A tower crane cooperative scheduling control method and related equipment

Abstract

The present application relates to a kind of tower crane cooperative scheduling control method and related equipment, comprising: the construction surface is divided into multiple construction units and constructs shared state vector, the procedure state of each unit, space-time resource constraint and equipment accessible state are uniformly represented;Collect equipment real-time operating state and environmental state, synchronization to shared digital model;Based on the shared state vector of update, extract each unit concrete workable time window constraint;With the optimization goal of minimizing equipment waiting time and concrete secondary disturbance cost, combined with time window constraint, jointly timing decision is carried out to feeding and smoothing action, and equipment task queue is generated;Control equipment executes procedure linkage according to queue;When abnormal disturbance event occurs, retrigger decision update queue.The present application realizes the efficient cooperation of feeding and smoothing by constructing unified scheduling dimension, dynamically calculating time window constraint and multi-agent reinforcement learning algorithm, improves construction quality and equipment utilization.

Description

Technical Field

[0001] This application relates to the field of intelligent construction and construction automation technology, and in particular to a collaborative scheduling and control method and related equipment for tower cranes. Background Technology

[0002] Tower cranes are crucial equipment for the vertical and horizontal transport of materials on construction sites, while ground leveling robots are used to level the construction surface after concrete pouring. With the development of intelligent construction technology, the remote control, automatic operation, path planning, and safety monitoring technologies for tower cranes have become increasingly sophisticated, and construction robots have also seen rapid development in autonomous operation, path control, and quality feedback.

[0003] Most related technologies focus on the operation control of a single tower crane or the operation control of a single construction robot. In concrete pouring and leveling scenarios, when the tower crane supplies material to the target construction area directly affects when the leveling robot can enter the construction area to carry out leveling operations; the current operation stage and progress of the leveling robot, in turn, affects the arrangement of subsequent material supply actions by the tower crane. If there is a lack of a unified process organization mechanism between the two, the material supply process and the leveling process are prone to disconnection. Summary of the Invention

[0004] This application provides a collaborative scheduling and control method and related equipment for tower cranes, which can be used to improve the operating efficiency of tower crane feeding and leveling robots.

[0005] Firstly, this application provides a collaborative scheduling and control method for tower cranes and ground leveling robots. The method includes: dividing the target construction surface into multiple construction units and constructing a shared state vector with each construction unit as the minimum scheduling object. The shared state vector is used to uniformly represent the process state, spatiotemporal resource constraints, and equipment reachability of the corresponding construction unit; collecting the real-time operating status of the tower crane and ground leveling robot, as well as the environmental status of the construction site, and synchronizing the real-time operating status and environmental status to a shared digital model to update the shared state vector of each construction unit; based on the updated shared state vector, extracting the concrete workable time window constraint for each construction unit; and minimizing the equipment waiting time cost and the secondary impact of fresh concrete mixing. The disturbance cost is the objective of collaborative optimization. Combined with the constraint of the concrete workable time window, joint timing decisions are made for the material supply and leveling actions of each construction unit to generate the material supply task queue of the tower crane and the operation task queue of the ground leveling robot. The tower crane is controlled to perform material supply actions to the target construction unit in each construction unit according to the material supply task queue. After the material is delivered, the ground leveling robot is triggered to enter the target construction unit according to the operation task queue to perform leveling actions, forming a process linkage between material supply and leveling. During the construction process, if an abnormal disturbance event that deviates from the shared state vector is detected on site, the joint timing decision is re-triggered based on the current state of the affected construction unit to update the material supply task queue and the operation task queue.

[0006] In the above embodiments, the system constructs a shared state vector with the construction unit as the minimum scheduling object, incorporating the spatiotemporal resource constraints and process states of the two types of equipment into a unified analysis dimension. Based on this, a concrete workable time window constraint is introduced, and joint timing decisions are made with the goal of minimizing equipment waiting time costs and secondary disturbance costs of fresh concrete. This ensures that the material supply and leveling actions are linked on the time axis, avoiding equipment idleness or material performance degradation caused by the disconnect between material supply and leveling, and improving the continuity and resource utilization of multi-equipment cross-process collaborative operations.

[0007] In conjunction with some embodiments of the first aspect, in some embodiments, minimizing the equipment waiting time cost and the secondary disturbance cost of fresh concrete is taken as the collaborative optimization objective. Combined with the concrete workable time window constraint, joint temporal decision-making is performed on the material supply action and leveling action of each construction unit to generate the material supply task queue of the tower crane and the work task queue of the ground leveling robot. Specifically, this includes: using the updated shared state vector as the environment state space, and constructing the action space by combining the tower crane's action of selecting the next target construction unit to perform material supply and the leveling robot's action of selecting the next target construction unit to perform leveling; constructing a reward function based on the equipment waiting time cost and the secondary disturbance cost of fresh concrete, and introducing the concrete workable time window constraint as the penalty boundary condition of the action space; using a multi-agent reinforcement learning algorithm to iteratively solve the environment state space, action space, and reward function, and outputting the action sequence that maximizes the global reward value; parsing and splitting the action sequence into the material supply task queue of the tower crane and the work task queue of the ground leveling robot.

[0008] In the above embodiments, the system addresses the high dimensionality and coupling between the state space and action space in multi-device collaborative scheduling. It uses the updated shared state vector as the environmental state space, constructs a reward function by combining device waiting time and secondary disturbance costs, and introduces a time window as a penalty boundary condition. A multi-agent reinforcement learning algorithm is employed to iteratively solve the aforementioned space and function, enabling the tower crane and leveling robot to autonomously explore the action sequence with the maximum global reward value while satisfying spatiotemporal boundary constraints. This reduces the limitations of traditional rule-based heuristic algorithms in complex and variable construction environments and improves the optimization capability and adaptability of joint temporal decision-making under multiple constraints.

[0009] In conjunction with some embodiments of the first aspect, in some embodiments, a reward function is constructed based on the equipment waiting time cost and the secondary disturbance cost of fresh concrete. Specifically, this includes: extracting the time difference between the moment when the tower crane completes the material supply of the current construction unit and the moment when the ground leveling robot enters the current construction unit to start leveling, as the process interval time; when the process interval time is within a first preset interval, calculating the equipment waiting time cost based on the sum of the process interval time and the tower crane's no-load slewing time; when the process interval time exceeds the first preset interval but is within the concrete workable time window constraint, calculating the concrete slump loss value corresponding to the process interval time based on the concrete rheological property curve, and mapping the concrete slump loss value to the additional energy consumption cost of the ground leveling robot, as the secondary disturbance cost of fresh concrete; and constructing a reward function by weighted summation of the equipment waiting time cost and the secondary disturbance cost of fresh concrete, taking the negative value.

[0010] In the above embodiments, the system uses the process interval time as the core evaluation indicator, quantifying the equipment waiting time cost and the secondary disturbance cost of fresh concrete according to its interval. When the interval time is long, the slump loss value is calculated based on the concrete rheological property curve and mapped to the additional energy consumption cost of the leveling robot, transforming the dynamic changes in material physical properties into calculable equipment operating costs. By constructing a reward function through a weighted summation of these two types of costs, the system guides scheduling decisions to achieve a balance between equipment operating efficiency and construction quality maintenance, improving the accuracy of the collaborative optimization objective in representing the interaction between underlying material properties and equipment energy consumption.

[0011] In conjunction with some embodiments of the first aspect, in some embodiments, based on the updated shared state vector, the concrete workable time window constraint for each construction unit is extracted. Specifically, this includes: extracting the current environmental state data of the target construction unit from the shared state vector, the current environmental state data including real-time ambient temperature, ambient humidity, and wind speed collected by site IoT sensors; inputting the current environmental state data and the concrete mix proportion parameters of the target construction unit into a preset concrete hydration kinetic model to calculate the initial setting time node of the concrete for the target construction unit; and generating the concrete workable time window constraint for the target construction unit with the moment when the tower crane completes the material supply action as the starting boundary and the initial setting time node of the concrete as the ending boundary.

[0012] In the above embodiments, the system extracts current environmental state data, including real-time temperature, humidity, and wind speed, and inputs this data, along with concrete mix proportion parameters, into a preset concrete hydration kinetic model to dynamically calculate the initial setting time of the concrete in the target construction unit. The operational time window is defined by the completion time of material supply by the tower crane and the initial setting time, allowing the time window constraints to be dynamically adjusted in response to changes in external weather conditions and material properties, thus improving the objectivity of the time window boundary condition setting and its environmental adaptability.

[0013] In conjunction with some embodiments of the first aspect, in some embodiments, after calculating the initial setting time of concrete for the target construction unit, the method further includes: acquiring real-time resistance feedback data of the ground leveling robot when performing leveling actions in adjacent already-operated construction units, the real-time resistance feedback data including the drive wheel torque of the ground leveling robot and the working current of the leveling mechanism; comparing the real-time resistance feedback data with the theoretical setting state of adjacent already-operated construction units to calculate the setting state deviation value; and compensating the environmental impact coefficient in the preset concrete hydration kinetics model based on the setting state deviation value to correct the initial setting time of concrete for subsequent unoperated construction units.

[0014] In the above embodiments, the system acquires real-time resistance feedback data of the ground leveling robot in adjacent completed construction units, compares it with the theoretical setting state to quantify the setting state deviation value. This deviation value is then used to compensate for the environmental impact coefficient in the preset concrete hydration kinetics model, taking into account the implicit effects caused by local microenvironmental differences or material batch fluctuations. This corrects the initial setting time of concrete in subsequent uncompleted construction units, improving the prediction accuracy and reliability of the hydration kinetics model during continuous construction operations.

[0015] In conjunction with some embodiments of the first aspect, in some embodiments, if an abnormal disturbance event deviating from the shared state vector is detected on site, then based on the current state of the affected construction unit, the joint timing decision is re-triggered to update the material supply task queue and the work task queue. Specifically, this includes: real-time monitoring of the operation data of the construction site; if an avoidance delay triggered by the tower crane anti-collision system, a power warning for the ground leveling robot, or a sudden weather change is identified, then an abnormal disturbance event deviating from the shared state vector is determined to have occurred; the estimated time delay caused by the abnormal disturbance event is calculated, and the estimated time delay is compared with the concrete workable time window constraint of the affected construction unit to assess the spatiotemporal impact range of the abnormal disturbance event; based on the spatiotemporal impact range, the joint timing decision is re-triggered to update the material supply task queue and the work task queue.

[0016] In the above embodiments, the system monitors abnormal disturbances such as tower crane avoidance, robot power warnings, or weather changes in real time and calculates the estimated time delays they cause. This delay is compared with the concrete workable time window constraints of the affected construction units to quantitatively assess the spatiotemporal impact of the abnormal events on the overall process. This provides a quantitative triggering basis and impact boundary definition for subsequent rescheduling decisions, avoiding the computational waste and system oscillations caused by blind replanning, and improving the collaborative scheduling system's ability to perceive and assess dynamic disturbances on site.

[0017] In conjunction with some embodiments of the first aspect, in some embodiments, joint timing decisions are re-triggered based on the spatiotemporal impact range to update the material supply task queue and the work task queue. Specifically, this includes: if the spatiotemporal impact range does not exceed the concrete workable time window constraint of the affected construction unit, then local timing fine-tuning is performed, keeping the original task allocation order unchanged, and only updating the expected execution timestamp of each task in the material supply task queue and the work task queue; if the spatiotemporal impact range exceeds the concrete workable time window constraint of the affected construction unit, then the currently unexecuted material supply task queue and the work task queue are frozen, the affected construction unit is marked as a high-priority emergency state, and global replanning is performed based on the current shared state vector to regenerate the material supply task queue and the work task queue.

[0018] In the above embodiments, when the impact does not exceed the time window constraint, the system only updates the expected execution timestamp through local timing fine-tuning to maintain the original task allocation order and thus maintain system stability. When the impact exceeds the time window constraint, the unexecuted queue is frozen and the affected units are marked as high-priority emergency states, triggering global replanning based on the current shared state vector. This hierarchical processing ensures that high-risk construction units are given priority while minimizing the interference of abnormal disturbances on the overall predetermined procedures, thus improving the system's scheduling recovery elasticity and fault tolerance when facing emergencies of varying degrees.

[0019] In a second aspect, embodiments of this application provide a tower crane collaborative scheduling and control device, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the tower crane collaborative scheduling and control device to perform the method described in the first aspect and any possible implementation thereof.

[0020] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a tower crane collaborative scheduling and control device, cause the tower crane collaborative scheduling and control device to execute the method described in the first aspect and any possible implementation thereof.

[0021] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a tower crane collaborative scheduling and control device, cause the tower crane collaborative scheduling and control device to perform the method described in the first aspect and any possible implementation thereof.

[0022] Understandably, the tower crane collaborative scheduling and control equipment provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0024] 1. This application incorporates two types of equipment into a unified dimension by constructing a shared state vector, and performs joint timing decisions by combining time window constraints and collaborative optimization objectives, so that material supply and leveling are linked, avoiding equipment idleness or material depletion caused by process disconnection, and improving the continuity of cross-process collaborative operations.

[0025] 2. This application constructs a reward function that includes equipment waiting time and secondary disturbance cost, and uses a multi-agent reinforcement learning algorithm for iterative solution, guiding scheduling decisions to achieve a balance between equipment operating efficiency and construction quality, thereby improving the optimization capability and adaptability of joint time-series decision-making in complex environments.

[0026] 3. This application dynamically calculates the workable time window and quantitatively assesses the spatiotemporal impact range of abnormal disturbance events for hierarchical rescheduling, avoiding system oscillations caused by blind replanning and improving the collaborative scheduling system's ability to perceive, assess, and recover from dynamic disturbances on site. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating a tower crane collaborative scheduling and control method in an embodiment of this application.

[0028] Figure 2 This is a schematic diagram of the physical device structure of a tower crane collaborative scheduling and control equipment in the embodiments of this application. Detailed Implementation

[0029] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0030] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0031] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a collaborative scheduling and control method for tower cranes in an embodiment of this application.

[0032] S101. Divide the target construction surface into multiple construction units and construct a shared state vector with the construction unit as the minimum scheduling object. The shared state vector is used to uniformly represent the process status, spatiotemporal resource constraints and equipment reachability status of the corresponding construction unit.

[0033] In this context, a construction unit refers to the smallest scheduling and management unit formed by dividing the target construction surface according to certain rules. Each construction unit has independent spatial boundaries, process status identifiers, and equipment accessibility determination indicators. A shared state vector is used to uniformly represent the multi-dimensional state information of the construction unit, including process status (e.g., awaiting material supply, material supplied but awaiting leveling, leveling in progress, completed), spatiotemporal resource constraints (e.g., time window parameters, space occupancy status), and equipment accessibility status (e.g., tower crane coverage indicator, robot path accessibility indicator). The process status indicates the current construction stage of the construction unit; spatiotemporal resource constraints refer to the constraints imposed on equipment scheduling by the unit in the time and space dimensions; and equipment accessibility status indicates whether the tower crane and the ground leveling robot can physically reach the unit to perform operations.

[0034] Construction unit division is performed during the construction preparation phase, serving as a spatial benchmark for subsequent scheduling decisions. Specifically, this application provides two construction unit division methods: the first is based on a fixed grid, uniformly dividing the target construction surface according to a preset grid size (e.g., 5m × 5m); the second is based on comprehensive factors, comprehensively considering the boundaries of concrete pouring sections, the building structure axis, the tower crane's working radius coverage, and the single-operation width of the ground leveling robot, dividing the construction surface into several irregular but logically consistent construction units. Each construction unit is assigned a unique number and its spatial range is marked in the global coordinate system. During shared state vector initialization, initial values ​​for each field are preset based on construction drawings and equipment parameters. For example, the process status of all construction units is initialized to "awaiting material supply," and the tower crane reachability marker is pre-determined based on its slewing radius and radius range. When the tower crane completes material supply to a unit, the robot enters a unit for operation, or environmental monitoring data is updated, the shared state vector fields of the corresponding construction unit are triggered for updating.

[0035] S102. Collect the real-time operating status of the tower crane and the ground leveling robot, as well as the environmental status of the construction site, and synchronize the real-time operating status and environmental status to the shared digital model to update the shared state vector of each construction unit.

[0036] Real-time operational status refers to the equipment condition data reported in real time by tower cranes and ground leveling robots through sensors and controllers during construction. Environmental status refers to external environmental parameters and dynamic information collected at the construction site through IoT sensors. Shared digital model refers to a digital twin model of the construction site built on a cloud or edge computing platform, used to aggregate, store, and synchronize multi-source heterogeneous data, providing a unified data view for scheduling decisions.

[0037] Data acquisition is continuous throughout the entire construction process. The sampling period is set according to the data type; for example, the sampling period for equipment position and attitude data is 1 second, and the sampling period for environmental temperature and humidity data is 5 minutes. Event-based data (such as anti-collision triggers) is reported in real time using an event-driven approach. Specifically, the real-time operating status of the tower crane includes at least: current slewing angle (acquired via a rotary encoder), boom amplitude or hook height (acquired via a displacement sensor), lifting load (acquired via a force sensor), busy / idle status indicator (determined based on control commands), and anti-collision system status (output by the tower crane safety monitoring system). The real-time operating status of the ground leveling robot includes at least: global coordinate position (obtained via GNSS-RTK or UWB positioning), body attitude angle (obtained via an IMU inertial measurement unit), travel speed (calculated via a wheel speed encoder), leveling mechanism working status (determined via operating current monitoring), remaining battery percentage (reported via the battery management system BMS), and fault codes (diagnostic output by the equipment controller). The environmental conditions at the construction site should include at least: real-time ambient temperature, relative humidity, wind speed and direction (collected by a weather station), location of temporary obstacles (identified by on-site cameras or lidar), and personnel-occupied areas (obtained through safety fence sensors or video analysis).

[0038] Data synchronization can be achieved in two ways: For devices with industrial Ethernet or 5G private network access capabilities, their controllers directly report data to the cloud platform, which then updates the shared digital model after timestamp alignment; for devices using fieldbus or wireless LAN, data is first aggregated to the edge gateway, where protocol conversion, data preprocessing, and timestamp synchronization are performed before being uploaded to the platform. When a device experiences a short-term data loss (e.g., due to wireless signal obstruction), the system activates a short-term prediction compensation mechanism: based on the device's historical trajectory and current task instructions, a Kalman filter algorithm is used to predict its short-term state (prediction duration not exceeding 10 seconds) to maintain the continuity of scheduling decisions, and state correction is performed after data recovery. For outliers (e.g., jumps caused by sensor malfunctions), the system uses a sliding window midpoint filter to remove them.

[0039] S103. Based on the updated shared state vector, extract the concrete workable time window constraint for each construction unit.

[0040] The concrete workable time window constraint refers to the time interval constraint formed from the moment the tower crane completes material supply until the moment when the concrete in the construction unit experiences a significant decrease in fluidity due to the progress of the hydration reaction, making it no longer suitable for the leveling robot to operate. This time window is used to constrain the entry time and leveling completion time of the ground leveling robot, ensuring that the leveling operation is completed within the window period when the concrete maintains good workability.

[0041] The time window extraction is triggered immediately after the material supply for each construction unit is completed, providing a time boundary constraint for the leveling operation of that unit. Specifically, the system extracts the current environmental state data and concrete mix proportion parameters from the shared state vector corresponding to the construction unit. These parameters are then substituted into a preset concrete hydration kinetic model to calculate the initial setting time node. This application preferably adopts an improved maturity-Arrhenius model: Initial setting time T_initial = f(T_env, RH, W / C, C_type, Admixture), where T_env represents the real-time ambient temperature, RH represents the ambient relative humidity, W / C represents the water-cement ratio (or water-cement ratio) of the concrete, C_type represents the cement type parameter, and Admixture represents the type and dosage parameter of admixtures (such as retarders and water-reducing agents). Ambient temperature affects the hydration rate through the activation energy parameter in the Arrhenius equation, humidity affects the effective water-cement ratio through the evaporation correction coefficient, and admixtures adjust the hydration process through a retardation coefficient. After calculating the initial setting time, the time T_supply, where the tower crane completes the material supply action and the unloading is confirmed by the load sensor, is used as the starting boundary of the time window, and the initial setting time T_initial is used as the ending boundary. This generates the workable time window constraints [T_supply, T_initial] for the concrete in this construction unit. Considering the need to reserve a safety margin for leveling operations, in practical applications, the ending boundary is set as a hard constraint boundary 15-30 minutes in advance to avoid quality risks caused by operations close to the initial setting time. This scheme explicitly incorporates the time-varying characteristics of concrete materials into the equipment scheduling constraints through a hydration kinetic model, allowing the time window boundary to be dynamically adjusted according to external environmental conditions and material properties, thereby maximizing equipment scheduling flexibility while ensuring construction quality.

[0042] In some preferred embodiments, the above steps specifically include: extracting the current environmental state data of the target construction unit from the shared state vector, wherein the current environmental state data includes real-time ambient temperature, ambient humidity and wind speed collected by IoT sensors at the construction site;

[0043] Input the current environmental status data and the concrete mix proportion parameters of the target construction unit into the preset concrete hydration kinetic model to calculate the initial setting time node of the concrete in the target construction unit; take the moment when the tower crane completes the feeding action as the starting boundary and the initial setting time node of the concrete as the ending boundary to generate the workable time window constraint of the concrete in the target construction unit.

[0044] The concrete workable time window constraint refers to the time interval constraint formed from the moment the tower crane completes material supply until the moment when the concrete in the construction unit experiences a significant decrease in fluidity due to the progress of the hydration reaction, making it unsuitable for the leveling robot to operate. This time window is used to constrain the entry time and completion time of the ground leveling robot, ensuring that the leveling operation is completed within the window period when the concrete maintains good workability. The initial setting time node of concrete refers to the point at which the concrete paste begins to lose plasticity and form preliminary structural strength. According to the GB / T 50080-2016 standard, it is determined to be initial setting when the concrete penetration resistance reaches 3.5MPa. The concrete hydration kinetic model is a mathematical model based on the cement hydration reaction mechanism, used to predict the setting time of concrete with a specific mix proportion under given environmental conditions. Commonly used models include the Maturity Method and the Arrhenius equation improved model.

[0045] The time window extraction is triggered immediately after the material supply for each construction unit is completed, providing time boundary constraints for the leveling operation of that unit. Specifically, the system extracts the current environmental state data from the shared state vector corresponding to the construction unit, including the ambient temperature (accuracy ±0.5℃), relative humidity (accuracy ±3%RH), and wind speed (accuracy ±0.3m / s) collected in real time by the construction site's IoT weather station. Simultaneously, it extracts the concrete mix proportion parameters for that construction unit, including the water-cement ratio, cement type and dosage, proportion of mineral admixtures such as fly ash and mineral powder, type and dosage of admixtures such as retarders and water-reducing agents, design slump, and pouring layer thickness.

[0046] Using the aforementioned environmental data and mix proportion parameters as input, the initial setting time of the concrete for this construction unit is calculated by substituting them into a preset concrete hydration kinetics model. This application preferably employs an improved maturity-Arrhenius model, specifically calculating the initial setting time T using the following formula. initial : .

[0047] Where T0 represents the standard reference temperature T ref (Typically taken as 20℃), under standard humidity and without the addition of admixtures, based on the current cement type parameters (C type The initial setting time of the base material is determined experimentally or by referring to tables, along with the water-cement ratio (W / C). a The apparent activation energy of the concrete hydration reaction (unit: J / mol) is expressed, and its value is related to the cement type and water-cement ratio; R is the ideal gas constant, with a value of 8.314 J / (mol·K); T env Real-time ambient temperature (unit: °C) collected by IoT sensors at the construction site; RH is the real-time relative humidity (expressed as a decimal, e.g., 80% is 0.80); kRH The preset evaporation correction coefficient is used to characterize the negative impact of water evaporation on the effective water-cement ratio and hydration rate when the ambient humidity is low. This is the admixture delay coefficient, determined based on the type and dosage parameters of the admixture (such as retarders and water-reducing agents). When no admixture is used... =1, when a retarder is added >1.

[0048] The above formula shows that ambient temperature exponentially affects the hydration rate through the activation energy parameter in the Arrhenius equation, humidity linearly affects the effective water-cement ratio through the evaporation correction coefficient, and admixtures adjust the hydration process through the retardation coefficient, thereby achieving accurate dynamic calculation of the initial setting time of concrete under complex on-site conditions.

[0049] After calculating the initial setting time, the time T_supply, where the tower crane completes the material supply action and the unloading is confirmed by the load sensor, is used as the starting boundary of the time window, and the initial setting time T_initial is used as the ending boundary. This generates the workable time window constraints [T_supply, T_initial] for the concrete in this construction unit. Considering that a safety margin needs to be reserved for leveling operations, in practical applications, the ending boundary is set as a hard constraint boundary 15-30 minutes in advance to avoid quality risks caused by operations close to the initial setting time.

[0050] This solution explicitly incorporates the time-varying characteristics of concrete materials into equipment scheduling constraints through a hydration kinetic model. This allows the time window boundaries to be dynamically adjusted based on external environmental conditions (temperature, humidity, wind speed) and the intrinsic properties of the material (mixing ratio), rather than using fixed durations or empirical values. This dynamic calculation automatically shortens the time window in high-temperature weather and automatically extends it in low-temperature weather or when a large amount of retarder is added, thereby maximizing equipment scheduling flexibility while ensuring construction quality.

[0051] In some other embodiments of this application, after calculating the initial setting time of the concrete in the target construction unit, the method further includes: obtaining real-time resistance feedback data of the ground leveling robot when it performs leveling action in an adjacent already-operated construction unit, wherein the real-time resistance feedback data includes the drive wheel torque of the ground leveling robot and the working current of the leveling mechanism.

[0052] The real-time resistance feedback data is compared with the theoretical setting state of adjacent completed construction units to calculate the setting state deviation value. Based on the setting state deviation value, the environmental impact coefficient in the preset concrete hydration kinetic model is compensated to correct the initial setting time node of concrete in subsequent uncompleted construction units.

[0053] Real-time resistance feedback data refers to the operational resistance-related parameters collected in real time by sensors in the drive system and actuators of the ground leveling robot during the leveling operation, used to characterize the current actual setting state of the concrete. Drive wheel torque is collected by a torque sensor installed on the output shaft of the robot's walking motor, reflecting the traction resistance experienced by the robot as it travels on the concrete surface. The operating current of the leveling mechanism is collected in real-time by a current transformer from the drive motor of the leveling vibration device or scraper, reflecting the medium resistance experienced by the leveling mechanism when working in the concrete. The theoretical setting state refers to the degree of setting that the construction unit should reach at the current moment, predicted according to the concrete hydration kinetics model, and can be characterized by penetration resistance or rheological parameters. The setting state deviation value represents the difference between the actual setting state and the theoretical prediction value, used to quantify the model prediction error.

[0054] This correction mechanism is triggered after the ground leveling robot completes the leveling operation of any construction unit, and is used to correct the time window prediction for subsequent uncompleted construction units. Specifically, the system acquires real-time resistance feedback data when the robot performs leveling actions in adjacent completed construction units. The sampling frequency can be selected as 10Hz. After applying a moving average filter to the resistance data throughout the entire operation, feature values ​​(such as average torque, peak current, and resistance change gradient) are extracted. Simultaneously, based on the material supply time of the construction unit, the environmental parameters at that time, and the mix proportion parameters, the theoretical setting state at that moment is calculated using a concrete hydration kinetic model (represented by equivalent penetration resistance). The real-time resistance feedback data is converted into the measured setting state through a pre-calibrated resistance-setting state mapping relationship, compared with the theoretical setting state, and the setting state deviation value Δ = measured setting state - theoretical setting state is calculated. If Δ is positive, it indicates that the actual setting speed is faster than the model prediction; if Δ is negative, it indicates that the actual setting speed is slower than the prediction.

[0055] Based on the deviation value of the setting state, the system adaptively compensates for the environmental influence coefficient in the preset concrete hydration kinetics model. For example, if the deviation values ​​of multiple consecutive construction units are all positive and close in value, it indicates that there is a local microenvironment (such as surface radiant heat, cross drafts) or batch fluctuations in materials (such as a low actual water-cement ratio) leading to accelerated setting. In this case, the temperature correction coefficient K_T or the hydration rate constant k in the model is adjusted upward (e.g., K_T' = K_T × (1 + α·Δ), where K_T' represents the new temperature correction coefficient after deviation compensation, and α is the compensation gain coefficient, calibrated through regression of historical data). The corrected model parameters are used to calculate the initial setting time node of concrete for subsequent unworked construction units, making the termination boundary of the time window more closely match the actual site conditions.

[0056] S104. With minimizing equipment waiting time cost and secondary disturbance cost of fresh concrete as the collaborative optimization objective, and combined with the concrete workable time window constraint, the joint timing decision is made on the material supply action and leveling action of each construction unit to generate the material supply task queue of the tower crane and the operation task queue of the ground leveling robot.

[0057] Joint timing decision-making refers to simultaneously determining the respective operational target construction unit sequences and corresponding expected execution times of tower cranes and ground leveling robots within a unified decision-making framework, enabling the operational sequences of the two types of equipment to coordinate. Equipment waiting time cost refers to the time cost incurred due to equipment idle waiting caused by the mismatch between the material supply and leveling action sequences, including the idle time of the tower crane waiting for the next task after completing material supply and the standby time of the robot waiting for material to arrive. Secondary disturbance cost of fresh concrete refers to the additional energy consumption and adverse effects on the internal structure of the concrete caused by the robot needing to apply greater vibration and leveling forces to complete the leveling operation when the leveling action is delayed and the concrete fluidity decreases. The material supply task queue is represented as an ordered sequence {(unit number_i, expected material supply start time_i, expected material supply completion time_i, preconditions_i, priority_i)}, and the operation task queue is represented as {(unit number_j, expected entry time_j, expected leveling completion time_j, preconditions_j, priority_j)}.

[0058] Joint timing decision-making is executed each time a global scheduling is triggered. The decision inputs include: the updated shared state vector of all construction units (including the current process state, remaining time window, and space occupancy status of each unit), the current position and attitude of the tower crane and the robot, and the task release status (which units have released their material supply requirements), and the predicted path conflict results between equipment (such as the overlap between the tower crane's slewing range and the robot's working area). The decision outputs are the tower crane's material supply task queue and the robot's work task queue. Each task item in the queue includes the target construction unit number, the expected start and end times, the pre-constraints that need to be met (such as the robot needing to wait for the adjacent unit to finish leveling before it can withdraw), and the priority identifier (normal / high priority emergency).

[0059] Specifically, this application provides two feasible implementation routes. The first route employs a multi-agent reinforcement learning algorithm to model the problem as a multi-agent Markov decision process, solving for the optimal action sequence through offline training and online inference. The second route combines engineering heuristics with mathematical programming: First, based on the spatial distribution of construction units, time window urgency, and current equipment location, heuristics (such as prioritizing units with the tightest time windows and closest distances) are used to generate an initial task allocation order; then, this order is used as the initial solution to establish a mixed-integer linear programming (MILP) model, with the objective function being the weighted sum of equipment waiting time cost and fresh concrete secondary disturbance cost, and constraints including time window constraints, equipment kinematic constraints, and task dependencies. Commercial solvers (such as Gurobi and CPLEX) are called for local optimization, outputting the optimized task queue. When the MILP solution times out or has no feasible solution, the system reverts to the initial solution generated by the heuristics. The system reserves a 5-10 minute buffer period between the material feeding and leveling actions to accommodate minor adjustments to the tower crane's material dropping position, robot path planning, and sudden short-term interference.

[0060] This solution couples the originally independent tower crane scheduling problem and robot scheduling problem into a unified multi-agent collaborative decision-making problem by sharing state vectors and joint optimization objectives. This enables synchronous optimization of material supply sequence and leveling sequence at the decision-making level. By introducing the secondary disturbance cost corresponding to the time-varying characteristics of concrete, the implicit constraint of material physical performance degradation is explicitly incorporated into the optimization objective. This guides the scheduling algorithm to seek a dynamic balance between equipment efficiency and construction quality, thus resolving the contradiction between "waiting waste" and "delay degradation" in multi-equipment cross-process collaboration.

[0061] In some preferred embodiments, the above steps specifically include: using the updated shared state vector as the environmental state space, and constructing the action space together with the action of the tower crane selecting the next target construction unit to perform material supply and the action of the leveling robot selecting the next target construction unit to perform leveling;

[0062] A reward function is constructed based on the cost of equipment waiting time and the cost of secondary disturbance of fresh concrete, and a workable time window constraint for concrete is introduced as a penalty boundary condition for the action space.

[0063] A multi-agent reinforcement learning algorithm is used to iteratively solve the environmental state space, action space, and reward function, and output the action sequence that maximizes the global reward value. The action sequence is then parsed and split into a material feeding task queue for a tower crane and a task queue for a ground leveling robot.

[0064] In this context, the environmental state space refers to the set of all state variables used to describe the current decision-making environment within a multi-agent reinforcement learning framework. In this scheme, it is composed of the updated shared state vectors of each construction unit. The action space refers to the set of actions that each agent can choose. In this scheme, the actions of the tower crane agent are "selecting construction unit number i to perform material supply" or "waiting," and the actions of the ground leveling robot agent are "selecting construction unit number j to perform leveling" or "waiting." The reward function is used to evaluate the immediate reward obtained by the agent after performing a specific action in a specific state, guiding the agent to learn the optimal strategy. The global reward value refers to the weighted cumulative sum of the immediate rewards of each agent in the multi-agent collaborative framework, representing the long-term benefit of the overall system. The action sequence refers to the ordered sequence of actions selected sequentially by the agent within multiple decision steps, corresponding to a complete scheduling scheme.

[0065] The solution process is executed when the scheduling decision is triggered, employing a multi-agent reinforcement learning algorithm based on a centralized training-distributed execution (CTDE) architecture. Specifically, the updated shared state vector of all construction units is expanded into a state vector s, whose dimensions include: the remaining time window length of each candidate construction unit, the estimated time for the tower crane to arrive at the unit (calculated based on the current slewing angle and amplitude), the estimated time for the robot to arrive at the unit (calculated based on the current position and path planning), the current occupancy status of the unit (idle / tower crane in operation / robot in operation), and the material arrival status (not in place / in place), etc. The action space encoding of the tower crane agent is a_tower ∈ {0, 1, 2, ..., N, N+1}, where 0~N correspond to feeding materials to one of the N candidate construction units, and N+1 corresponds to waiting actions; the action space encoding of the ground leveling robot agent is a_robot ∈ {0, 1, 2, ..., N, N+1}, with similar meaning. To avoid illegal actions (such as selecting a unit that has not yet been fed or selecting a unit that is beyond the reach of the equipment), the system sets up an action masking mechanism in the action space: based on the current state vector, the Q value or policy probability corresponding to actions that do not meet the preconditions is set to negative infinity or zero, forcing the agent to select only from the set of legal actions.

[0066] The reward function is constructed based on the equipment waiting time cost C_wait and the secondary disturbance cost of fresh concrete C_disturb, and a concrete workable time window constraint is introduced as a penalty boundary condition: if the selected leveling action causes the leveling completion time to exceed the time window termination boundary, a large negative penalty P_penalty (e.g., -1000) is applied; otherwise, the reward is r = -(w1·C_wait + w2·C_disturb), where w1 and w2 are weight coefficients determined through expert experience or hyperparameter search (e.g., w1=0.4, w2=0.6). The global reward is defined as the weighted sum of the rewards of the two agents R_global = r_tower + r_robot, where r_tower represents the immediate reward value obtained by the tower crane agent after performing the action, and r_robot represents the immediate reward value obtained by the ground leveling robot agent after performing the action.

[0067] This application preferably uses the QMIX or MAPPO algorithm for training. Training samples are derived from historical construction data playback and virtual construction scenarios generated from simulation environments. During training, the agent interacts with the state, performs actions, and receives rewards in the environment. Gradient descent is used to optimize the neural network parameters, maximizing the cumulative global reward. After training convergence, the trained policy network is deployed to the scheduling system. During online inference, the current state vector is input into the policy network, which outputs the action distribution of each agent. A greedy or sampling strategy is used to select actions, gradually building a complete material supply task queue and a work task queue. If the reinforcement learning model fails during online inference (e.g., encountering an abnormal state not covered by the training set), the system automatically reverts to engineering heuristics for decision-making, ensuring system robustness.

[0068] In some other preferred embodiments, in the above embodiments, a reward function is constructed based on the equipment waiting time cost and the secondary disturbance cost of fresh concrete. Specifically, this includes: extracting the time difference between the moment when the tower crane completes the material supply of the current construction unit and the moment when the ground leveling robot enters the current construction unit and begins leveling, as the process interval time; when the process interval time is within a first preset range, calculating the equipment waiting time cost based on the sum of the process interval time and the tower crane's no-load slewing time.

[0069] When the interval between processes exceeds the first preset range but is within the time window constraint for concrete operation, the slump loss value of concrete corresponding to the interval between processes is calculated based on the concrete rheological property curve. The slump loss value of concrete is then mapped to the additional energy consumption cost of the ground leveling robot as the secondary disturbance cost of fresh concrete. The equipment waiting time cost and the secondary disturbance cost of fresh concrete are weighted and summed, and the negative value is used to construct the reward function.

[0070] The process interval time refers to the time difference between T_supply_done (the moment the tower crane completes the material supply action of the current construction unit, unloads the hook, and confirms the material's arrival) and T_leveling_start (the moment the ground leveling robot enters the construction unit and the leveling mechanism begins work), expressed as Δt = T_leveling_start - T_supply_done. The first preset interval refers to the ideal coordinated range of the process interval time, within which the robot can follow up on the operation in a timely manner, the equipment waiting time is short, and the concrete fluidity remains good. It is usually set to [0, 20 minutes]. The upper limit of this interval is determined based on historical construction data and statistics on the initial slump of the concrete. The tower crane's no-load slewing time refers to the time required for the hook to slew back to the next construction unit to be supplied after the tower crane completes the material supply of the current construction unit. It is calculated based on the difference in slewing angle between the two units and the tower crane's rated slewing speed. The concrete rheological property curve represents the decay relationship of concrete slump or yield stress over time, usually exhibiting an exponential or hyperbolic decreasing trend, obtained by laboratory rheometer testing or field slump loss testing.

[0071] The reward function is triggered at the start of each decision step to quantify the immediate cost of the current action selection. Specifically, the system first extracts the process interval Δt of the construction unit involved in the current decision. It then determines the range of Δt: if Δt ≤ 20 minutes (within the first preset range), it indicates a close connection between material supply and leveling, where the main cost comes from equipment waiting. The system calculates the equipment waiting time cost C_wait = k1·(Δt + T_rotation), where T_rotation is the time it takes for the tower crane to rotate empty to the next target unit after completing the current material supply, and k1 is the unit cost coefficient for waiting time (unit: yuan / minute), reflecting equipment rental costs and schedule pressure. This formula reflects that if the robot's entry delay Δt is too long, the tower crane cannot immediately execute the next task during this period and must wait for the robot to complete leveling and withdraw before rotating to the next unit. Therefore, the waiting cost is proportional to the sum of Δt and rotation time.

[0072] If Δt > 20 minutes and Δt ≤ T_time window (outside the first preset interval but still within the time window constraint), it indicates that the interval between material supply and leveling is relatively long, and the concrete fluidity has significantly decreased. The system queries or calculates the concrete slump loss value ΔSlump = Slump_initial - Slump(Δt) at time Δt from the pre-stored concrete rheological property curve. Slump loss leads to increased concrete viscosity and yield stress, requiring the robot leveling mechanism to apply a higher vibration frequency and pushing force to achieve the designed flatness. The system converts the slump loss value into the additional energy consumption cost of the ground leveling robot through a pre-calibrated mapping function: C_disturb = k2·f(ΔSlump), where f(ΔSlump) can be linearly mapped as f(ΔSlump) = a·ΔSlump + b, where a and b are pre-calibrated linear mapping coefficients (constant parameters) based on experimental data, used to linearly convert the slump loss value into the robot's additional energy consumption. Alternatively, a nonlinear fitting curve based on robot power-slump test data can be used, where k2 is the energy consumption cost coefficient (unit: yuan / kWh). For example, when the slump decreases from 180mm to 120mm, the measured energy consumption of robot leveling increases by approximately 30%, then f(60mm) = 0.3·E_baseline energy consumption. This mapping transforms changes in material physical properties into measurable economic costs, enabling scheduling algorithms to explicitly balance "rapid equipment rotation" with "maintaining material properties."

[0073] The final reward function is constructed as: r = -(w1·C_wait + w2·C_disturb), taking a negative value to align with the optimization direction of "maximizing cumulative reward" in reinforcement learning. The weights w1 and w2 are set according to the actual project situation; for example, increasing w1 in projects with high time pressure prioritizes reducing waiting time, while increasing w2 in projects with high-quality requirements prioritizes reducing secondary disturbances. If Δt exceeds the time window constraint (Δt > T_time window), a large penalty term P_penalty (e.g., -1000) is added to the reward function to force the algorithm to avoid such actions. This segmented cost calculation and weighted fusion allows the reward function to finely characterize the multi-dimensional costs under different process connection states, providing clear optimization signals for multi-agent learning.

[0074] S105. Control the tower crane to perform material supply actions to the target construction unit in each construction unit according to the material supply task queue, and after the material is supplied, trigger the ground leveling robot to enter the target construction unit according to the work task queue to perform leveling actions, forming a process linkage between material supply and leveling.

[0075] In this context, "material delivery" refers to the tower crane transporting concrete hoppers to the target construction unit, unloading the material by lowering the hook, and confirming with sensors or a vision system that the material has fallen into the designated area. "Process linkage" refers to the coordinated operation mode where the material delivery process and the leveling process are linked in time and space. The completion status of the previous process serves as a prerequisite for the start of the next process, and the two are automatically linked through digital signal transmission.

[0076] This execution control continues after the joint timing decision generates the task queue, spanning the entire construction operation process. Specifically, based on the generated material supply task queue, the system sequentially issues material supply instructions to the tower crane control system. These instructions include: the target construction unit number, the three-dimensional coordinates (X, Y, Z) of the target material drop point, the target slewing angle, the target luffing radius or hook target height, and the expected arrival time. Upon receiving the instructions, the tower crane control system executes automated or semi-automated operations: the hook grabs the hopper, slewing to the target angle, luffing to the target radius, descending to the target height, and opening the hopper's bottom door to unload the material.

[0077] The determination of material supply arrival adopts a multi-source information fusion method: The first method is based on the load sensor. When the load suddenly changes from full load (e.g., 2 tons) to empty load (e.g., 0.2 tons) and lasts for more than 3 seconds, the unloading is determined to be complete. The second method is based on the vision camera installed on the hook or hopper. The image recognition algorithm detects the opening status of the hopper bottom door and the falling status of the concrete to confirm that the material has fallen into the construction unit. The third method combines GPS-RTK or UWB positioning system to verify that the deviation between the current position of the hopper and the target material drop point is within the allowable range (e.g., horizontal deviation <0.5m). If any two of the three methods are met simultaneously, or if a single method is confirmed twice consecutively, the system determines that the "material supply arrival" condition is met, and the tower crane controller reports a "material supply completion signal" to the cloud platform or edge gateway.

[0078] After receiving the material supply completion signal, the platform immediately executes the linkage trigger verification process: First, update the "material arrival status" field in the shared state vector of the construction unit to "in place", and at the same time extract the robot entry threshold conditions of the unit for item-by-item verification, including: (1) material arrival status = in place; (2) no obstacles in the area, that is, the "obstacle occupancy flag" of the construction unit and its adjacent channel units is "none"; (3) the current time is within the concrete workable time window constraint of the unit; (4) the ground leveling robot battery is ≥30% and there is no fault alarm; (5) the robot's current posture is normal (roll angle and pitch angle are within the safe range). If all the above conditions are met, the platform issues an entry operation instruction to the ground leveling robot control system in the order of "tower crane delivery completion signal → platform verification → robot entry instruction issuance". The instruction content includes: target construction unit number, entry path waypoint sequence, leveling operation parameters (vibration frequency, leveling speed, number of operation passes), and expected operation time.

[0079] After receiving the command, the robot control system activates the path planning module to calculate the obstacle avoidance path, drives the robot to the boundary of the target construction unit, and enters the unit to activate the leveling mechanism (vibrator, scraper, laser leveler, etc.) to perform leveling operations according to preset parameters. After the operation is completed, the robot sends the leveling completion signal and flatness detection data (such as laser scanning point cloud) back to the platform. The platform updates the "process status" of the shared state vector of the construction unit to "completed" and releases the space occupancy flag of the unit, allowing subsequent processes or equipment to enter.

[0080] The system is equipped with a manual takeover interface: when any abnormality occurs in any part of the automated process (such as visual recognition failure or robot path planning timeout), the system will automatically pause the current task, push alarm information to the on-site management personnel, and wait for manual confirmation or remote intervention; after the handling is completed, the personnel can manually trigger the "resume execution" command through the management platform, and the system will continue to execute from the breakpoint.

[0081] S106. During the construction process, if an abnormal disturbance event deviating from the shared state vector is detected on site, the joint timing decision is re-triggered based on the current state of the affected construction unit to update the material supply task queue and the operation task queue.

[0082] Among them, abnormal disturbance events refer to sudden events during construction execution where, due to equipment failure, sudden environmental changes, safety avoidance, or other reasons, the actual execution state deviates from the original expected task queue, and the impact exceeds the system's automatic absorption threshold. Deviation from the shared state vector refers to inconsistencies between real-time monitoring data and the expected or planned state stored in the shared state vector. Examples include excessive deviations between the actual and planned equipment positions, operation durations far exceeding expectations, and sudden changes in environmental parameters causing time window constraints to fail. Re-triggering the joint timing decision means that after detecting an abnormal disturbance, the system suspends the subsequent execution of the original task queue and, based on the current actual state, re-invokes the joint timing decision algorithm to generate an updated task queue to adapt to the new site conditions.

[0083] This monitoring and rescheduling mechanism operates in real-time throughout the entire construction process, with a monitoring cycle selectable as 1 second. Specifically, the system collects operational data from the construction site in real time and compares it with the expected state in the shared state vector to identify abnormal disturbance event types. The first type of abnormality is the avoidance delay triggered by the tower crane anti-collision system. When multiple tower cranes exist on site, or when the working range of a tower crane conflicts with the space of temporary facilities (such as construction elevators or material hoists), the anti-collision system, based on real-time position monitoring, forces the tower crane to slow down or stop if a collision risk is determined, and triggers the avoidance coordination process. The system identifies this type of abnormality by listening to the "anti-collision triggered" event flag reported by the tower crane controller and records the duration of the avoidance delay. The second type of abnormality is the ground leveling robot's battery level warning. The system monitors the remaining battery percentage reported by the robot's battery management system (BMS) in real time. When the battery level is ≤20%, a warning is triggered; when the battery level is ≤10%, a forced return to charging is triggered. The battery level warning will cause the robot to interrupt its current task or abandon subsequent tasks and return to the charging station, causing the execution of the work task queue to be interrupted. The third type of anomaly is sudden weather change. The system monitors the construction site's weather station data in real time. When the ambient temperature changes by ≥5℃ within 30 minutes, the wind speed suddenly increases to ≥8m / s, or a rainfall signal is detected, a sudden weather change is identified. Weather changes significantly alter the concrete hydration rate and the workable time window, causing the original time window constraints to fail. In addition, the system can also identify other anomalies such as personnel intrusion into the work area, interruption of concrete pumping, robot slippage and getting stuck, and tower crane unbalanced load.

[0084] Upon detecting any type of abnormal disturbance, the system initiates an impact assessment process. First, the estimated time delay caused by the abnormal disturbance is calculated: for collision avoidance, the delay is the avoidance duration plus the tower crane's preparation time to resume normal operation; for robot battery warnings, the delay is the total time for the robot to return to the charging station from its current location, complete charging, and then return to the work area (estimated based on historical data or real-time path planning); for sudden weather changes, the delay is the estimated time waiting for weather conditions to return to safe operating standards. The system compares the estimated time delay with the concrete workable time window constraint of the affected construction unit: it extracts the expected leveling completion time T_plan of the affected unit in the original task queue, adds the estimated delay to obtain a new expected completion time T_new = T_plan + Δt_delay, and determines whether T_new exceeds the time window termination boundary T_window of that unit. If T_new ≤ T_window, the delay can still be absorbed within the time window; if T_new > T_window, the delay has caused the time window constraint to fail.

[0085] In some preferred embodiments, the above steps specifically include: real-time monitoring of the operation data of the construction site; if the avoidance delay triggered by the tower crane anti-collision system, the power warning of the ground leveling robot, or the sudden weather change is detected, it is determined that an abnormal disturbance event deviating from the shared state vector has occurred.

[0086] Calculate the estimated time delay caused by the abnormal disturbance event, and compare the estimated time delay with the concrete workable time window constraint of the affected construction unit to assess the spatiotemporal impact range of the abnormal disturbance event; based on the spatiotemporal impact range, re-trigger joint timing decision to update the material supply task queue and the work task queue.

[0087] Among them, abnormal disturbance events refer to sudden events during construction execution where, due to equipment failure, sudden environmental changes, safety avoidance, or other reasons, the actual execution state deviates from the original expected task queue, and the impact exceeds the system's automatic absorption threshold. Deviation from the shared state vector refers to inconsistencies between real-time monitoring data and the expected or planned state stored in the shared state vector, such as excessive deviation between the actual and planned equipment positions, operation duration far exceeding the expected duration, or sudden changes in environmental parameters causing time window constraints to fail. Expected time delay refers to the time loss caused by the abnormal disturbance event from its occurrence to its complete elimination, including the event duration, equipment recovery time, and time to return to the work position. The spatiotemporal impact range refers to the ripple effect of the abnormal disturbance event on the set of construction units in both time and space dimensions, specifically manifested in which construction units' time window constraints are at risk of failure, which equipment movement paths need to be replanned, and which areas' resource occupancy status needs to be adjusted.

[0088] This monitoring and evaluation mechanism operates in real time throughout the entire construction process, with a monitoring cycle typically set to 1 second. Specifically, the system collects operational data from the construction site in real time and compares it with the expected state in the shared state vector to identify the types of abnormal disturbance events. When an avoidance delay triggered by the tower crane's collision avoidance system, a ground leveling robot battery warning, or a sudden weather change is detected, an abnormal disturbance event deviating from the shared state vector is determined. The system then initiates an impact assessment process to calculate the estimated time delay: for collision avoidance, the delay is the avoidance duration plus the tower crane's preparation time to resume normal operation; for robot battery warnings, the delay is the total time it takes for the robot to return to the charging station from its current location, complete charging, and then return to the work area; for sudden weather changes, the delay is the estimated time waiting for weather conditions to return to safe operating standards. The system compares the estimated time delay with the concrete workable time window constraint of the affected construction unit, extracts the expected leveling completion time T_plan of the affected unit in the original task queue, and adds the estimated delay to obtain the new expected completion time T_new = T_plan + Δt_delay. It then determines whether T_new exceeds the time window termination boundary T_window of the unit and evaluates the spatiotemporal impact range of the abnormal disturbance event.

[0089] In other embodiments of this application, in the above embodiments, joint timing decisions are re-triggered based on the spatiotemporal impact range to update the material supply task queue and the work task queue. Specifically, if the spatiotemporal impact range does not exceed the concrete workable time window constraint of the affected construction unit, local timing fine-tuning is performed to keep the original task allocation order unchanged and only update the expected execution timestamp of each task in the material supply task queue and the work task queue.

[0090] If the spatiotemporal impact exceeds the concrete workable time window constraint of the affected construction unit, then the currently unexecuted material supply task queue and work task queue are frozen, the affected construction unit is marked as a high-priority emergency state, and a global replanning is performed based on the current shared state vector to regenerate the material supply task queue and work task queue.

[0091] Local timing fine-tuning refers to a lightweight scheduling correction method that maintains the original task allocation order and only shifts the expected execution timestamps of each task in the task queue when the impact of abnormal disturbances is controllable. Global replanning refers to a deep scheduling reconstruction method that, when abnormal disturbances cause time window constraints to fail, suspends the subsequent execution of the original task queue, and re-invokes the joint timing decision algorithm based on the current actual state to reorder and reschedule all unexecuted tasks.

[0092] This hierarchical rescheduling mechanism is triggered after the system completes the assessment of the impact of abnormal disturbances. Specifically, if the spatiotemporal impact range does not exceed the concrete workable time window constraint of the affected construction unit (i.e., T_new ≤ T_window), it means that the delay can still be absorbed within the time window. The system performs local timing fine-tuning: keeping the original task allocation order of the material supply task queue and the work task queue unchanged, and only shifting the expected execution timestamps of subsequent tasks uniformly according to the expected time delay (e.g., adding Δt_delay to the start time of all subsequent tasks), maintaining the relative timing relationship between tasks and the process connection logic, and minimizing interference with the global established process.

[0093] If the spatiotemporal impact range exceeds the concrete workable time window constraint of the affected construction unit (i.e., T_new > T_window), it indicates that the delay has caused the time window to fail. The system performs a global replanning: freezes the currently unexecuted material supply task queue and work task queue, marks the affected construction unit as a high-priority emergency state in the shared state vector, and then calls the joint timing decision algorithm again based on the shared state vector of all current construction units (including completed, executing, and unexecuted state updates) to regenerate the material supply task queue and work task queue, ensuring that high-priority emergency units can obtain equipment resource allocation first.

[0094] The tower crane collaborative scheduling and control device in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link to relevant documentation]. Figure 2 This is a schematic diagram of the physical device structure of a tower crane collaborative scheduling and control equipment in the embodiments of this application.

[0095] It should be noted that, Figure 2 The structure of the tower crane collaborative scheduling and control device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0096] like Figure 2 As shown, the tower crane collaborative scheduling and control equipment includes a Central Processing Unit (CPU) 201, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 202 or programs loaded from storage section 208 into Random Access Memory (RAM) 203, such as executing the methods described in the above embodiments. The RAM 203 also stores various programs and data required for system operation. The CPU 201, ROM 202, and RAM 203 are interconnected via a bus 204. An input / output (I / O) interface 205 is also connected to the bus 204.

[0097] The following components are connected to I / O interface 205: input section 206 including audio input devices, push-button switches, etc.; output section 207 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 208 including a hard disk, etc.; and communication section 209 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 209 performs communication processing via a network such as the Internet. Drive 210 is also connected to I / O interface 205 as needed. Removable media 211, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 210 as needed so that computer programs read from them can be installed into storage section 208 as needed.

[0098] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 209, and / or installed from removable medium 211. When the computer program is executed by central processing unit (CPU) 201, it performs the various functions defined in the present invention.

[0099] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0100] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0101] Specifically, the tower crane collaborative scheduling and control device in this embodiment includes a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it implements the tower crane collaborative scheduling and control method provided in the above embodiment.

[0102] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the tower crane collaborative scheduling and control device described in the above embodiments; or it may exist independently and not assembled into the tower crane collaborative scheduling and control device. The storage medium carries one or more computer programs, which, when executed by a processor of the tower crane collaborative scheduling and control device, cause the tower crane collaborative scheduling and control device to implement the tower crane collaborative scheduling and control method provided in the above embodiments.

[0103] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0104] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0105] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for collaborative scheduling and control of a tower crane and a ground leveling robot, characterized in that, The method includes: The target construction surface is divided into multiple construction units, and a shared state vector is constructed with the construction unit as the minimum scheduling object. The shared state vector is used to uniformly represent the process status, spatiotemporal resource constraints, and equipment reachability status of the corresponding construction unit. The real-time operating status of the tower crane and the ground leveling robot, as well as the environmental status of the construction site, are collected, and the real-time operating status and the environmental status are synchronized to the shared digital model to update the shared state vector of each construction unit. Based on the updated shared state vector, extract the concrete workable time window constraint for each construction unit; With minimizing equipment waiting time cost and secondary disturbance cost of fresh concrete as the collaborative optimization objective, and combined with the concrete workable time window constraint, the material supply action and leveling action of each construction unit are jointly timed to generate the material supply task queue of the tower crane and the operation task queue of the ground leveling robot. The tower crane is controlled to perform the material feeding action to the target construction unit in each of the construction units according to the material feeding task queue. After the material is supplied, the ground leveling robot is triggered to enter the target construction unit according to the work task queue to perform the leveling action, thus forming a process linkage between material feeding and leveling. If an abnormal disturbance event deviating from the shared state vector is detected during construction, the joint timing decision is re-triggered based on the current state of the affected construction unit to update the material supply task queue and the operation task queue.

2. The method according to claim 1, characterized in that, The optimization objective is to minimize equipment waiting time and secondary disturbance cost of fresh concrete. Combined with the concrete workable time window constraint, joint timing decisions are made for the material supply and leveling actions of each construction unit to generate the material supply task queue for the tower crane and the work task queue for the ground leveling robot. Specifically, this includes: The updated shared state vector is used as the environment state space, and the actions of the tower crane selecting the next target construction unit to perform material supply and the actions of the leveling robot selecting the next target construction unit to perform leveling are jointly constructed as the action space. A reward function is constructed based on the equipment waiting time cost and the secondary disturbance cost of the fresh concrete, and the concrete workable time window constraint is introduced as a penalty boundary condition for the action space. A multi-agent reinforcement learning algorithm is used to iteratively solve the environment state space, the action space, and the reward function, and outputs the action sequence that maximizes the global reward value. The action sequence is parsed and split into the material feeding task queue of the tower crane and the operation task queue of the ground leveling robot.

3. The method according to claim 2, characterized in that, The construction of the reward function based on the equipment waiting time cost and the secondary disturbance cost of the freshly mixed concrete specifically includes: Extract the time difference between the moment when the tower crane completes material supply to the current construction unit and the moment when the ground leveling robot enters the current construction unit to start leveling, and use it as the process interval time; When the process interval time is within the first preset range, the equipment waiting time cost is calculated based on the sum of the process interval time and the tower crane's no-load slewing time. When the process interval exceeds the first preset range and is within the concrete workable time window constraint, the concrete slump loss value corresponding to the process interval is calculated based on the concrete rheological property curve, and the concrete slump loss value is mapped to the additional energy consumption cost of the ground leveling robot as the secondary disturbance cost of the fresh concrete. The reward function is constructed by weighted summing of the equipment waiting time cost and the secondary disturbance cost of the fresh concrete, and taking the negative value of the sum.

4. The method according to claim 1, characterized in that, The step of extracting the concrete workable time window constraint for each construction unit based on the updated shared state vector specifically includes: The current environmental status data of the target construction unit is extracted from the shared state vector. The current environmental status data includes real-time ambient temperature, ambient humidity and wind speed collected by the construction site Internet of Things sensors. The current environmental state data and the concrete mix proportion parameters of the target construction unit are input into a preset concrete hydration kinetic model to calculate the initial setting time of the concrete in the target construction unit. Using the moment when the tower crane completes the material feeding action as the starting boundary and the initial setting time of the concrete as the ending boundary, the workable time window constraint of the concrete for the target construction unit is generated.

5. The method according to claim 4, characterized in that, After calculating the initial setting time of the concrete for the target construction unit, the method further includes: The real-time resistance feedback data of the ground leveling robot when it performs the leveling action in an adjacent already-operated construction unit is obtained. The real-time resistance feedback data includes the drive wheel torque of the ground leveling robot and the working current of the leveling mechanism. The real-time resistance feedback data is compared with the theoretical condensation state of the adjacent completed construction units to calculate the condensation state deviation value. Based on the aforementioned setting state deviation value, the environmental impact coefficient in the preset concrete hydration kinetics model is compensated to correct the initial setting time node of the concrete in subsequent unworked construction units.

6. The method according to claim 1, characterized in that, If an abnormal disturbance event deviating from the shared state vector is detected on-site, the joint timing decision is re-triggered based on the current state of the affected construction unit to update the material supply task queue and the work task queue, specifically including: Real-time monitoring of operational data at the construction site; if the avoidance delay triggered by the tower crane anti-collision system, the power warning of the ground leveling robot, or a sudden weather change is detected, it is determined that an abnormal disturbance event deviating from the shared state vector has occurred. Calculate the expected time delay caused by the abnormal disturbance event, and compare the expected time delay with the concrete workable time window constraint of the affected construction unit to evaluate the spatiotemporal impact range of the abnormal disturbance event. Based on the spatiotemporal influence range, the joint timing decision is retried to update the material supply task queue and the operation task queue.

7. The method according to claim 6, characterized in that, The step of re-triggering the joint timing decision based on the spatiotemporal influence range to update the material supply task queue and the job task queue specifically includes: If the spatiotemporal impact range does not exceed the concrete workable time window constraint of the affected construction unit, then local timing fine-tuning is performed, keeping the original task allocation order unchanged, and only updating the expected execution timestamps of each task in the material supply task queue and the work task queue. If the spatiotemporal impact range exceeds the concrete workable time window constraint of the affected construction unit, then the currently unexecuted material supply task queue and the work task queue are frozen, the affected construction unit is marked as a high-priority emergency state, and a global replanning is performed based on the current shared state vector to regenerate the material supply task queue and the work task queue.

8. A collaborative scheduling and control device for tower cranes, characterized in that, The tower crane collaborative scheduling and control device includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the tower crane collaborative scheduling and control device to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the tower crane collaborative scheduling and control device, the tower crane collaborative scheduling and control device performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the tower crane collaborative scheduling and control equipment, the tower crane collaborative scheduling and control equipment performs the method as described in any one of claims 1-7.

Images