Truck crane group operation safety monitoring system and method based on multi-source perception and collaborative decision

Abstract

The invention discloses a truck-mounted crane group operation safety monitoring system and method based on multi-source perception and collaborative decision, and particularly relates to the field of safety monitoring, the truck-mounted crane group operation safety monitoring system comprises a distributed perception module, an edge calculation module, a graded early warning and execution module, a digital twin platform module and a man-machine interaction module; collected multi-source data are fused through Kalman filtering and then transmitted to an edge computing node; generating a dynamic safety envelope region based on an adaptive kinematics model, predicting a group conflict risk through a graph theory algorithm, and calculating collision time; the digital twin platform fuses a BIM model and real-time data, hoisting path rehearsal, load swing prediction and foundation bearing capacity visualization are carried out, an improved RRT * algorithm is adopted to plan a collision-free path, and a Lagrange dynamics model is adopted to predict a swing track; the grading early warning module triggers differential response according to the risk grade; and risk warning and path guidance are provided through the AR-HUD and the three-dimensional situation billboard.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology, and more specifically, to a safety monitoring system and method for group operations of truck cranes based on multi-source perception and collaborative decision-making. Background Technology

[0002] In numerous fields such as modern construction engineering, logistics and transportation, and large equipment installation, truck cranes play an indispensable role as a key lifting equipment. They possess significant advantages such as mobility, high lifting height, and wide operating range, enabling them to adapt to various complex and changing operating environments and meet lifting needs in different scenarios. With the continuous advancement of global infrastructure construction and the expansion of industrial production scale, the frequency of truck crane use is increasing, the scale of operations is becoming larger, and group operation scenarios are becoming more common. For example, in projects such as large bridge construction, high-rise building construction, and port cargo handling, multiple truck cranes are often required to work together to complete tasks such as lifting and transporting large components.

[0003] However, the complexity of group operations with mobile cranes also presents significant challenges to operational safety. On one hand, the dynamic changes in the crane's boom swing and the lifting and lowering of loads during operation create a constantly shifting working environment, increasing the uncertainty of safety risks. On the other hand, when multiple mobile cranes operate simultaneously, there are spatial interferences and coordination issues related to the operational process. For example, the booms of different mobile cranes may collide due to limited working space, or improper coordination during lifting may lead to serious accidents such as the load becoming unbalanced and falling. These safety hazards can not only cause equipment damage and personal injury, but also delay project progress, result in significant economic losses, and even attract public attention and negative impacts. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a safety monitoring system and method for group operations of truck cranes based on multi-source perception and collaborative decision-making, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Distributed sensing module: used to collect real-time operational environment data and status data;

[0007] Edge computing module: used to process sensor data in real time, generate dynamic safety envelopes and predict group conflict risks;

[0008] Tiered early warning and execution module: Triggers different response measures based on the risk level;

[0009] Digital Twin Platform Module: Based on BIM model and real-time sensor data, it constructs a three-dimensional operation scene to perform hoisting path pre-simulation, load swing trajectory prediction and foundation bearing capacity visualization;

[0010] Human-computer interaction module: including driver AR-HUD interface and command center 3D situation dashboard, providing risk source highlighting, recommended avoidance routes and global operation heat map.

[0011] Preferably, the multi-source sensing module consists of a multi-source sensor group installed on each truck crane, including millimeter-wave radar, binocular vision sensor, IMU inertial unit, and boom angle encoder.

[0012] Comprehensive perception of the working environment and equipment status is achieved through multi-sensor data fusion; millimeter-wave radar detects obstacles within a 50-meter radius at a scanning frequency of 20Hz, accurately identifying the relative positions of other equipment, personnel, and temporary structures, with a ranging accuracy of ±3cm; a binocular vision sensor installed at the end of the boom tracks the hook's movement trajectory in real time based on deep learning algorithms, calculates the swing amplitude and direction of the load through stereo matching technology, and detects potential collision risks in the surrounding area by combining feature point recognition technology; a high-precision IMU inertial unit continuously monitors changes in vehicle body posture at a sampling frequency of 200Hz, and, together with real-time boom length and elevation angle data provided by the boom angle encoder, constructs an accurate kinematic model of the equipment.

[0013] Preferably, in the edge computing module, the edge computing module processes multi-source sensing data in real time through a distributed architecture to construct a three-dimensional safety protection system; the dynamic safety envelope generator adopts an adaptive kinematic model to calculate the safe working space based on the real-time collected boom parameters and environmental data; the adaptive kinematic model is updated every 100ms to generate a safety envelope that dynamically changes with the working conditions; the group risk prediction engine analyzes the spatial relationship of the envelopes of multiple devices through graph theory algorithms, and when overlapping envelopes are detected, a time collision prediction model is used.

[0014] Preferably, the graded early warning and execution module adopts a multi-level response mechanism, and uses intelligent decision-making algorithms to achieve closed-loop control from risk early warning to automatic intervention; based on real-time risk assessment results, the operational risks are divided into three levels and differentiated response strategies are triggered.

[0015] When the calculated TTC ij When the value is less than 5, the risk level is 1, and dangerous actions are automatically cut off; when the calculated value is 5 ≤ ​​TTC... ij <15, risk level 2, then cab vibration + voice prompt; when the calculated TTC ij If the risk level is 3 at 3 PM, the command center will issue a pop-up alarm.

[0016] At Level 1 risk, a linkage response is triggered, sending an emergency braking command to the hydraulic control system via the CAN bus. At Level 2 risk, preventative intervention is initiated, with the cab's tactile feedback device generating a vibration pattern at a specific frequency, simultaneously displaying a yellow buffer envelope on the AR interface. Voice prompts, using natural language generation technology, provide specific guidance such as "Equipment is approaching from the right rear; a 15° left turn is recommended." The execution unit limits the operating speed in the dangerous direction. At Level 3 risk, a management-level response is triggered, automatically displaying a live video stream and risk parameter trend chart on the command center's large screen. The location of the risk source is marked in the BIM model, and an electronic work order containing avoidance suggestions is generated and pushed to the management personnel's terminal. Simultaneously, the risk event is recorded in the blockchain evidence storage unit. For foundation risks, a pressure redistribution algorithm is activated.

[0017] Preferably, in the digital twin platform module, a high-fidelity dynamic virtual operation scene is constructed by deeply integrating the BIM model with real-time multi-source sensor data. First, the engineering BIM model is imported as the basic scene framework, and key structural features are retained through lightweight processing, and a two-way data channel with the physical equipment is established. Real-time data fusion adopts a spatiotemporal registration algorithm to map UWB positioning data, IMU attitude data and visual recognition results to the virtual scene. For hoisting path planning, the platform integrates a kinematics solver and generates a collision-free path based on the improved RRT* algorithm, whose cost function considers equipment motion constraints.

[0018] The load swing prediction uses a Lagrange dynamics model, and the specific calculation method is as follows:

[0019]

[0020] Where β2 represents the angular acceleration of the load swing, g represents the gravitational acceleration, l represents the distance from the hook to the load, β represents the angle of the load swing, β1 represents the angular velocity of the load swing, and a x Let a be the acceleration of the hook in the x-direction. y Let represent the acceleration of the hook in the y direction, m represent the load mass, and c represent the damping coefficient.

[0021] By calculating the hook acceleration a in real time x a y Given the angle β of the load swing, predict the swing trajectory for the next 5 seconds and visualize it in a probabilistic cloud map in a 3D scene.

[0022] Preferably, in the human-computer interaction module, the AR-HUD system at the driver terminal uses waveguide projection technology to overlay key information at a virtual image distance of 7 meters in front of the driver's field of vision. Spatial registration technology is used to accurately align the virtual information with the real scene, with a positioning error of less than 0.3°. It receives risk prediction data from the graded warning and execution module in real time. When a potential collision is detected, a three-level visual warning strategy is adopted: Level I risk triggers a red pulse border warning, and simultaneously projects a 3D arrow on the windshield indicating the emergency avoidance direction; Level II risk displays a yellow semi-transparent envelope outline, overlaid with a green guide line for the recommended path. Path planning considers the kinematic constraints of the boom, and the calculation method is as follows:

[0023]

[0024] Where Γ0 represents the optimal path, ∫ δ ds represents the total path length, and λ represents the weighting factor. The maximum value of the boom angular acceleration is represented by μ, which is the boom kinematic constraint; μ represents the boom rotation angle; and T1 represents the total time for path planning.

[0025] The 3D situation dashboard deployed in the command center is built on WebGL technology, supporting free switching between multiple perspectives and dynamic adjustment of detail levels.

[0026] The technical effects and advantages of this invention are as follows:

[0027] This invention utilizes a distributed sensing module to collect environmental and equipment data in real time, which is then fused using Kalman filtering to construct a three-dimensional safety envelope. An edge computing module employs an adaptive kinematics model to predict collision risks and combines graph theory algorithms to analyze multi-device interactions. A digital twin platform integrates BIM and sensor data, using an improved RRT* algorithm to plan paths, a Lagrange model to predict load sway, and a finite element method to assess foundation stress. Based on the risk level, braking, AR prompts, or command center intervention are triggered, and the actuators employ a dual-redundancy design to ensure reliability. A human-machine interaction module provides AR-HUD real-time alerts and a WebGL three-dimensional situation dashboard, supporting heatmap analysis and multimodal interaction. This invention achieves closed-loop management from perception and decision-making to control, improving the safety and collaborative efficiency of group operations. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the module connection of the present invention.

[0029] Figure 2 Schematic diagram of the method of the present invention. Detailed Implementation

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0031] Please see Figure 1 As shown, the present invention provides a safety monitoring system for group operations of truck cranes based on multi-source perception and collaborative decision-making, including a distributed perception module, an edge computing module, a hierarchical early warning and execution module, a digital twin platform module, and a human-computer interaction module.

[0032] Distributed sensing module: used to collect real-time operational environment data and status data;

[0033] The multi-source sensing module consists of a multi-source sensor group installed on each truck crane, including millimeter-wave radar, binocular vision sensor, IMU inertial unit, and boom angle encoder.

[0034] Comprehensive perception of the working environment and equipment status is achieved through multi-sensor data fusion; millimeter-wave radar detects obstacles within a 50-meter radius at a scanning frequency of 20Hz, accurately identifying the relative positions of other equipment, personnel, and temporary structures, with a ranging accuracy of ±3cm; a binocular vision sensor installed at the end of the boom tracks the hook's movement trajectory in real time based on deep learning algorithms, calculates the swing amplitude and direction of the load through stereo matching technology, and detects potential collision risks in the surrounding area by combining feature point recognition technology; a high-precision IMU inertial unit continuously monitors changes in vehicle body posture at a sampling frequency of 200Hz, and, together with real-time boom length and elevation angle data provided by the boom angle encoder, constructs an accurate kinematic model of the equipment.

[0035] All sensor data are timestamped through a time synchronization module and multi-source data are fused using a Kalman filter algorithm to eliminate measurement errors from a single sensor. The fused data contains multi-dimensional information such as device pose, distribution of environmental obstacles, and load motion status, and is transmitted to the edge computing node via 5G communication with a delay of less than 50ms.

[0036] Edge computing module: used to process sensor data in real time, generate dynamic safety envelopes and predict group conflict risks;

[0037] In the edge computing module, multi-source sensing data is processed in real time through a distributed architecture to construct a three-dimensional safety protection system; the dynamic safety envelope generator adopts an adaptive kinematic model to calculate the safe working space based on the real-time collected boom parameters and environmental data. The specific calculation method is as follows:

[0038]

[0039] Among them, V safe (t) represents the safety value at time t, L(t) represents the real-time boom length, θ(t) represents the current elevation angle, ω(t) represents the slewing angular velocity, and Q(t) represents the load weight. max The maximum lifting weight is represented by ε(·), which represents the environmental correction term. wind Expressed as the combined wind speed, α ground This is expressed as the inclination of the foundation.

[0040] The adaptive kinematics model is updated every 100ms, generating a safety envelope that dynamically changes with operating conditions. The group risk prediction engine analyzes the spatial relationship of the envelopes of multiple devices using graph theory algorithms. When overlapping envelopes are detected, a time-based collision prediction model is used. The specific calculation method is as follows:

[0041]

[0042] Among them, TTC ij P represents the estimated collision time. i 、P j Represented as a position vector, v i v j Represented as a velocity vector, overlap ij V is represented by the overlapping volume of the envelope regions of device i and device j. envelope It is expressed as the envelope volume of a single device.

[0043] Tiered early warning and execution module: Triggers different response measures based on the risk level;

[0044] The graded early warning and execution module adopts a multi-level response mechanism and uses intelligent decision-making algorithms to achieve closed-loop control from risk early warning to automatic intervention. Based on real-time risk assessment results, the operational risks are divided into three levels and differentiated response strategies are triggered.

[0045] When the calculated TTC ij When the value is less than 5, the risk level is 1, and dangerous actions are automatically cut off; when the calculated value is 5 ≤ ​​TTC... ij <15, risk level 2, then cab vibration + voice prompt; when the calculated TTC ij If the risk level is 3 at 3 PM, the command center will issue a pop-up alarm.

[0046] At Level 1 risk, a linkage response is triggered, sending an emergency braking command to the hydraulic control system via the CAN bus. The control logic is as follows:

[0047]

[0048] Among them, F brake The braking force, k, represents the force required for emergency braking. p Represented as the proportional gain coefficient, k d It is represented as the differential gain coefficient, and u(t) is represented as the error function, which is the difference between the danger angle and the current angle;

[0049] u(t)=θ danger -θ current

[0050] Where u(t) represents the error function, θ danger Represented as the danger angle, i.e., the critical value for triggering emergency braking; θ current This is represented as the current angle, i.e., the actual angle value of the device;

[0051] Simultaneously, the audible and visual alarms are activated, broadcasting a coordinated avoidance signal to adjacent devices, and a laser projection warning area is displayed via AR-HUD. The actuator employs a dual-redundancy design; when the main control system fails, the backup PLC immediately takes over control.

[0052] At Level 2 risk, preventative intervention is initiated. The cab tactile feedback device generates a vibration pattern at a specific frequency, simultaneously displaying a yellow buffer envelope on the AR interface. Voice prompts utilize natural language generation technology, outputting specific guidance such as "Equipment is approaching from the right rear; a 15° left turn is recommended." The execution unit limits the operating speed in the dangerous direction; the specific calculation method for its speed constraint function is as follows:

[0053]

[0054] Among them, v limit This represents the restricted operating speed, v. max This is expressed as the maximum permissible speed. It is represented as an exponential decay function, which is used to calculate the speed limit factor. t0 represents the reference time point, and TTC0 represents the danger time threshold.

[0055] When a Level 3 risk is detected, a management-level response is triggered. The command center's large screen automatically displays a live video stream and a risk parameter trend chart. The location of the risk source is marked in the BIM model, and an electronic work order containing mitigation suggestions is generated and pushed to the management personnel's terminal. Simultaneously, the risk event is recorded in the blockchain evidence storage unit. For foundation risks, a pressure redistribution algorithm is initiated. The specific calculation method is as follows:

[0056]

[0057] Wherein, ΔP b This represents the pressure adjustment amount for the b-th region, where M represents the proportionality coefficient. Expressed as the overall stress change of the foundation, A c Let A represent the area of ​​the c-th region. b Let d represent the area of ​​the b-th region. b Represented as the depth factor of the b-th region;

[0058] The status of each actuator is verified every 30 seconds, and reliability is ensured through a heartbeat mechanism; all intervention operations are logged in detail, including trigger time, risk type, and execution effect parameters.

[0059] Digital Twin Platform Module: Based on BIM model and real-time sensor data, it constructs a three-dimensional operation scene to perform hoisting path pre-simulation, load swing trajectory prediction and foundation bearing capacity visualization;

[0060] In the digital twin platform module, a high-fidelity dynamic virtual operation scene is constructed by deeply integrating BIM models with real-time multi-source sensor data. First, the engineering BIM model is imported as the basic scene framework. Key structural features are preserved through lightweight processing, and a two-way data channel with the physical equipment is established. Real-time data fusion employs a spatiotemporal registration algorithm to map UWB positioning data, IMU attitude data, and visual recognition results into the virtual scene. For hoisting path planning, the platform integrates a kinematics solver, generating a collision-free path based on an improved RRT* algorithm. Its cost function considers equipment motion constraints, and the specific calculation method is as follows:

[0061]

[0062] Where J1 represents the cost function of the path, L path Represented as path length, Let θ0(t) represent the minimum safe distance between the path and the i-th obstacle, θ0(t) represent the motion smoothness, and w1, w2, and w3 represent the weighting factors.

[0063] The load swing prediction uses a Lagrange dynamics model, and the specific calculation method is as follows:

[0064]

[0065] Where β2 represents the angular acceleration of the load swing, g represents the gravitational acceleration, l represents the distance from the hook to the load, β represents the angle of the load swing, β1 represents the angular velocity of the load swing, and a x Let a be the acceleration of the hook in the x-direction. y Let represent the acceleration of the hook in the y direction, m represent the load mass, and c represent the damping coefficient.

[0066] By calculating the hook acceleration a in real time x a yGiven the angle β of the load swing, predict the swing trajectory in the next 5 seconds and visualize it in the form of a probability cloud map in a 3D scene;

[0067] The foundation analysis sub-unit couples pressure sensor data with the geological BIM model and uses the finite element method to calculate the soil stress distribution in real time. The specific calculation method is as follows:

[0068]

[0069] Where σ(x,y) represents the soil stress distribution function, D represents the stress concentration factor, and P a The measured value of the a-th pressure sensor, r a The distance from the a-th pressure sensor to the calculation point, γ represents the attenuation coefficient;

[0070] When a stress concentration area is detected to exceed the threshold, an early warning is automatically triggered and an outrigger adjustment suggestion is generated; a multi-view visualization interface is provided, supporting the command center to conduct immersive scene inspections through VR devices, and key risk areas are displayed using heat maps overlaid.

[0071] Human-computer interaction module: including driver AR-HUD interface and command center 3D situation dashboard, providing risk source highlighting, recommended avoidance routes and global operation heat map.

[0072] In the human-computer interaction module, at the driver's terminal, the AR-HUD system uses waveguide projection technology to overlay key information at a virtual image distance of 7 meters in front of the driver's field of vision. Spatial registration technology precisely aligns the virtual information with the real scene, achieving a positioning error of less than 0.3°. It receives risk prediction data from the graded warning and execution module in real time. When a potential collision is detected, a three-level visual warning strategy is adopted: Level I risk triggers a red pulse border warning, simultaneously projecting a 3D arrow on the windshield indicating the emergency avoidance direction; Level II risk displays a yellow semi-transparent envelope outline, overlaid with a green guide line for the recommended path. Path planning considers the boom's kinematic constraints, and the calculation method is as follows:

[0073]

[0074] Where Γ0 represents the optimal path, ∫ δ ds represents the total path length, and λ represents the weighting factor. The maximum value of the boom angular acceleration is represented by μ, which is the boom kinematic constraint; μ represents the boom rotation angle; and T1 represents the total time for path planning.

[0075] The 3D situation dashboard deployed in the command center is built on WebGL technology, supporting free switching between multiple perspectives and dynamic adjustment of detail levels. The dashboard integrates the following intelligent visualization components: a group operation heatmap, which calculates risk distribution using a kernel density estimation method. The specific calculation method is as follows:

[0076]

[0077] Where f(x,y) represents the risk distribution function of the group operation heatmap, q represents the number of devices in the group operation, h represents the bandwidth parameter, and ||(x,y)-(x m ,y m || represents the Euclidean distance, and K(·) represents the kernel function;

[0078] The equipment status matrix displays key parameters such as load rate and stability coefficient of each crane in real time;

[0079] The predictive situation simulation window simulates changes in the operational scenario over the next 30 seconds based on digital twin data; it supports multimodal interaction, allowing commanders to quickly access millimeter-wave radar point cloud data or ground stress distribution maps of areas of interest via gesture recognition and voice commands; all early warning information adopts a tiered push mechanism, with Level I alarms simultaneously triggering AR-HUD, situation dashboards, and multi-terminal vibration alerts to ensure zero omission of critical information; historical operational data automatically generates visual logs, supporting backtracking and analysis of the operational process via a timeline.

[0080] Please see Figure 2 As shown in this embodiment, it should be specifically explained that the present invention provides a safety monitoring method for mobile crane group operations based on multi-source perception and collaborative decision-making, including the following steps:

[0081] Step A1: Collect real-time operational environment and status data;

[0082] Step A2: Process sensor data in real time, generate dynamic safety envelopes, and predict group conflict risks;

[0083] Step A3: Trigger different response measures based on the risk level;

[0084] Step A4: Based on the BIM model and real-time sensor data, construct a three-dimensional operation scene to perform hoisting path pre-simulation, load swing trajectory prediction, and foundation bearing capacity visualization;

[0085] Step A5: Through the driver's AR-HUD interface and the command center's 3D situation dashboard, provide risk source highlighting, recommended avoidance routes, and a global operation heat map.

[0086] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A safety monitoring system for group operations of truck cranes based on multi-source perception and collaborative decision-making, characterized in that, include: Distributed sensing module: used to collect real-time operational environment data and status data; Edge computing module: used to process sensor data in real time, generate dynamic safety envelopes and predict group conflict risks; Tiered early warning and execution module: Triggers different response measures based on the risk level; Digital Twin Platform Module: Based on BIM model and real-time sensor data, it constructs a three-dimensional operation scene to perform hoisting path pre-simulation, load swing trajectory prediction and foundation bearing capacity visualization; Human-computer interaction module: including driver AR-HUD interface and command center 3D situation dashboard, providing risk source highlighting, recommended avoidance routes and global operation heat map.

2. The safety monitoring system for truck crane group operations based on multi-source perception and collaborative decision-making as described in claim 1, characterized in that: The system achieves comprehensive perception of the working environment and equipment status through multi-sensor data fusion; the millimeter-wave radar detects obstacles within a 50-meter radius at a scanning frequency of 20Hz, identifies the relative positions of other equipment, personnel and temporary structures, and the binocular vision sensor installed at the end of the boom tracks the hook's movement trajectory in real time based on deep learning algorithms. It calculates the swing amplitude and direction of the load through stereo matching technology, and detects potential collision risks in the surrounding area by combining feature point recognition technology. A high-precision IMU inertial unit continuously monitors the vehicle's attitude changes at a sampling frequency of 200Hz. Combined with real-time boom length and elevation angle data provided by the boom angle encoder, a kinematic model of the equipment is constructed.

3. The safety monitoring system for truck crane group operations based on multi-source perception and collaborative decision-making as described in claim 1, characterized in that: In the edge computing module, multi-source sensing data is processed in real time through a distributed architecture to construct a three-dimensional safety protection system; the dynamic safety envelope generator adopts an adaptive kinematic model to calculate the safe working space based on the real-time collected boom parameters and environmental data. The specific calculation method is as follows: Among them, V safe (t) represents the safety value at time t, L(t) represents the real-time boom length, θ(t) represents the current elevation angle, ω(t) represents the slewing angular velocity, and Q(t) represents the load weight. max The maximum lifting weight is represented by ε(·), which represents the environmental correction term. wind Expressed as the combined wind speed, α ground This is expressed as the inclination of the foundation.

4. The safety monitoring system for truck crane group operations based on multi-source perception and collaborative decision-making according to claim 1, characterized in that: The graded early warning and execution module adopts a multi-level response mechanism and uses intelligent decision-making algorithms to achieve closed-loop control from risk early warning to automatic intervention. Based on real-time risk assessment results, the operational risks are divided into three levels and differentiated response strategies are triggered. At Level 1 risk, a linkage response is triggered, sending an emergency braking command to the hydraulic control system via the CAN bus. The control logic is as follows: Among them, F brake The braking force, k, represents the braking force during emergency braking. p Represented as the proportional gain coefficient, k d Let t be the differential gain coefficient, and u(t) be the error function. When the risk level is 2, preventative intervention is initiated, and the cab tactile feedback device generates a vibration pattern at a specific frequency; The execution unit limits the operating speed in dangerous directions, and the specific method for calculating its speed constraint function is as follows: Among them, v limit This represents the restricted operating speed, v. max This is expressed as the maximum permissible speed. It is represented as an exponential decay function, which is used to calculate the speed limit factor. t0 represents the reference time point, and TTC0 represents the danger time threshold.

5. A safety monitoring system for truck crane group operations based on multi-source perception and collaborative decision-making as described in claim 4, characterized in that: When the risk level is 2, preventative intervention is initiated, and the cab tactile feedback device generates a vibration pattern at a specific frequency; The execution unit limits the operating speed in dangerous directions, and the specific method for calculating its speed constraint function is as follows: Among them, v limit This represents the restricted operating speed, v. max This is expressed as the maximum permissible speed. It is represented as an exponential decay function, which is used to calculate the speed limit factor. t0 represents the reference time point, and TTC0 represents the danger time threshold.

6. A safety monitoring system for truck crane group operations based on multi-source perception and collaborative decision-making as described in claim 4, characterized in that: When a Level 3 risk is detected, a management-level response is triggered. The command center's large screen automatically displays a live video stream and a risk parameter trend chart. The location of the risk source is marked in the BIM model, and an electronic work order containing mitigation suggestions is generated and pushed to the management personnel's terminal. Simultaneously, the risk event is recorded in the blockchain evidence storage unit. For foundation risks, a pressure redistribution algorithm is initiated. The specific calculation method is as follows: Wherein, ΔP b Let A represent the pressure adjustment in the b-th region, M represent the proportionality coefficient, Δφ represent the overall stress change in the foundation, and A represent the overall stress change in the foundation. c Let A represent the area of ​​the c-th region. b Let d represent the area of ​​the b-th region. b It is represented as the depth factor of the b-th region.

7. A safety monitoring system for truck crane group operations based on multi-source perception and collaborative decision-making as described in claim 1, characterized in that: In the digital twin platform module, a high-fidelity dynamic virtual operation scene is constructed by deeply integrating BIM models with real-time multi-source sensor data. First, the engineering BIM model is imported as the basic scene framework. Key structural features are preserved through lightweight processing, and a two-way data channel with the physical equipment is established. Real-time data fusion employs a spatiotemporal registration algorithm to map UWB positioning data, IMU attitude data, and visual recognition results into the virtual scene. For hoisting path planning, the platform integrates a kinematics solver, generating a collision-free path based on an improved RRT* algorithm. Its cost function considers equipment motion constraints, and the specific calculation method is as follows: Where J1 represents the cost function of the path, L path Represented as path length, Let θ0(t) represent the minimum safe distance between the path and the i-th obstacle, θ0(t) represent the motion smoothness, and w1, w2, and w3 represent the weighting factors. The load swing prediction adopts the Lagrange dynamic model. The foundation analysis sub-unit couples the pressure sensor data with the geological BIM model and uses the finite element method to calculate the soil stress distribution in real time.

8. A safety monitoring system for truck crane group operations based on multi-source perception and collaborative decision-making as described in claim 1, characterized in that: In the human-machine interaction module, at the driver's terminal, the AR-HUD system uses optical waveguide projection technology to overlay key information at a virtual image distance of 7 meters in front of the driver's field of vision. It receives risk prediction data from the graded warning and execution module in real time. When a potential collision is detected, a three-level visual warning strategy is adopted, and a green guide line of the recommended path is overlaid. The path planning takes into account the kinematic constraints of the boom. The specific calculation method is as follows: Where Γ0 represents the optimal path, ∫ δ ds represents the total path length, and λ represents the weighting factor. The maximum value of the boom angular acceleration is represented by μ, which is the boom kinematic constraint; μ represents the boom rotation angle; and T1 represents the total time for path planning.

9. A method for safety monitoring of mobile crane group operations based on multi-source perception and collaborative decision-making, using a mobile crane group operation safety monitoring system based on multi-source perception and collaborative decision-making as described in any one of claims 1-8, characterized in that: Step A1: Collect real-time operational environment and status data; Step A2: Process sensor data in real time, generate dynamic safety envelopes, and predict group conflict risks; Step A3: Trigger different response measures based on the risk level; Step A4: Based on the BIM model and real-time sensor data, construct a three-dimensional operation scene to perform hoisting path pre-simulation, load swing trajectory prediction, and foundation bearing capacity visualization; Step A5: Through the driver's AR-HUD interface and the command center's 3D situation dashboard, provide risk source highlighting, recommended avoidance routes, and a global operation heat map.

Images