Intelligent safety monitoring and anti-collision system for tower crane and elevator

Abstract

The invention provides an intelligent safety monitoring and anti-collision system for a tower crane and an elevator, and relates to the technical field of building construction safety monitoring. The system constructs a five-layer architecture of multi-source sensing, edge computing, converged communication, a cloud platform and application interaction, a tower crane end is integrated with a torque sensor, a rotary encoder, a wind speed sensor, a height encoder and a GNSS / RTK dual-mode positioning module, an elevator end is provided with a load sensor, a double-shaft inclination sensor and a network camera, and data intercommunication is achieved through a communication network. A distributed space-time cooperation mechanism is utilized, an LSTM neural network based on an attention mechanism is adopted to predict a device track, a dynamic collision cone model introducing wind speed disturbance compensation is constructed, and an improved ORCA algorithm introducing jerk constraint is operated to realize optimal avoidance of a group tower. According to the system, the BIM three-dimensional geometric model and real-time working condition data are fused to construct the digital twinborn body, acousto-optic broadcast linkage alarm is supported, and the system is suitable for intelligent safety management of large capital construction projects.

Description

Technical Field

[0001] This invention relates to the field of construction safety monitoring technology, specifically to an intelligent safety monitoring and collision avoidance system for tower cranes and hoists. Background Technology

[0002] With the rapid development of large-scale infrastructure projects, tower cranes and hoists, as core construction equipment, are increasingly frequently used in collaborative operations. However, due to the complex operating trajectories of the equipment, limited working space, and variable environmental interference, collision accidents are prone to occur during cross-operation, resulting in equipment damage, personnel injuries and fatalities, and project delays.

[0003] Existing technologies suffer from the following drawbacks: Most rely on GNSS / RTK dual-mode positioning devices, triggering alarms based on preset fixed safety distance thresholds. This approach fails to consider the dynamic motion characteristics of the devices and environmental disturbances, resulting in low collision prediction accuracy, frequent false alarms or missed alarms, and an inability to adapt to dynamic changes in real-world working conditions. Some systems upload device location data to cloud servers for collision risk analysis and command issuance, but this architecture suffers from significant response delays, making timely intervention difficult in emergency avoidance scenarios. Furthermore, high network transmission dependence means data loss is likely in areas with weak or congested signals at construction sites, failing to meet the real-time requirements of safety responses. In common construction site environments such as strong winds, rain, and fog, the accuracy of traditional positioning devices decreases, leading to distorted collision predictions.

[0004] Therefore, there is an urgent need for an intelligent safety solution that can achieve real-time prediction of dynamic collision risks, layered real-time response, adaptation to complex environments, and collaborative monitoring across all scenarios, in order to fill the technological gap in the industry. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an intelligent safety monitoring and anti-collision system for tower cranes and hoists, which solves the problem of collaborative safety management in the scenario of tower crane groups and hoists working together on construction sites, and maximizes operation efficiency while ensuring absolute safety.

[0006] The technical solution adopted in this invention is to provide an intelligent safety monitoring and anti-collision system for tower cranes and elevators, including a multi-source heterogeneous sensor network, an edge gateway, a cloud decision platform, and a BIM model server;

[0007] Sensor networks are deployed on the tower crane and elevator body to collect operating parameters and spatial location data, which are then transmitted to the edge gateway via the network. The edge gateway has a built-in dynamic collision cone modeling module and an improved ORCA (Optimal Reciprocal Collision Avoidance) obstacle avoidance algorithm that introduces jerk constraints. The edge gateway calculates the collision risk between devices in real time and directly triggers local braking when the risk value exceeds the emergency threshold. The cloud platform runs an LSTM trajectory prediction model and a dual digital twin engine to generate a global risk heatmap and optimize the edge gateway model parameters; The system employs a hierarchical collaborative decision-making mechanism to establish an early warning response and continuous learning loop.

[0008] Preferably, a dynamic collision cone modeling and coordinate transformation algorithm is adopted to abstract each tower crane into a moving rigid body model, calculate the relative collision cone for any two tower cranes, and introduce a wind speed disturbance compensation mechanism to improve the prediction robustness.

[0009] Preferably, the improved ORCA obstacle avoidance algorithm running on the edge gateway achieves speed smoothing by introducing jerk constraints; Construct a quadratic programming objective function that includes an efficiency term and a smoothing term, where the efficiency term is used to minimize the deviation between the computation speed and the preferred speed, and the smoothing term is used to penalize the rate of change of acceleration; Solve the quadratic programming objective function and output the optimal avoidance speed command that satisfies the physical characteristics of the motor, so as to eliminate the mechanical stress on the tower crane structure and the current impact on the drive motor caused by sudden speed changes.

[0010] Preferably, a dual digital twin approach is adopted. The first layer is a BIM-based digital twin of equipment, and the second layer is a digital twin of personnel integrating multi-dimensional data such as identity information and qualification certificates. The equipment digital twin includes component attributes, kinematic pair constraints, and mechanical parameters, and can map working condition parameters in real time and generate risk heat maps by predicting the future movement trajectory of the equipment. The personnel digital twin achieves position updates through UWB tags, GNSS positioning, and inertial navigation. It uses the Extended Kalman Filter (EKF) algorithm to fuse UWB, GNSS, and inertial navigation data, and maps the personnel positions to the BIM model coordinate system through coordinate transformation.

[0011] Preferably, a distributed state broadcast mechanism is used to ensure communication between tower cranes. When local computing resources are insufficient, a state snapshot is sent to the cloud for offloading processing. The cloud generates a future trajectory heatmap for each tower crane using an LSTM trajectory prediction model, and sends parameter updates to correct the edge gateway model when the predicted trajectory deviates significantly from the current edge gateway calculation result.

[0012] Preferably, its characteristic is a hierarchical system decision-making mechanism, including: The first-level edge-end emergency braking decision-making directly triggers local braking when the risk value exceeds the emergency threshold; The second layer edge obstacle avoidance speed planning and decision-making is performed by running an improved ORCA algorithm to solve for the optimal avoidance speed. The third layer of cloud-based model parameters are optimized and decisions are made to generate a global risk heatmap; The first-level decision has the highest priority and cannot be overridden or interrupted by the second or third-level decisions.

[0013] Preferably, the system performs a continuous learning closed loop, including edge sample collection, cloud training, model distribution, and edge optimization processes. The system dynamically adjusts model parameters based on real-time operating conditions and rationally allocates computing resources between the edge gateway and the cloud.

[0014] Beneficial effects: 1. Achieve accurate prediction and avoidance of dynamic collision risks. By constructing a dynamic collision cone model and introducing a wind speed disturbance compensation mechanism, the system can analyze the relative motion trajectory of the tower crane and the elevator and environmental interference factors in real time, automatically generate collision risk assessment results, and dynamically plan the optimal avoidance path based on the improved ORCA algorithm, fundamentally avoiding the risk of spatial and temporal intersection of equipment in collaborative operations, and solving the shortcomings of traditional fixed safety distance schemes that cannot adapt to dynamic working conditions.

[0015] 2. To ensure safe response and localized decision-making, a real-time collision risk calculation module is deployed on the edge gateway. When a high-risk state is detected, the system directly triggers a local braking command, significantly shortening the safety response time. This design avoids response delays caused by network latency, ensuring the immediacy of safe operations in high-risk scenarios.

[0016] 3. Enhance system robustness in complex environments. Through wind speed disturbance compensation mechanism and dual digital twin technology, the system can synchronously map equipment operating status, personnel location and environmental parameters, so that safety monitoring can maintain high reliability under severe weather conditions such as strong winds. This solves the industry pain point of high misjudgment rate and high missed judgment rate of existing technology in complex construction site environments.

[0017] 4. Construct a collaborative safety monitoring system covering the entire scenario, integrating BIM models and IoT sensing data. The system realizes dynamic monitoring of "equipment-personnel-environment" in a three-in-one manner: the digital twin of equipment simulates the movement trajectory in real time to generate a risk heat map, and the digital twin of personnel accurately identifies violations through multi-source positioning technology, expanding the scope of safety monitoring from a single device to the entire construction scenario, and solving the problem of fragmented collaborative safety information across devices. Attached Figure Description

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a diagram of the tower crane safety monitoring system architecture of the present invention; Figure 2This is the architecture diagram of the elevator safety monitoring subsystem of the present invention; Figure 3 This is the architecture diagram of the hook video safety auxiliary subsystem of the present invention; Figure 4 This is a flowchart of the algorithm for dynamic collision cone modeling and coordinate transformation of the present invention; Figure 5 This is a flowchart of the dual digital twin method of the present invention; Figure 6 This is a flowchart of the edge gateway and cloud decision collaboration method of the present invention. Detailed Implementation

[0019] To better understand the purpose, system architecture, and functional implementation of this embodiment, the embodiments and features described herein can be combined with each other without conflict. The exemplary embodiments disclosed herein will be described below with reference to the accompanying drawings, including specific technical details disclosed to aid understanding; however, these details should be considered exemplary rather than restrictive. Therefore, those skilled in the art should understand that various improvements and adjustments can be made to the embodiments described herein without departing from the scope and core ideas of the invention. Similarly, for clarity, detailed descriptions of well-known technologies, functions, and structures are omitted in the following description.

[0020] Example 1 Figure 1 This is the architecture diagram of the tower crane safety monitoring system of the present invention.

[0021] Figure 2 This is the architecture diagram of the elevator safety monitoring subsystem of the present invention.

[0022] Figure 3 This is the architecture diagram of the hook video safety auxiliary subsystem of the present invention.

[0023] like Figure 1 As shown, a multi-source heterogeneous sensor network is deployed to collect equipment operating conditions, positioning data, and behavioral data. In terms of hardware configuration, each tower crane is equipped with torque sensors, slewing sensors, wind speed sensors, height sensors, and amplitude sensors to ensure that the position, attitude, motion parameters, and environmental parameters of the equipment can be acquired and transmitted to the cloud. The tower cranes communicate in real time with each other via a wireless sensor network. Figure 2 As shown, for the elevator, a combined GNSS and inertial navigation positioning system is used, and UWB positioning tags are provided for the operators to achieve sub-meter level positioning. After data acquisition, the raw data needs to be classified and processed, including time synchronization, coordinate transformation, outlier filtering, and data fusion. For example... Figure 3As shown, the tower crane hook video subsystem uses precision sensors to collect real-time data on hook height and trolley amplitude. After calculation, it obtains the angle and distance parameters between the hook and the camera. Based on this, it controls the camera lens tilt angle and magnification in real time, ensuring a clear video image of the load below the hook is displayed on the monitor in the tower crane operator's cab. This guides the operator's lifting operations and significantly improves safety. The video images are stored on the hard drive built into the video recorder for easy accident cause location. They can also be transmitted via 5G network to the ground project site and remote monitoring platform to build a complete tower crane video monitoring platform.

[0024] The multi-source data fusion center comprehensively processes data from various sensors, combining data from torque sensors, rotation sensors, wind speed sensors, height sensors, and amplitude sensors to form a comprehensive and accurate description of the equipment's status. The dynamic collision cone model abstracts the tower crane as a moving rigid body and considers its geometry, kinematic characteristics, and environmental disturbances to establish an accurate collision risk model. This model not only considers factors such as position, velocity, acceleration, jerk, and attitude, but also specifically introduces a wind speed disturbance compensation mechanism, improving the robustness of the prediction.

[0025] Example 2 Figure 4 This is a flowchart of the dynamic collision cone modeling and coordinate transformation algorithm of the present invention.

[0026] like Figure 4 As shown, the tower crane is abstracted as a multi-link rigid body system, whose motion state is described by position, velocity, acceleration, jerk, and attitude parameters. Position is represented by a three-dimensional vector, and attitude is represented by a unit quaternion to ensure the numerical stability of rotational operations. The kinematic equations describe the changes of position, velocity, acceleration, jerk, and attitude over time, where linear velocity is the time derivative of position, and angular velocity is related to the change of attitude quaternion. To achieve precise control, the system establishes coordinate transformation equations from the global coordinate system to the device coordinate system. Global coordinates are converted to local device coordinates through rotation matrices and translation vectors, facilitating subsequent collision detection and obstacle avoidance calculations.

[0027] The safety distance constraint is a key indicator for judging the collision risk. The collision risk function is obtained by calculating the difference between the actual distance between two devices and the preset safety distance, as shown in equation (1) below.

[0028] (1) in, This is the minimum safe distance between the two tower cranes. This represents the relative positions of the two tower cranes. This represents the distance between the two tower cranes. When the risk function is greater than zero, the system determines that there is a collision risk and needs to initiate an obstacle avoidance procedure.

[0029] Wind speed has a significant impact on the motion of the tower crane, and the system quantifies this impact through a wind force calculation model. The wind force is proportional to the air density, drag coefficient, windward area, and the square of the wind speed, and its direction is consistent with the wind direction. The calculated wind force is divided by the equivalent mass of the tower crane to obtain the wind-induced acceleration. When constructing the compensated collision cone, the system incorporates this acceleration into the motion equation to correct the predicted trajectory, as shown in equation (2) below.

[0030] (2) in, This represents the relative positions of the two tower cranes. Let be the relative speed between the two tower cranes. This compensation mechanism, which compensates for wind-induced acceleration, significantly improves the accuracy of collision prediction under adverse weather conditions, enabling the system to adapt to various complex environmental conditions and avoid misjudgments or omissions caused by wind factors.

[0031] This invention employs an improved ORCA obstacle avoidance algorithm for linear programming. The improved ORCA algorithm, based on the standard ORCA, incorporates an efficiency term. With smoothing terms The composite objective function is shown in equation (3) below.

[0032] (3) in, The velocity vector to be optimized The preferred speed is determined by the current mission objective. The current speed is obtained by encoding and data acquisition calculation. The speed from the previous moment is stored in the edge gateway's memory. This is a smoothing coefficient, with a value ranging from 0.1 to 0.5, which is dynamically adjusted according to the mechanical characteristics of the tower crane.

[0033] Define the jerk as the derivative of acceleration with respect to time, and its discrete form is shown in the following equation: (4) Introduce a motor jerk constraint, as shown in equation (5): (5) in, To control the cycle, This is the maximum permissible acceleration of the motor, set according to the inverter parameters, and typically ranges from 0.5 to 2.0 m / s². 3 The smoothing term optimizes the smoothness of the velocity curve, while the jerk constraint ensures physical realizability.

[0034] The corresponding quadratic programming problem is shown in equation (6): (6) The interior point method is used to solve the above quadratic programming problem. The iteration termination condition is that the duality gap is less than 10. -6 If the number of iterations exceeds 100, the optimal speed obtained is the target speed for the tower crane's next control cycle. The optimal speed is decomposed into slewing speed and luffing speed components, which are sent to the tower crane's frequency converter via the CAN bus to drive the motor. During speed switching, there are no sudden changes in actual acceleration, thus avoiding mechanical shock.

[0035] Multiple tower cranes need to share status information for collaborative obstacle avoidance. The system employs a distributed status broadcasting mechanism, where each tower crane periodically broadcasts its own status, including position, velocity, acceleration, and attitude, while simultaneously receiving status information from other tower cranes. To handle communication latency, the system uses timestamps and status prediction techniques to estimate the actual status of other devices at the current moment. A distributed consensus protocol ensures that all devices have a consistent understanding of the system state, maintaining system stability even in the event of partial communication failures. The communication reliability model considers distance and environmental factors, adaptively adjusting transmission power and retransmission strategies to guarantee the reliable transmission of critical safety information. The entire communication system design meets actual latency requirements, ensuring real-time obstacle avoidance decisions.

[0036] Example 3 Figure 5 This is a flowchart of the dual digital twin method of the present invention.

[0037] like Figure 5 As shown, in the dual digital twin parallel construction framework, the edge gateway is deployed in the tower crane cab and the elevator control cabinet to be responsible for real-time data processing, while the cloud platform runs on the hyperconverged server cluster in the temporary computer room to be responsible for global optimization and model training.

[0038] The left-hand equipment digital twin branch uses Revit to create LOD300 precision models of tower cranes, hoists, and hooks. Component attributes include engineering attributes such as material density, moment of inertia, motor parameters, and reduction ratio. The lightweight BIM model is distributed to the edge gateway via a 5G network for local caching, and can still be loaded during network outages.

[0039] Real-time parameters of the equipment are collected by sensors and linked to the corresponding components in the BIM model. In tower cranes, torque sensors collect real-time torque, encoders collect slewing angle, tilt sensors collect amplitude angle, and wind speed sensors collect operating wind speed. For hoists, load cells collect car load, and tilt sensors collect body tilt angle; these real-time parameters are then written into the component attribute library of the BIM model.

[0040] The motion trajectory of the device in the next 10 seconds is predicted by the LSTM neural network, as shown in equation (7).

[0041] (7) in, This refers to the length of the tower crane boom. For the amplitude angle, For the turning angle, This is the horizontal projection length of the end of the crane boom. The horizontal component is perpendicular to the crane boom. This represents the vertical height component of the crane boom tip, while the future trajectory is a continuous set of crane boom tip positions at multiple future moments. LSTM is used to predict the amplitude angle at a future moment based on historical operating data. and turning angle Substitute the predicted amplitude and slewing angles into equation (7) to obtain the coordinates of the corresponding future time. Connect the end coordinates of different future times in chronological order to form the future motion trajectory of the crane boom end.

[0042] Based on the future trajectory projection results, the collision probability of each region of the device is calculated and visualized, as shown in the following formula.

[0043] (8) in, This represents the real-time distance between devices. The system sets a safety distance threshold for equipment. Higher calculated probabilities are indicated by a darker color on the heatmap, suggesting a higher risk in the area the trajectory traverses. If the trajectory enters a dangerous area, the system will trigger an early warning to prevent collisions. The system uses the separating axis theorem to determine if the OBBs of the equipment's BIM model overlap. Each equipment's OBB consists of a center, a positive half-axis, and a half-length. The system calculates the OBB projection range for any separating axis. If the projection ranges of all separating axes overlap, a spatial overlap risk is identified, triggering an audible and visual alarm for the tower crane or elevator, and simultaneously pushing a warning to the cloud. If no risk is detected, the system continues to monitor equipment operating conditions for risk.

[0044] The right-hand digital twin branch integrates multi-dimensional data of construction personnel and constructs digital profiles. It uses multi-source positioning technology of UWB+GNSS+inertial navigation to update personnel pose, as shown in the following formula.

[0045] (9) in, This is a local location captured by a UWB tag. The global location collected by the GNSS / RTK dual-mode positioning module. The location of occluded scenes is filled in for inertial navigation. , and These are the weighting coefficients. The Extended Kalman Filter (EKF) algorithm is used to fuse UWB, GNSS, and inertial navigation data. Through coordinate transformation, the personnel positions are mapped to the BIM model coordinate system, achieving accurate spatial matching by unifying the coordinates of the personnel position point cloud with the BIM model.

[0046] Example 4 Figure 6 This is a flowchart of the edge gateway and cloud decision collaboration method of the present invention.

[0047] like Figure 6 As shown, the system performs real-time data acquisition and preliminary processing from sensors at the edge gateway. It utilizes a collision detection algorithm for rapid response. In the left branch, when assessing collision risk, it performs rapid collision detection using a directional bounding box, calculates the minimum distance between devices, compares it to a preset safety threshold, and analyzes the movement direction and speed trends of the devices. When a collision risk exists, the system issues correction parameters and updates the edge gateway's prediction model to improve its accuracy. If the risk is low or effectively controlled, the existing model parameters are maintained. In the right branch detection, when insufficient edge gateway computing resources are detected, a snapshot of the device status is uploaded to the cloud. An LSTM model is run in the cloud to generate accurate trajectory predictions and risk heatmaps. The deviation between the cloud prediction results and the edge gateway prediction results is compared to evaluate model performance. Based on the deviation analysis results, correction parameters are issued to update the edge gateway model. Through collaborative learning between the edge and cloud, the edge gateway performs real-time monitoring and preliminary judgment, deciding whether cloud support is needed based on the actual situation. The cloud provides more accurate prediction results, and the edge gateway model is adjusted based on the deviation, continuously optimizing model performance and improving prediction accuracy. Ultimately, the system dynamically adjusts model parameters based on real-time operating conditions, rationally allocates computing resources between edge gateways and the cloud, and continuously improves the system's predictive capabilities and response speed through continuous learning, ensuring effective prevention of safety incidents in various complex environments.

[0048] This invention proposes a hierarchical collaborative decision-making mechanism. The first layer is an edge-end emergency braking decision layer, which triggers emergency braking when any condition is met, including: When collision time Triggered in seconds, as shown in the following formula: (10) When relative distance meters and relative speed trigger; When the collision risk function And relative speed Triggered, in For minimum safe distance, , For the equivalent radius of the tower crane, This is for buffer distance.

[0049] When any of the above emergency threshold conditions are met, the edge gateway immediately generates a braking command. This braking command is a hardware-level emergency stop signal, directly output to the tower crane inverter. The actions performed include cutting off the motor power supply, activating the mechanical brake, and recording the braking moment's operating data to local storage. The generation and output of the braking command are entirely independent of the edge gateway, without going through the cloud decision-making platform. The first-level decision has the highest priority and enters a locked state upon triggering. In the locked state, the second-level decision is prohibited from modifying the speed command, the third-level decision is prohibited from issuing any control commands, and the lock can only be released by a manual emergency reset signal. The non-overriding mechanism is implemented through a hardware interlock circuit within the edge gateway. When the first-level decision is triggered, a hardware interrupt signal is sent to the second and third-level decision modules, forcibly terminating their calculation process. Simultaneously, a high-priority braking signal is sent to the inverter to ensure the priority execution of the braking command.

[0050] The second-layer decision-making is deployed at the edge gateway and is responsible for obstacle avoidance speed planning in medium-risk scenarios. It is activated when the first-layer decision-making fails to trigger and a potential collision risk is detected. The criterion for determining a potential collision risk is 2s ≤ <5s or 5m≤ For distances less than 10m, the edge gateway runs an improved ORCA obstacle avoidance algorithm to solve for the optimal obstacle avoidance speed. This optimal speed is decomposed into slewing speed and luffing speed, and sent to the tower crane inverter via the CAN bus as a speed limit command. The second-level decision continuously monitors the triggering status of the first-level decision. If the first-level decision is triggered, the second-level decision immediately terminates its current calculation, releases computing resources, and confirms the execution status of the braking command.

[0051] The third-layer decision-making is deployed on a cloud-based decision-making platform, responsible for non-real-time global optimization and model updates. The cloud platform continuously monitors the operational status of the edge gateways and initiates the third-layer decision-making when the following conditions are met: edge gateway CPU utilization < 80%; no first-layer decision trigger; no second-layer decision running for more than 5 seconds. A large-parameter LSTM trajectory prediction model runs in the cloud, generating a global risk heatmap covering all work areas of the construction site. The third-layer decision-making includes a continuous learning loop encompassing edge sample collection, cloud training, model deployment, and edge optimization.

[0052] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0053] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. An intelligent safety monitoring and collision avoidance system for tower cranes and hoists, characterized in that, This includes multi-source heterogeneous sensor networks, edge gateways, cloud-based decision platforms, and BIM model servers; Sensor networks are deployed on the tower crane and elevator body to collect operating parameters and spatial location data, which are then transmitted to the edge gateway via the network. The edge gateway has a built-in dynamic collision cone modeling module and an improved ORCA obstacle avoidance algorithm that introduces jerk constraints, enabling the edge gateway to calculate the collision risk between devices in real time. The cloud platform runs an LSTM trajectory prediction model and a dual digital twin engine to generate a global risk heatmap and optimize the edge gateway model parameters; The system achieves early warning response through a hierarchical collaborative decision-making mechanism, which includes emergency braking decision at the edge, obstacle avoidance speed planning decision at the edge, and cloud-based model parameter optimization decision.

2. The system according to claim 1, characterized in that, A dynamic collision cone modeling and coordinate transformation algorithm is adopted to abstract each tower crane into a moving rigid body model. The relative collision cone is calculated for any two tower cranes, and a wind speed disturbance compensation mechanism is introduced to improve the prediction robustness.

3. The system according to claim 1, characterized in that, The improved ORCA obstacle avoidance algorithm running on the edge gateway achieves speed smoothing by introducing jerk constraints; Construct a quadratic programming objective function that includes an efficiency term and a smoothing term, where the efficiency term is used to minimize the deviation between the computation speed and the preferred speed, and the smoothing term is used to penalize the rate of change of acceleration; Solve the quadratic programming objective function and output the optimal avoidance speed command that satisfies the physical characteristics of the motor, so as to eliminate the mechanical stress on the tower crane structure and the current impact on the drive motor caused by sudden speed changes.

4. The system according to claim 1, characterized in that, A dual digital twin approach is proposed, consisting of a first layer of equipment digital twins based on BIM technology and a second layer of personnel digital twins that integrates multi-dimensional data such as identity information and qualification certificates.

5. The system according to claim 4, characterized in that, The digital twin of the equipment includes component attributes, kinematic pair constraints, and mechanical parameters. It can map operating parameters in real time and generate a risk heat map by predicting the future motion trajectory of the equipment.

6. The system according to claim 4, characterized in that, The digital twin of personnel achieves location updates through UWB tags, GNSS positioning, and inertial navigation. It uses the extended Kalman filter (EKF) algorithm to fuse UWB, GNSS, and inertial navigation data, and maps the personnel's position to the BIM model coordinate system through coordinate transformation.

7. The system according to claim 2, characterized in that, A distributed status broadcast mechanism is used to ensure communication between tower cranes. When local computing resources are insufficient, status snapshots are sent to the cloud for offloading processing.

8. The system according to claim 7, characterized in that, The cloud uses an LSTM trajectory prediction model to generate a heat map of the future trajectory for each tower crane, and sends parameters to update and correct the edge gateway model when the predicted trajectory deviates significantly from the current edge gateway calculation results.

9. The system according to claim 1, characterized in that, The hierarchical system decision-making mechanism includes: The first-level edge emergency braking decision-making process directly triggers local braking when the risk value exceeds the set emergency threshold. The second layer edge obstacle avoidance speed planning and decision-making is performed by running an improved ORCA algorithm to solve for the optimal avoidance speed. The third layer of cloud-based model parameters are optimized and decisions are made to generate a global risk heatmap; The first-level decision has the highest priority and cannot be overridden or interrupted by the second or third-level decisions.

10. The system according to claim 9, characterized in that, The third layer of cloud-based model parameter optimization decision-making includes a continuous learning closed loop: edge sample acquisition, cloud training, model deployment, and edge optimization. Model parameters are dynamically adjusted based on real-time operating conditions, and computing resources of edge gateways and the cloud are allocated rationally.

Images