations, if the estimated time for solving step three again exceeds the emergency response window... Then, step five skips the directed graph and solves it directly, switching to inverse kinematics yield path. (Disaster relief window) Before construction, the edge gateway writes parameters into a table according to component type, clamping method, and equipment level. The edge gateway estimates the time required for the current re-solution based on the rolling average time of the most recent local re-solutions; if the estimated time exceeds the emergency response window... When this happens, stop resolving and sending the current map and directly switch to the inverse kinematics yield path.

[0176]

[0177] in, This represents the joint increment vector of the assembly tool, used to drive the hydraulic servo to perform the retraction. This represents the generalized inverse of the Jacobian matrix of the current assembly machine joint; This represents the yield displacement vector along the outer normal of the danger side, calculated from the overlap depth of the expansion envelope and the envelope of the adjacent obstacle.

[0178] The edge gateway calculates the normal yield vector based on the current component's center of gravity, gripping point, and arm end posture, and sends it to the hydraulic servo drive via the fieldbus master station. This causes the component to first yield a safe displacement away from the danger surface, and then waits for the directed graph edge weights to be restored. In the degradation-assurance implementation, if the reverse overwrite engine cannot perform a complete expansion calculation due to a computing power alarm, it calls the deflection angle-impedance multiple digital dictionary in the local read-only memory, based on the accumulated offset tensor of the machine's posture micro-shaking. The corresponding sway level directly updates the impedance weights. Multiply by the curing factor to obtain the coarse-grained polarization impedance weight. Although this path abandons the solution of continuous expansion volume, it still maintains the core structure of the directed graph with static anti-collision constraints described in the upward write-back of physical residuals.

[0179] Specifically, polarization impedance weighting The path solution result can be changed immediately after the register is overwritten. Impedance ripple broadcasting further expands the single-point danger into local coordinated contraction. Emergency retreat switching and degradation backup paths ensure that step five can still be implemented within a very small safety window.

[0180] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0181] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0182] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0183] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0184] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for topology analysis and intelligent scheduling of electromechanical installation tasks based on BIM digital twins, characterized by: Performed by the edge gateway, including: A static anti-collision constraint directed graph is pre-stored, which is compiled by dimensionality reduction of component sweeping interference; visual suspension is triggered based on on-site optical data, and the current target node is identified as a deadlock abnormal node when the waveform at the bottom layer of the working tool does not meet the real chain breakage criterion; Extract local spillover subgraphs around deadlock anomaly nodes, overwrite the anomaly edges with impedance masks and generate alternative flow instruction sequences; issue alternative flow instruction sequences and update graph edge impedances according to execution hysteresis; generate machine attitude micro-shaking cumulative offset tensor and safety envelope approaching limit alarm flag when continuous micro-shaking causes the safety envelope to approach the limit. In response to the safety envelope approaching the limit alarm, a virtual anti-collision envelope expansion calculation is performed based on the cumulative offset tensor of the machine's attitude micro-shaking. The expansion result is converted into polarization impedance weights and written back to the edge weights of the static anti-collision constraint directed graph, triggering a rerouting or retreat.

2. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 1, characterized in that: The static anti-collision constraint directed graph is generated by the edge gateway based on the overlap results of the pre-stored component sweep space domain and the envelope of the adjacent obstacle. The edge gateway defines the segment nodes as lifting position, rotation position, wall-hugging transition position, entry position and landing position, and generates directed edges and corresponding edge weight index items based on the continuous action relationship between the segment nodes.

3. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 2, characterized in that: When the glare exponential index exceeds a preset threshold, the edge gateway suspends the visual access decision and writes the lock window start point. The edge gateway collects the underlying execution waveform of the operating equipment within the locked window. When the underlying execution waveform does not meet the true link breakage criteria, the current target node is identified as a deadlock abnormal node.

4. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 3, characterized in that: The edge gateway calls the adjacency list of the static anti-collision constraint directed graph around the deadlock abnormal node, extracts the local spillover subgraph according to the abnormal out-degree and skip-level boundary, and writes the directed edges pointing to the blocked bit segment into the abnormal edge set. After performing impedance mask overwriting on the edge weights corresponding to the abnormal edge set, a replacement flow instruction sequence is generated.

5. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 4, characterized in that: The edge gateway breaks down the alternative flow instruction sequence into back segment, lift segment, turn segment, and position segment, and updates the graph edge impedance based on the start deviation, position deviation, and holding deviation of the actual action relative to the planned action; when continuous micro-shaking causes the remaining clearance to decrease, the smoothing result of the segment is frozen and an alarm flag indicating that the machine attitude micro-shaking cumulative offset tensor and safety envelope approach limit are generated.

6. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 1, characterized in that: The edge gateway pre-stores an edge weight index for each directed edge. The edge weight index records at least the current edge identifier, the target edge weight address, and the set of adjacent edge identifiers that share the start or end point with the current edge, as well as the corresponding task number and graph version number, for subsequent writing of polarization impedance weights.

7. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 6, characterized in that: When generating the alternative flow instruction sequence, the edge gateway synchronously writes the task number, current edge identifier, deadlock abnormal node identifier, abnormal edge set, alternative flow instruction sequence, and frozen status identifier; and writes the fields to the local task buffer and execution buffer for use by the execution controller and field terminals.

8. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 7, characterized in that: When the security envelope approaching limit alarm flag is set, the edge gateway outputs the current edge flag, the prohibited edge flag, the alternative flow instruction sequence, and the frozen status to the field terminal. And until the alarm sign indicating that the safety envelope is approaching its limit is lifted, the on-site terminal should maintain the "no passage" display status for the current side.

9. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 6, characterized in that: The edge gateway sends an edge weight overwrite instruction frame through the fieldbus master station; the edge weight overwrite instruction frame includes at least the current edge identifier, the target edge weight address, the polarization impedance weight, and the timestamp; and when it receives an overwrite failure response, it maintains the current frozen state and stops the current graph re-solution.

10. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 1, characterized in that: The edge gateway connects the optical acquisition terminal, waveform acquisition terminal, execution controller, geometric mirror cache, and fieldbus master station; the optical acquisition terminal is used to provide field optical data, the waveform acquisition terminal is used to provide the underlying execution waveform of the work equipment, and the geometric mirror cache is used to provide the original envelope and the envelope of the adjacent obstacles corresponding to the current edge.

11. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 3, characterized in that: The width of the lock window is 400 to 600 milliseconds; the edge gateway only receives the underlying execution waveform of the working machine within the lock window, and excludes the execution waveform outside the lock window from the current visual suspension cycle, and uses the start point of the lock window as the unified time reference for the waveform determination.

12. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 4, characterized in that: When extracting the local sweep subgraph, the edge gateway limits the upper limit of the out-degree of the immediate successor of the deadlock anomalous node to 8 to 12, and limits the maximum hop level of the candidate node relative to the deadlock anomalous node to two levels; nodes that exceed the boundary are not written into the local sweep subgraph.

13. The electromechanical installation task topology analysis and intelligent scheduling method according to claim 1, characterized in that: If the estimated time for resolving the map exceeds the preset emergency response window, the edge gateway will stop resolving the map and convert the virtual anti-collision envelope expansion solution into a yield displacement command along the outer normal of the dangerous side. This command will be sent to the hydraulic servo drive to execute the yield action, while maintaining the frozen state of the current edge.

Images