Road engineering construction resource dynamic optimization scheduling system based on Internet of Things and implementation method

Abstract

The invention discloses a highway engineering construction resource dynamic optimization scheduling system based on the Internet of Things and an implementation method, and the system comprises a multi-dimensional perception layer, an edge calculation transmission layer, an intelligent decision scheduling layer and an execution feedback layer, and each layer forms a perception-calculation-decision-execution-feedback closed-loop scheduling system. The multi-dimensional sensing layer is deployed on a highway construction area and various construction resources; the edge calculation transmission layer performs real-time preprocessing and local calculation on the multi-source data acquired by the sensing layer; a dynamic optimization scheduling model based on reinforcement learning is built in the intelligent decision scheduling layer, and a resource scheduling optimization scheme is output; the execution feedback layer is used for driving the construction resources to execute the scheduling instruction. According to the method, the core problems of one-sided perception, calculation lag, decision-making rigidity, feedback deficiency and the like in a traditional mode are solved, road construction resource scheduling is promoted to be transformed from experience-driven to data-driven and intelligent decision-making, and finally optimal configuration of construction resources and efficient management and control of the construction process are achieved.

Description

Technical Field

[0001] This invention relates to the field of highway engineering construction management, and more specifically to a dynamic optimization scheduling system and implementation method for highway engineering construction resources based on the Internet of Things. Background Technology

[0002] Highway construction is characterized by dispersed work areas, complex and variable construction environments, diverse resource types (including construction equipment, materials, and personnel), and high dynamism. The efficiency of resource allocation directly determines the project schedule, cost control, operational safety, and overall construction quality. With the continuous expansion of highway construction scale and the popularization of intelligent construction concepts, traditional highway construction resource allocation models have gradually exposed many unavoidable technical bottlenecks, failing to meet the demands of modern construction for efficient, precise, and dynamic allocation.

[0003] At the level of construction resource perception, existing scheduling models largely rely on manual inspections, single-point monitoring, or data collection from single types of sensors, resulting in incomplete perception dimensions and narrow data coverage. For example, monitoring of the construction environment is often limited to basic meteorological data, lacking accurate acquisition of key environmental parameters such as terrain slope, road surface smoothness, temporary obstacles, and underground pipelines; monitoring of construction equipment is mostly limited to location positioning or simple operational status feedback, making it difficult to achieve real-time capture of equipment workload, fuel / electricity reserves, and potential faults; material tracking relies heavily on ledger records, failing to achieve dynamic positioning and information traceability throughout the entire material flow process; and the management of workers lacks real-time monitoring of their physiological state and labor intensity, leading to delayed safety warnings. This singular and fragmented perception method results in incomplete and inaccurate data collection during construction, providing weak basic data support for subsequent scheduling decisions.

[0004] In the data processing and transmission stage, traditional methods often involve directly uploading sensor data to a remote cloud for centralized processing, without establishing localized real-time processing nodes. Since highway construction areas are frequently located in remote regions with unstable network coverage, 5G signals are easily attenuated due to terrain obstruction. Direct transmission of large amounts of unstructured data (such as images and videos) not only consumes extremely high network bandwidth but also significantly reduces data timeliness due to transmission delays. Furthermore, existing transmission methods often use fixed protocols, failing to dynamically adjust transmission strategies based on network quality. When network signals are weak, data loss or transmission interruptions are likely, further impacting the timely issuance of scheduling instructions. In addition, the collected data lacks targeted preprocessing, resulting in abundant redundant information and significant noise interference, directly affecting the effectiveness of subsequent data applications.

[0005] At the scheduling decision-making level, existing scheduling schemes are mostly based on experience or static model calculations, lacking the ability to dynamically respond to real-time data during construction. Traditional decision-making models often optimize only around a single objective (such as schedule or cost), failing to build a multi-dimensional optimization system encompassing "time-cost-efficiency-safety," and particularly neglecting the core weight of construction safety in scheduling decisions. When faced with sudden construction scenarios such as equipment failure, material shortages, and extreme weather, the model response is lagging, unable to quickly generate adaptive adjustment schemes. Furthermore, existing models lack an effective mechanism for transferring historical experience, requiring lengthy debugging periods in the initial deployment of new construction scenarios to achieve ideal scheduling results, indicating insufficient adaptability and self-learning capabilities.

[0006] In the execution and feedback phases, the existing scheduling model's "decision-execution" link is mostly unidirectional, lacking effective execution status feedback and closed-loop optimization mechanisms. After scheduling instructions are issued, it is impossible to collect resource execution status and actual effects in real time, making it difficult to dynamically verify the rationality of the scheduling plan. When deviations occur between the actual execution status and the scheduling plan, the model cannot be promptly re-optimized, leading to the continuous accumulation of scheduling errors and further impacting construction efficiency and quality. In addition, existing safety early warning mechanisms are mostly independent modules, disconnected from the resource scheduling system. When personnel safety hazards, equipment conflicts, or environmental risks are identified, they can only issue simple early warning signals, failing to link with the scheduling model to generate targeted risk avoidance and adjustment plans, resulting in limited safety assurance capabilities.

[0007] In summary, the current field of highway construction resource scheduling urgently needs an integrated scheduling system capable of multi-dimensional accurate perception, real-time efficient data processing, intelligent dynamic decision-making, and closed-loop optimization. This system would address core issues in traditional methods, such as incomplete perception, computational lag, rigid decision-making, and lack of feedback. It would also drive the transformation of highway construction resource scheduling from "experience-driven" to "data-driven, intelligent decision-making," ultimately achieving optimal allocation of construction resources and efficient management of the construction process. Therefore, developing an IoT-based dynamic optimization scheduling system for highway construction resources has significant engineering practical value and technological innovation implications. Summary of the Invention

[0008] Therefore, to address the aforementioned shortcomings of existing systems, this invention provides an Internet of Things-based dynamic optimization scheduling system and implementation method for highway construction resources. This system solves core problems in traditional methods such as incomplete perception, lagging computation, rigid decision-making, and lack of feedback, promoting the transformation of highway construction resource scheduling from "experience-driven" to "data-driven, intelligent decision-making," ultimately achieving optimal allocation of construction resources and efficient management of the construction process.

[0009] This invention is implemented by constructing a dynamic optimization scheduling system for highway engineering construction resources based on the Internet of Things (IoT). The system comprises a multi-dimensional perception layer, an edge computing transmission layer, an intelligent decision-making scheduling layer, and an execution feedback layer. Each layer interacts with data and transmits instructions through IoT protocols, forming a closed-loop scheduling system of perception-computation-decision-execution-feedback. The multi-dimensional perception layer is deployed in the highway construction area and on various construction resources to collect construction environment parameters, resource status data and construction progress information; The edge computing transmission layer performs real-time preprocessing and local computation on the multi-source data collected by the perception layer, filters key data and uploads it to the intelligent decision scheduling layer through an adaptive transmission protocol, and at the same time receives scheduling instructions and sends them to the execution feedback layer. The intelligent decision-making and scheduling layer has a built-in dynamic optimization scheduling model based on reinforcement learning. It combines preprocessed real-time data with construction plan baseline data to output resource scheduling optimization schemes. The execution feedback layer is used to drive the execution of scheduling instructions for construction resources and collect resource execution status data in real time and feed it back to the multi-dimensional perception layer to realize dynamic verification and adjustment of scheduling effect.

[0010] Furthermore, the multi-dimensional perception layer includes environmental perception units, equipment perception units, material perception units, and personnel perception units; The environmental perception unit uses a combination of millimeter-wave radar and high-definition camera to collect terrain slope, road surface flatness, real-time meteorological data and temporary obstacle information of the construction area. The millimeter-wave radar is used to acquire three-dimensional terrain data, and the high-definition camera, together with image recognition algorithm, can identify obstacle types. The equipment sensing unit collects real-time location, workload, fuel / electricity remaining, fault diagnosis codes, and work efficiency data of the equipment through an IoT terminal installed on the construction machinery. The IoT terminal integrates a Beidou-3 positioning module and a CAN bus data acquisition module to achieve full-dimensional monitoring of equipment status. The material sensing unit uses a combination of RFID tags and ultra-wideband positioning. The RFID tags have built-in information on material type, specifications, quantity and storage location, while the ultra-wideband positioning module collects real-time location data of material transport vehicles and on-site material piles, enabling full traceability of material flow. The personnel sensing unit collects real-time data on the worker's location, heart rate, body temperature, and labor intensity through the positioning module and physiological state monitoring sensors built into the smart safety helmet. When physiological parameters exceed safety thresholds, an early warning is automatically triggered.

[0011] Furthermore, the edge computing transport layer includes edge computing nodes and an adaptive transport module; Edge computing nodes are deployed around the construction area and adopt a heterogeneous computing architecture. They extract edge features from unstructured data collected by the perception layer, clean, denoise and standardize the structured data, remove redundant data and retain valid data with a confidence level of ≥95%, thereby reducing data transmission bandwidth consumption. The adaptive transmission module has a built-in transmission protocol switching mechanism, and at the same time, it reduces the amount of transmitted data by 40%-60% through data compression algorithms.

[0012] Furthermore, the intelligent decision-making and scheduling layer includes a data fusion module, a dynamic optimization scheduling model, and a scheme output module; The data fusion module uses DS evidence theory to fuse multi-source data uploaded from the edge computing transmission layer. Combined with the baseline construction period, resource allocation standards and cost budget data in the construction plan database, it constructs a scheduling decision matrix that includes three-dimensional objectives of "time-cost-efficiency". The dynamic optimization scheduling model is based on the DQN algorithm in reinforcement learning and introduces a construction scenario adaptation factor to improve the reward function: the reduction in resource idle rate, the duration of early / delayed construction period, the amount of cost savings, and the incidence of safety accidents are used as core reward parameters, with the incidence of safety accidents having the highest weight (0.4), followed by the construction period parameter (0.3), the cost parameter having a weight of 0.2, and the efficiency parameter having a weight of 0.1. The model updates the state space through real-time data iteration to achieve online self-learning and dynamic optimization of the scheduling strategy. When the construction scenario undergoes sudden changes (such as equipment failure or material shortage), the model response time is ≤500ms. The solution output module transforms the optimized scheduling strategy into structured instructions, including equipment scheduling instructions (work area allocation, operating parameter adjustment), material scheduling instructions (transportation route planning, replenishment time nodes), and personnel scheduling instructions (job responsibility allocation, shift arrangement), and simultaneously generates a visual report of the scheduling solution.

[0013] Furthermore, the execution feedback layer includes an instruction execution unit and a status feedback unit; The instruction execution unit connects with the construction equipment control system, material transportation management system, and personnel management system through an Internet of Things terminal, converting scheduling instructions into equipment control signals (such as excavator operating range and transport vehicle speed) and management instructions, wherein the execution accuracy error of the equipment control signals is ≤2%; The status feedback unit collects resource status data in real time after the command is executed, including actual equipment operating parameters, actual material transportation volume and actual personnel working status. It feeds back to the intelligent decision-making and scheduling layer through the edge computing transmission layer to form a scheduling closed loop. When the actual execution status deviates from the scheduling plan by ≥10%, the model is triggered to re-optimize the process.

[0014] An implementation method for the above-mentioned Internet of Things-based dynamic optimization scheduling system for highway construction resources is characterized by the following steps: Step 1: Multi-dimensional perception data collection; Deploy perception devices in the highway construction area and construction resources to simultaneously collect construction environment parameters, resource status data and construction progress information to form multi-source raw data; Step 2: Edge data processing and transmission; The edge computing nodes around the construction area perform real-time preprocessing on the multi-source raw data, filter out valid data with a confidence level of ≥95%, and then the adaptive transmission module dynamically selects the transmission protocol according to the network quality and uploads it to the intelligent decision-making terminal, while receiving the scheduling instructions issued by the intelligent decision-making terminal. Step 3: Intelligent decision-making generates scheduling schemes; the intelligent decision-making terminal processes the uploaded data through data fusion technology, combines the construction plan baseline data to construct a three-dimensional decision matrix of "time-cost-efficiency", and uses a dynamic optimization scheduling model based on reinforcement learning to output a structured resource scheduling optimization scheme; Step 4: Execution feedback and closed-loop optimization; drive the construction resource execution scheduling plan, collect resource execution status data in real time and feed it back to the perception link to verify the scheduling effect. When the actual execution deviates from the plan by ≥10%, the model is triggered to re-optimize, forming a closed-loop scheduling system of perception-computation-decision-execution-feedback.

[0015] Furthermore, the multi-dimensional sensing data acquisition described in step 1 of the above method specifically includes: 1) Environmental perception: A combination of millimeter-wave radar and high-definition camera is used. The millimeter-wave radar acquires three-dimensional terrain data of the construction area to extract terrain slope and road surface smoothness parameters. The high-definition camera, in conjunction with image recognition algorithm, collects real-time meteorological data and temporary obstacle information and identifies obstacle types. 2) Equipment sensing: Through an IoT terminal that integrates a Beidou-3 positioning module and a CAN bus data acquisition module, the real-time location, workload, remaining fuel / electricity, fault diagnosis codes, and work efficiency data of the construction equipment are collected; 3) Material sensing: The system combines RFID tags with ultra-wideband positioning. The RFID tags store information on material type, specifications, quantity, and storage location, while the ultra-wideband positioning module collects real-time location data of material transport vehicles and on-site material piles. 4) Personnel perception: Through the positioning module and physiological sensors built into the smart safety helmet, real-time location, heart rate, body temperature and labor intensity data of the workers are collected. When the physiological parameters exceed the safety threshold, a local warning is automatically triggered. The environmental perception also includes: collecting soil moisture content data in the roadbed operation area through a soil moisture sensor, and locating the underground pipeline position through an electromagnetic induction underground pipeline detection module with a positioning error of ≤5cm.

[0016] Furthermore, the edge data processing described in step 2 of the above method specifically includes: The edge computing nodes adopt a heterogeneous computing architecture to extract edge features from unstructured data (images, videos) and clean, denoise, and standardize the format of structured data (device parameters, environmental values) to remove redundant data; the effective data screening is based on a data confidence level of ≥95%. The adaptive transmission described in step 2 specifically includes: The adaptive transmission module has a built-in protocol switching mechanism. When the 5G signal strength is ≥-85dBm, it uses the 5G protocol for transmission; when the 5G signal strength is -100dBm≤5G signal strength<-85dBm, it switches to industrial Ethernet; when the network signal strength is <-100dBm, it uses the LoRaWAN protocol for transmission; and before transmission, the data volume is reduced by 40%-60% through a data compression algorithm.

[0017] Furthermore, the data fusion technology described in step 3 of the above method is the DS evidence theory. This theory is used to fuse edge transmission data with baseline construction period, resource allocation standards, and cost budget data in the construction plan database to construct a three-dimensional decision matrix. The dynamic optimization scheduling model described in step 3 is based on the reinforcement learning DQN algorithm, and introduces a construction scenario adaptation factor to improve the reward function: The core parameters of the reward function include the reduction in resource idle rate, the duration of early / delayed construction, the amount of cost savings, and the accident rate. The weights of each parameter are as follows: accident rate 0.4, construction period parameter 0.3, cost parameter 0.2, and efficiency parameter 0.1. The model updates the state space iteratively through real-time data, and the response time is ≤500ms when there are sudden changes in the construction scenario. Before deployment, the dynamic optimization scheduling model described in step 3 needs to be pre-trained using a historical data training module based on a database containing 500+ typical highway construction cases. The transfer learning algorithm is then used to transfer historical scheduling experience to new construction scenarios, thereby improving the initial scheduling optimization effect of new scenarios by more than 30%. The structured resource scheduling optimization scheme described in step 3 includes: equipment scheduling instructions (work area allocation, operating parameter adjustment), material scheduling instructions (transportation route planning, replenishment time nodes), and personnel scheduling instructions (job responsibility allocation, shift arrangement), and simultaneously generates a visual scheduling report.

[0018] Furthermore, step 4 of the above method specifically includes: converting the scheduling instructions into equipment control signals and management instructions through an IoT terminal, with the execution accuracy error of the equipment control signals being ≤2%; the status feedback is achieved by collecting actual operating parameters of the equipment, actual material transportation volume, and actual working status of personnel, and the feedback data is uploaded to the intelligent decision-making terminal via the edge computing transmission layer.

[0019] The present invention has the following advantages: The ingenuity of this system, combined with the characteristics of highway engineering construction scenarios, is mainly reflected in the following aspects: First, the closed-loop architecture reconstructs the construction scheduling logic, breaking through the limitations of the traditional decentralized model. Due to the core challenges of dispersed geographical locations, resource mobility, and variable environments in highway construction, a closed-loop system of "multi-dimensional perception - edge computing - intelligent decision-making - execution feedback" has been established, integrating previously isolated equipment monitoring, material management, and personnel scheduling. Seamless data flow is achieved across all layers through IoT protocols, solving the problems of "lagging perception data and disconnect between decision-making and execution" in traditional construction. This forms a dynamic and adaptive scheduling closed loop, adapting to the linear operations and multi-process collaboration requirements of highway construction.

[0020] Secondly, the multi-dimensional perception layer enables precise capture of all construction elements, filling the gaps in traditional monitoring. Comprehensive Perception Coverage: Integrating four major perception units—environment, equipment, materials, and personnel—this approach breaks through the limitations of traditional construction methods that focus solely on equipment or safety monitoring. For example, the environmental perception unit combines millimeter-wave radar (3D terrain) with high-definition cameras (obstacle recognition), adapting to the complex terrain and numerous temporary obstacles encountered during highway construction. The underground pipeline detection module (positioning error ≤5cm) accurately avoids common pipeline damage risks encountered during municipal / highway construction.

[0021] Scenario-based adaptation of sensing technology: The equipment sensing unit integrates Beidou-3 and CAN bus to solve the monitoring problem of wide movement range and many status parameters of highway construction equipment; Material sensing adopts RFID + ultra-wideband positioning to realize full traceability of bulk / large materials such as steel bars and cement in long-distance transportation (large span between highway construction material yard and work surface), solving the problem of "inconsistency between accounts and actual goods and delayed supply" of traditional materials.

[0022] Third, edge computing combined with intelligent decision-making solves the core bottleneck of "poor network and slow response" in highway construction. Edge computing adapts to field construction scenarios: Heterogeneous computing nodes are deployed in the construction area to preprocess unstructured data such as images and videos locally, and only valid data with a confidence level of ≥95% is uploaded, reducing bandwidth dependence in remote areas of highway construction; Adaptive transmission protocols (dynamic switching between 5G / Industrial Ethernet / LoRaWAN) solve the problem of network signal fluctuations in construction scenarios such as mountainous areas and tunnels, ensuring the stability of data transmission.

[0023] The reinforcement learning model focuses on the core construction objectives: based on the DQN algorithm, it introduces a construction-oriented reward function of "safety (weight 0.4) - schedule (0.3) - cost (0.2) - efficiency (0.1)", which is in line with the management requirements of "safety first and schedule rigidity" in highway construction; the model response time is ≤500ms, which can quickly respond to sudden situations in highway construction such as equipment failure and material shortage, far exceeding the efficiency of traditional manual scheduling.

[0024] Transfer learning lowers the deployment threshold for new scenarios: Based on a pre-trained model with 500+ highway construction cases, historical experience can be reused for new road sections through transfer learning, improving the initial optimization effect of new scenarios by more than 30% and solving the problems of "highly personalized and long model implementation cycle" in highway construction projects.

[0025] Fourth, implement an execution-feedback linkage mechanism to ensure the accurate implementation of dispatching plans. High-precision execution control: The execution unit converts scheduling instructions into equipment control signals (error ≤2%), adapting to the precision operation requirements of highway construction equipment such as excavators and road rollers (such as subgrade compaction control). Dynamic deviation correction: When the actual execution deviates from the plan by ≥10% (such as insufficient material transportation), the model is automatically re-optimized to solve the common problem of "disconnect between plan and reality" in highway construction and avoid resource waste or project delays.

[0026] Fifth, coordinate safety early warning and dispatch to build a construction safety defense line. The safety early warning subsystem does not issue warnings in isolation, but rather works in conjunction with the scheduling model. When risks such as equipment collisions or abnormal physiological conditions of personnel are identified, it triggers both audible and visual alarms and automatically generates a risk avoidance scheduling plan (such as instructing excavators to avoid the danger). This breaks through the limitations of the traditional highway construction model of "separation of early warning and response" and constructs a safety closed loop of "early warning-decision-risk avoidance". Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the overall system architecture of this application; Figure 2 This is a schematic diagram of the multi-dimensional perception layer and edge computing transmission layer in this system; Figure 3 This is a schematic diagram of the intelligent decision-making scheduling layer and the execution feedback layer in this system; Figure 4 This is a schematic diagram illustrating the execution process of the implementation method in this application. Detailed Implementation

[0028] The following will be combined with the appendix Figures 1-4This invention will be described in detail, and the technical solutions in the embodiments of this invention will be clearly and completely described. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0029] Example 1: This invention provides a dynamic optimization scheduling system for highway engineering construction resources based on the Internet of Things, such as... Figures 1-3 As shown, it can be implemented in the following manner; the system includes a multi-dimensional perception layer 10, an edge computing transmission layer 20, an intelligent decision scheduling layer 30, and an execution feedback layer 40. Each layer realizes data interaction and command transmission through the Internet of Things protocol, forming a closed-loop scheduling system of perception-computation-decision-execution-feedback; it is described in detail below; The multi-dimensional perception layer is deployed in the highway construction area and on various construction resources to collect construction environment parameters, resource status data and construction progress information. The construction resources include construction equipment, materials and workers. The edge computing transmission layer performs real-time preprocessing and local computation on the multi-source data collected by the perception layer, filters key data and uploads it to the intelligent decision scheduling layer through an adaptive transmission protocol, and at the same time receives scheduling instructions and sends them to the execution feedback layer. The intelligent decision-making and scheduling layer has a built-in dynamic optimization scheduling model based on reinforcement learning. It combines preprocessed real-time data with construction plan baseline data to output resource scheduling optimization schemes. The execution feedback layer is used to drive the execution of scheduling instructions for construction resources and collect resource execution status data in real time and feed it back to the multi-dimensional perception layer to realize dynamic verification and adjustment of scheduling effect.

[0030] In this embodiment, the multi-dimensional perception layer includes an environmental perception unit, an equipment perception unit, a material perception unit, and a personnel perception unit; The environmental perception unit uses a combination of millimeter-wave radar and high-definition camera to collect terrain slope, road surface flatness, real-time meteorological data and temporary obstacle information of the construction area. The millimeter-wave radar is used to acquire three-dimensional terrain data, and the high-definition camera, together with image recognition algorithm, realizes obstacle type identification. The equipment sensing unit collects real-time location, workload, remaining fuel / electricity, fault diagnosis codes, and work efficiency data of the equipment through an IoT terminal installed on the construction machinery. The IoT terminal integrates a Beidou-3 positioning module and a CAN bus data acquisition module to achieve full-dimensional monitoring of the equipment status. The material sensing unit adopts a combination of RFID tags and ultra-wideband positioning. The RFID tags have built-in information on material type, specifications, quantity and storage location. The ultra-wideband positioning module collects the location data of material transport vehicles and on-site material piles in real time, realizing full traceability of material flow. The personnel sensing unit collects real-time location, heart rate, body temperature, and labor intensity data of workers through the positioning module and physiological state monitoring sensor built into the smart safety helmet. When physiological parameters exceed safety thresholds, an early warning is automatically triggered.

[0031] In this embodiment, the edge computing transport layer includes edge computing nodes and an adaptive transport module; The edge computing nodes are deployed around the construction area and adopt a heterogeneous computing architecture. They perform edge feature extraction on the unstructured data (images and videos) collected by the perception layer, and clean, denoise and standardize the structured data (equipment parameters and environmental values). Redundant data is removed and valid data with a confidence level of ≥95% is retained, thereby reducing the data transmission bandwidth usage. The adaptive transmission module has a built-in transmission protocol switching mechanism that dynamically selects 5G, industrial Ethernet, or LoRaWAN protocol based on the network quality of the construction area: when the 5G signal strength is ≥-85dBm, the 5G protocol is used to achieve high-speed data transmission; when the 5G signal strength is <-85dBm but ≥-100dBm, it switches to industrial Ethernet; when the network signal strength is <-100dBm, the LoRaWAN protocol is used to ensure stable transmission of low-speed data, while reducing the amount of transmitted data by 40%-60% through data compression algorithms.

[0032] In this embodiment, the intelligent decision-making and scheduling layer includes a data fusion module, a dynamic optimization scheduling model, and a scheme output module; The data fusion module uses DS evidence theory to fuse multi-source data uploaded from the edge computing transmission layer. Combined with the baseline construction period, resource allocation standards and cost budget data in the construction plan database, it constructs a scheduling decision matrix that includes three-dimensional objectives of "time-cost-efficiency". The dynamic optimization scheduling model is based on the DQN algorithm in reinforcement learning and introduces a construction scenario adaptation factor to improve the reward function: the reduction in resource idle rate, the duration of early / delayed construction, the amount of cost savings, and the accident rate are used as core reward parameters, with the accident rate having the highest weight (0.4), followed by the construction period parameter (0.3), the cost parameter having a weight of 0.2, and the efficiency parameter having a weight of 0.1. The model updates the state space through real-time data iteration to achieve online self-learning and dynamic optimization of the scheduling strategy. When the construction scenario undergoes sudden changes (such as equipment failure or material shortage), the model response time is ≤500ms. The output module of the scheme transforms the optimized scheduling strategy into structured instructions, including equipment scheduling instructions (work area allocation, operating parameter adjustment), material scheduling instructions (transportation route planning, replenishment time nodes) and personnel scheduling instructions (job responsibility allocation, shift arrangement), and simultaneously generates a visual report of the scheduling scheme.

[0033] The intelligent decision-making and scheduling layer also includes a historical data training module. This module pre-trains the dynamic optimization scheduling model based on the historical database of highway engineering construction (containing 500+ typical construction cases). Through transfer learning algorithms, the scheduling experience in the historical cases is transferred to the new construction scenario, which improves the scheduling optimization effect of the model by more than 30% in the initial deployment stage of the new scenario.

[0034] In this embodiment, the execution feedback layer includes an instruction execution unit and a status feedback unit; The instruction execution unit connects with the construction equipment control system, material transportation management system, and personnel management system through an Internet of Things terminal, converting scheduling instructions into equipment control signals (such as excavator operating range and transport vehicle speed) and management instructions, wherein the execution accuracy error of the equipment control signals is ≤2%; The status feedback unit collects resource status data in real time after the command is executed, including actual equipment operating parameters, actual material transportation volume and actual personnel working status. It feeds back to the intelligent decision-making and scheduling layer through the edge computing transmission layer to form a scheduling closed loop. When the actual execution status deviates from the scheduling plan by ≥10%, the model is triggered to re-optimize the process.

[0035] In this embodiment, the environmental sensing unit also integrates a soil moisture sensor and an underground pipeline detection module. The soil moisture data is used to optimize the construction parameters of the earthmoving equipment, and the underground pipeline detection module uses electromagnetic induction technology to locate the underground pipeline, with a positioning error of ≤5cm, thus avoiding pipeline damage during construction.

[0036] Example 2, as Figure 1 As shown, based on the system of the above-described embodiment 1, a safety early warning subsystem 50 is also included. This subsystem is based on the data collected by the multi-dimensional perception layer and combined with the risk assessment model of the intelligent decision-making and scheduling layer. When construction resource conflicts (such as equipment collision risks), personnel safety hazards (such as abnormal physiological states) or environmental risks (such as extreme weather) are identified, an early warning signal is issued through the sound and light alarm device and the intelligent safety helmet terminal, and at the same time, the scheduling model is triggered to generate a risk avoidance adjustment plan.

[0037] Example 3, an implementation method of the above-mentioned Internet of Things-based dynamic optimization scheduling system for highway construction resources, includes the following steps: See Figure 4 As shown; Step 1: Multi-dimensional perception data collection - Deploy perception devices in the highway construction area and construction resources (including construction equipment, materials, and workers) to simultaneously collect construction environment parameters, resource status data, and construction progress information, forming multi-source raw data; Step 2: Edge data processing and transmission - The edge computing nodes around the construction area perform real-time preprocessing of multi-source raw data, filter out valid data with a confidence level of ≥95%, and then the adaptive transmission module dynamically selects the transmission protocol according to the network quality to upload to the intelligent decision-making terminal, while receiving the scheduling instructions issued by the intelligent decision-making terminal. Step 3: Intelligent decision-making generates scheduling schemes—The intelligent decision-making terminal processes the uploaded data through data fusion technology, combines the construction plan baseline data to construct a three-dimensional decision matrix of "time-cost-efficiency", and uses a dynamic optimization scheduling model based on reinforcement learning to output a structured resource scheduling optimization scheme; Step 4: Execution Feedback and Closed-Loop Optimization – Drive the execution and scheduling of construction resources, collect resource execution status data in real time and feed it back to the perception stage to verify the scheduling effect. When the actual execution deviates from the plan by ≥10%, the model is triggered to re-optimize, forming a closed-loop scheduling system of perception-computation-decision-execution-feedback.

[0038] The multi-dimensional sensing data acquisition described in step 1 of Example 3 specifically includes: 1) Environmental perception: A combination of millimeter-wave radar and high-definition camera is used. The millimeter-wave radar acquires three-dimensional terrain data of the construction area to extract terrain slope and road surface smoothness parameters. The high-definition camera, in conjunction with image recognition algorithm, collects real-time meteorological data and temporary obstacle information and identifies obstacle types. 2) Equipment sensing: Through an IoT terminal that integrates a Beidou-3 positioning module and a CAN bus data acquisition module, the real-time location, workload, remaining fuel / electricity, fault diagnosis codes, and work efficiency data of the construction equipment are collected; 3) Material sensing: The system combines RFID tags with ultra-wideband positioning. The RFID tags store information on material type, specifications, quantity, and storage location, while the ultra-wideband positioning module collects real-time location data of material transport vehicles and on-site material piles. 4) Personnel perception: Through the positioning module and physiological sensors built into the smart safety helmet, real-time location, heart rate, body temperature and labor intensity data of the workers are collected. When the physiological parameters exceed the safety threshold, a local warning is automatically triggered. The environmental perception also includes: collecting soil moisture content data in the roadbed operation area through a soil moisture sensor, and locating the underground pipeline position through an electromagnetic induction underground pipeline detection module with a positioning error of ≤5cm.

[0039] The edge data processing described in step 2 of Example 3 specifically includes: The edge computing nodes adopt a heterogeneous computing architecture to extract edge features from unstructured data (images, videos) and clean, denoise, and standardize the format of structured data (device parameters, environmental values) to remove redundant data; the effective data screening is based on a data confidence level of ≥95%. The adaptive transmission described in step 2 specifically includes: The adaptive transmission module has a built-in protocol switching mechanism. When the 5G signal strength is ≥-85dBm, it uses the 5G protocol for transmission; when the 5G signal strength is -100dBm≤5G signal strength<-85dBm, it switches to industrial Ethernet; when the network signal strength is <-100dBm, it uses the LoRaWAN protocol for transmission; and before transmission, the data volume is reduced by 40%-60% through a data compression algorithm.

[0040] The data fusion technology described in step 3 of Example 3 is the DS evidence theory. This theory is used to fuse edge transmission data with baseline construction period, resource allocation standards, and cost budget data in the construction plan database to construct a three-dimensional decision matrix. The dynamic optimization scheduling model described in step 3 is based on the reinforcement learning DQN algorithm, and introduces a construction scenario adaptation factor to improve the reward function: The core parameters of the reward function include the reduction in resource idle rate, the duration of early / delayed construction, the amount of cost savings, and the accident rate. The weights of each parameter are as follows: accident rate 0.4, construction period parameter 0.3, cost parameter 0.2, and efficiency parameter 0.1. The model updates the state space iteratively through real-time data, and the response time is ≤500ms when there are sudden changes in the construction scenario. Before deployment, the dynamic optimization scheduling model described in step 3 needs to be pre-trained using a historical data training module based on a database containing 500+ typical highway construction cases. The transfer learning algorithm is then used to transfer historical scheduling experience to new construction scenarios, thereby improving the initial scheduling optimization effect of new scenarios by more than 30%. The structured resource scheduling optimization scheme described in step 3 includes: equipment scheduling instructions (work area allocation, operating parameter adjustment), material scheduling instructions (transportation route planning, replenishment time nodes), and personnel scheduling instructions (job responsibility allocation, shift arrangement), and simultaneously generates a visual scheduling report.

[0041] The execution of scheduling instructions in step 4 of Example 3 specifically includes: converting scheduling instructions into equipment control signals and management instructions through an IoT terminal, with the execution accuracy error of the equipment control signals being ≤2%; the status feedback is achieved by collecting actual equipment operating parameters, actual material transport volume, and actual personnel working status, and the feedback data is uploaded to the intelligent decision-making terminal via the edge computing transmission layer.

[0042] In the implementation of Example 3, a safety linkage early warning step is also included: based on the data collected in Step 1 and the risk assessment model of the intelligent decision-making terminal, when construction resource conflicts, personnel safety hazards or environmental risks are identified, an early warning signal is issued through the sound and light alarm device and the intelligent safety helmet terminal, and the dynamic optimization scheduling model is triggered to generate a risk avoidance adjustment plan.

[0043] The following describes the patent implementation process and application of the Internet of Things-based dynamic optimization scheduling system for highway construction resources. I. Patent Implementation Process The implementation process of the system involved in this patent follows the core logic of "technology decomposition - module development - integration and debugging - verification and deployment". It revolves around four core levels: multi-dimensional perception, edge computing transmission, intelligent decision-making and scheduling, and execution feedback, as well as supporting subsystems. The technology transformation and engineering deployment are completed in stages, and the specific process is as follows: (I) Preliminary preparations and needs breakdown Scenario Requirements Survey and Parameter Determination: Considering the diversity of highway construction projects (such as expressway reconstruction and expansion, and new rural road construction), data on construction specifications, resource allocation benchmarks, and environmental interference factors under different scenarios were collected to clarify the core indicator thresholds for each scenario. For example, in municipal highway construction, it was determined that the accuracy of underground pipeline detection must be ≤5cm; in mountainous highway construction, the triggering conditions for extreme weather warnings and the environmental adaptability parameters for equipment operation were determined. Simultaneously, by combining over 500 typical historical construction cases, common requirements (such as equipment idle rate control and schedule assurance) and personalized requirements (such as efficiency optimization for special terrain) were extracted to provide a basis for system module design.

[0044] Technical Standards and Protocol Adaptation Planning: Establish unified standards for data interaction at all levels, clarify the selection of IoT protocols (e.g., MQTT protocol for sensing layer devices, HTTP / 2 protocol for edge computing and decision layers); determine data format specifications (JSON format for structured data, H.265 encoding standard for unstructured data); define acceptance criteria for key technical parameters, such as model response time ≤ 500ms, device control accuracy error ≤ 2%, data transmission compression ratio 40%-60%, etc., to ensure consistency in the development direction of each module.

[0045] Hardware and software selection and supply chain setup: Based on the requirements of the perception layer, suitable hardware devices were selected, such as millimeter-wave radar (selected to support 3D terrain data acquisition, ranging accuracy ±0.1m), BeiDou-3 positioning modules (supporting centimeter-level positioning), RFID tags (UHF passive tags, reading distance ≥5m), and smart safety helmets (integrating heart rate sensors and LoRaWAN communication modules); edge computing nodes used heterogeneous computing servers (equipped with CPU+GPU architecture, supporting real-time image feature extraction); the decision layer adopted a cloud computing platform (supporting distributed training and high-concurrency processing). Simultaneously, partnerships were established with equipment manufacturers and telecommunications operators to ensure hardware supply and network coverage.

[0046] (II) Core Module Development and Unit Testing 1. Development of a multi-dimensional perception layer The development tasks are broken down into functional units, and each unit is implemented as follows: Environmental Sensing Unit: Integrating millimeter-wave radar, high-definition cameras, soil moisture sensors, and underground pipeline detection modules, this unit develops a data fusion and acquisition program to simultaneously collect data on terrain slope (sampling frequency 10Hz), road surface smoothness (error ≤0.5mm / m), meteorological data (temperature accuracy ±0.3℃), soil moisture (accuracy ±1%), and underground pipeline location (positioning error ≤5cm). Through image recognition algorithms (based on a YOLOv8-trained obstacle recognition model), the unit processes camera data to automatically classify obstacle types (such as rocks and temporary barriers), achieving an accuracy rate ≥98%.

[0047] Equipment sensing unit: An IoT terminal is installed on construction machinery (excavators, road rollers, etc.), integrating a Beidou-3 positioning module and a CAN bus data acquisition module. The terminal firmware program is developed to realize the data acquisition of equipment location (update frequency 1Hz), workload (error ≤3%), fuel / electricity remaining (accuracy ±1%), fault diagnosis codes, and operation efficiency (such as excavator bucket capacity utilization). The terminal supports IP67 protection level and is suitable for construction dust and vibration environments.

[0048] Material sensing unit: RFID tags are affixed to materials (steel bars, cement, sand, etc.) and the material type, specifications, quantity and storage location information are written in; ultra-wideband positioning base stations are deployed around transport vehicles and material piles, and positioning data acquisition programs are developed to obtain the material location in real time (positioning accuracy ±10cm). Combined with RFID readers (deployed at the entrance / exit of the material yard), automatic recording of material entry and exit is realized, forming a material flow traceability chain.

[0049] Personnel sensing unit: The smart safety helmet integrates a GPS / BeiDou positioning module, a heart rate sensor (measurement range 30-200 beats / min, accuracy ±1 beat / min) and a body temperature sensor (measurement range 32-42℃, accuracy ±0.1℃). The terminal program is developed to collect personnel location and physiological data in real time. When the heart rate is ≥120 beats / min or the body temperature is ≥38.5℃, a local audible and visual alarm is triggered and the alarm information is uploaded.

[0050] Unit testing: Each sensing unit is tested independently to verify the accuracy, stability, and environmental adaptability of data acquisition. For example, in a simulated construction scenario, equipment data is collected continuously for 24 hours to confirm that the data loss rate is ≤0.5%; the underground pipeline detection module is tested under different terrain conditions to ensure that the positioning error is ≤5cm.

[0051] 2. Development of the edge computing transport layer Edge computing nodes: Data processing programs are developed based on heterogeneous computing architectures. OpenCV is used to extract edge features from image / video data (preserving target contours and key parameters while discarding redundant pixels). Python Pandas library is used to clean structured data (removing outliers and filling in missing values), denoise (using Kalman filtering algorithm), and standardize the format. A data confidence screening threshold of ≥95% is set to ensure the validity of uploaded data.

[0052] Adaptive Transmission Module: A transmission protocol switching program is developed to monitor network quality in the construction area in real time (5G signal strength, Ethernet connection status, LoRaWAN signal quality). Based on preset rules, it dynamically selects the transmission protocol: 5G protocol (transmission rate ≥ 100Mbps) is used when 5G signal strength is ≥ -85dBm; industrial Ethernet (transmission latency ≤ 10ms) is switched when 5G signal strength is -100dBm ≤ -85dBm; and LoRaWAN protocol (communication distance ≥ 3km) is used when signal strength is < -100dBm. The LZ4 data compression algorithm is also integrated, reducing the amount of transmitted data by 40%-60%.

[0053] Unit testing: Simulate different network environments (strong signal, weak signal, no signal switching) to test the response time of protocol switching (≤100ms) and the stability of data transmission; verify data processing efficiency to ensure that there is no lag when 10 channels of high-definition video streams and 100 channels of sensor data are processed concurrently, and the data processing latency is ≤50ms.

[0054] 3. Development of the Intelligent Decision-Making and Scheduling Layer Data fusion module: Based on DS evidence theory, a multi-source data fusion program is developed to integrate the perception data uploaded from the edge layer with the baseline construction period, resource allocation standards, and cost budget data in the construction plan database. This constructs a three-dimensional decision matrix of "time-cost-efficiency" to solve the problems of data redundancy and conflict. The consistency of the fused data is ≥99%.

[0055] Dynamic optimization scheduling model: Based on the reinforcement learning DQN algorithm, the model program was developed using the TensorFlow framework. A construction scenario adaptation factor was introduced to improve the reward function, and the core reward indicators were set as the accident incidence rate (weight 0.4), schedule parameter (weight 0.3), cost parameter (weight 0.2), and efficiency parameter (weight 0.1). An iterative update program was developed to ensure a response time ≤500ms in the event of sudden changes in the construction scenario.

[0056] Historical data training module: Based on a historical database containing 500+ typical construction cases, a model pre-training program is developed. Transfer learning algorithms (such as the ResNet transfer learning framework) are used to transfer historical scheduling experience to new scenarios, improving the initial deployment effect of the model by more than 30%.

[0057] Solution Output Module: Develop instruction conversion and visualization programs to convert the scheduling strategies output by the model into structured instructions (equipment operation areas, material transportation routes, personnel shift arrangements, etc.), and generate scheduling solution reports (including resource configuration diagrams, project schedule curves, cost analysis tables, etc.) through the ECharts visualization component.

[0058] Unit testing: The model is tested using historical construction data to verify the optimization effect of the scheduling scheme (e.g., resource idle rate reduction ≥20%, cost savings ≥15%); simulates sudden scenarios such as equipment failure and material shortage to test the model's response time and the rationality of scheme adjustments; and verifies the generation efficiency and data accuracy of visualization reports.

[0059] 4. Development of the Execution Feedback Layer and Security Early Warning Subsystem Execution Feedback Layer: Develop instruction execution programs to interface with construction equipment control systems (such as excavator PLC controllers), material transportation management systems, and personnel management systems, converting scheduling instructions into equipment control signals (execution accuracy error ≤2%); develop status feedback programs to collect resource status data in real time after instruction execution, triggering model re-optimization when the actual deviation from the plan is ≥10%.

[0060] Safety Early Warning Subsystem: Based on the data from the perception layer, a risk assessment model is developed to identify equipment collision risks (calculating safe distances through location data), personnel safety hazards (exceeding physiological parameters), and environmental risks (extreme weather, pipeline conflicts). An early warning program is developed to issue early warning signals through audible and visual alarm devices and smart safety helmet terminals, and to trigger the model to generate avoidance plans.

[0061] Unit testing: test the accuracy of command execution (e.g., excavator operating range adjustment error ≤2%); simulate deviation scenarios (e.g., material transportation volume is 15% less than planned) to verify the triggering mechanism of model re-optimization; simulate risk scenarios such as personnel physiological abnormalities and close contact with equipment to test the early warning response time (≤500ms) and the effectiveness of risk avoidance schemes.

[0062] (III) System Integration and Joint Testing Module integration: The connection between the sensing layer and the edge computing layer is realized through the IoT gateway, the data interaction between the edge computing layer and the decision layer is realized through message queue (such as RabbitMQ), and the instruction transmission between the decision layer and the execution feedback layer is realized through the API interface. A complete closed-loop architecture of "sensing-computing-decision-execution-feedback" is built to ensure smooth data flow at each level without data loss or delay.

[0063] Full-process integration testing: In a simulated construction scenario (a test environment was set up including 10 construction devices, 5 types of materials, and 20 workers), full-process integration testing was conducted: the perception layer collected environmental, equipment, material, and personnel data, which were processed by the edge layer and uploaded to the decision layer. The decision layer generated a scheduling plan and issued it to the execution layer. After execution, the execution layer fed back status data to verify the integrity and stability of the closed-loop scheduling. The testing focused on the system's collaborative response capabilities under sudden scenarios (such as equipment failure or extreme rainstorms) to ensure the orderly linkage of all modules.

[0064] Performance optimization and adjustment: Optimize the issues found during joint debugging, such as improving the data processing efficiency of edge computing nodes, optimizing the weight of the model reward function, and enhancing the smoothness of transmission protocol switching, to ensure that the system's various performance indicators meet the preset standards (such as model response time ≤ 500ms, data transmission compression ratio ≥ 40%, and early warning accuracy ≥ 99%).

[0065] (iv) On-site trial operation and formal deployment Pilot Scenario Selection and Deployment: A typical highway construction scenario (such as the K10-K15 section of a highway reconstruction and expansion project) was selected as a pilot site. On-site hardware deployment (installation of sensing equipment, setup of edge computing nodes, and optimization of network coverage) and software system deployment (building of a cloud-based decision-making platform and installation of terminal programs) were completed. System operation training was conducted for construction personnel to clarify the procedures for receiving dispatch instructions, operating equipment, and responding to anomalies.

[0066] Trial Operation and Effect Monitoring: The trial operation period is set at 30 days. The system's operating status is monitored in real time, and key indicators are recorded, such as resource idle rate, project progress, cost consumption, and safety accident rate, and compared with traditional scheduling methods. For example, monitoring whether excavator idle time has decreased from 2 hours per day to less than 1 hour, and whether material transportation costs have decreased by more than 15%. Simultaneously, feedback from construction personnel is collected to optimize the system's user interface and scheduling logic.

[0067] Formal Deployment and Iterative Maintenance: After final optimization based on the trial operation results, the system will be formally deployed throughout the construction project. A system operation and maintenance mechanism will be established to regularly inspect hardware equipment (e.g., cleaning sensor lenses and checking positioning module signals) and update the software system (e.g., optimizing model parameters based on new construction scenarios and adding data statistics functions) to ensure long-term stable operation of the system.

[0068] II. Application Implementation (I) Overview of Application Scenarios Taking a provincial highway reconstruction and expansion project as an application case, this project, with a total length of 80km, covers three main construction sections: roadbed engineering, pavement engineering, and bridge engineering. The construction area involves both plains and hilly terrain, requiring 60 pieces of construction equipment such as excavators, road rollers, and pavers, more than 10 types of materials including steel bars, cement, and asphalt, and 200 workers. The project was originally planned for an 18-month construction period with a budgeted cost of 320 million yuan. Traditional scheduling methods suffered from high equipment idle rates (approximately 25%), delayed material supply, and unreasonable personnel division of labor, resulting in a one-month delay in the initial construction period and a 5% cost overrun. To solve these problems, this project introduced the IoT-based dynamic optimization scheduling system for construction resources described in this patent, achieving intelligent and dynamic resource scheduling.

[0069] (II) Implementation of system applications at all levels 1. Field applications of multi-dimensional perception layers Sensing devices are deployed across the construction area and resources to achieve full-scene data collection: An environmental sensing unit was deployed every 500 meters along the roadbed construction section. This unit integrated millimeter-wave radar, high-definition cameras, soil moisture sensors, and underground pipeline detection modules to collect real-time data on terrain slope (maximum 30° in hilly sections), road surface smoothness, real-time weather (such as rainfall and wind speed), soil moisture, and the location of underground pipelines. The underground pipeline detection module successfully located three unmarked communication fiber optic cables within the construction area, preventing the risk of them being severed during construction.

[0070] Equipment sensing terminals were installed on 60 construction machines to collect real-time data such as the operating load of excavators (e.g., boom digging force), the number of compaction cycles of road rollers, the paving speed of pavers, and the remaining fuel level. For example, if the load of an excavator drops to 60% after 4 hours of continuous operation, a fatigue operation warning will be automatically uploaded.

[0071] RFID tags are affixed to various materials, and ultra-wideband positioning base stations and RFID readers are deployed in material yards, transport vehicles, and construction sections to track the transportation status of materials in real time. For example, after steel bars leave the material yard, the system can display the location of the transport vehicle and the estimated time of arrival at the bridge construction section, solving the previous problem of delayed steel bar supply.

[0072] Two hundred workers were equipped with smart safety helmets, which collect real-time data on their location (differentiating between roadbed, pavement, and bridge construction areas), heart rate, and body temperature. In hot weather, when a worker's heart rate rose to 130 beats per minute, the system immediately triggered an alert, and management promptly arranged for the worker to rest, preventing heatstroke.

[0073] 2. Application safeguards for the edge computing transport layer Five edge computing nodes were deployed around the construction area, covering the entire construction section, to process the multi-source data collected by the sensing layer in real time. Edge features are extracted from construction scene videos captured by high-definition cameras, retaining only key information such as equipment location and personnel actions, reducing the data volume by 50%; structured data such as equipment parameters and environmental values ​​are cleaned to remove outliers caused by sensor jitter, retaining valid data with a confidence level of ≥95%.

[0074] The adaptive transmission module dynamically switches protocols based on network quality: in plain construction sections where the 5G signal is stable (strength ≥ -70dBm), the 5G protocol is used to achieve high-speed data transmission and support real-time video streaming; in hilly construction sections where the 5G signal is weak (strength -90dBm), it automatically switches to industrial Ethernet; in tunnel construction areas where the signal is extremely poor (strength < -100dBm), the LoRaWAN protocol is used to ensure stable transmission of critical data such as equipment fault codes and personnel locations.

[0075] 3. Core Applications of the Intelligent Decision-Making and Scheduling Layer Based on the processed real-time data and the baseline data of the construction plan, the cloud-based decision-making platform outputs an optimized scheduling solution: The data fusion module uses the DS evidence theory to fuse multi-source data and combines it with the original project plan (6 months for subgrade engineering, 8 months for pavement engineering, and 4 months for bridge engineering) to construct a decision matrix and clarify the priority of resource needs for each construction section.

[0076] After being pre-trained on historical data, the dynamic optimization scheduling model improved the initial deployment optimization effect by 35%. During the trial operation phase, the model found that the idle rate of excavators in the roadbed construction section was high (30%), while there were insufficient excavators in the bridge construction section. It immediately generated a scheduling plan: transferred two idle excavators from the roadbed section to the bridge section and adjusted their working range and digging speed; at the same time, it optimized the material transportation route, changing the cement transportation route from detouring through the national highway to using the nearest rural road, shortening the transportation time by 20 minutes. When a road roller suddenly malfunctioned, the model responded within 400ms, scheduling a nearby standby road roller to take over the operation, avoiding delays in the construction period.

[0077] The solution output module generates and issues structured instructions: equipment instructions specify the excavator's operating area and the road roller's compaction speed; material instructions specify the cement replenishment time (7:00 AM and 12:00 PM daily) and the transport vehicle number; personnel instructions specify the personnel shift arrangements for the bridge construction section (one shift every 6 hours). Simultaneously, a daily scheduling report is generated, visually displaying resource allocation and project progress.

[0078] 4. Application of the Execution Feedback Layer and Security Early Warning Subsystem The execution feedback layer translates scheduling instructions into equipment control signals and management commands. The excavator automatically adjusts its working range according to the commands (error ≤ 1.5%), and the transport vehicles travel along the planned route. The status feedback unit collects execution data in real time. If it finds that the amount of cement transported in a certain batch is 12% less than planned (deviation ≥ 10%), it immediately triggers the model to be re-optimized, adjusting the subsequent supply amount and the number of transport vehicles.

[0079] When the safety early warning subsystem detects that two excavators are too close (≤5m, posing a collision risk), it immediately issues an audible and visual alarm through the equipment terminal. At the same time, it triggers the model to generate a risk avoidance plan, instructing one of the excavators to stop working and retreat to a safe distance. When heavy rain is expected, it combines meteorological data to issue early warnings, dispatches material transport vehicles to nearby safe areas, and suspends roadbed earthwork operations to avoid the risk of slope collapse.

[0080] (III) Application Implementation Results Resource utilization efficiency has been significantly improved: the idle rate of construction equipment has decreased from 25% to 8%, and the operating efficiency of core equipment such as excavators and road rollers has increased by 30%; the material turnover rate has increased by 25%, and the backlog of materials such as cement and steel bars has decreased by 40%; the division of labor among personnel has become more reasonable, the labor intensity has decreased by 15%, and the satisfaction of construction personnel has improved.

[0081] Effective control of schedule and cost: Through dynamic optimization and scheduling, the one-month delay in the project was recovered, and the final total project duration was controlled at 17.5 months, which was 0.5 months shorter than the original plan; material transportation costs were reduced by 18%, equipment maintenance costs were reduced by 12%, the total budget cost was reduced by 6%, and expenditures were reduced by approximately 19.2 million yuan.

[0082] Construction safety has been greatly improved: the safety early warning subsystem triggered 32 warnings in total, including 8 warnings of abnormal physiological conditions of personnel, 15 warnings of equipment collision risks, and 9 warnings of environmental risks. All of these were handled in a timely manner to avoid safety accidents, and the project achieved the goal of "zero safety accidents" throughout the entire process.

[0083] Improved management intelligence: Through the visualization report of scheduling plan, managers can keep track of construction progress and resource status in real time, improving decision-making efficiency by 50%; the accumulation of historical data provides valuable experience for subsequent similar projects, and the model's initial deployment optimization effect was improved by 40% in another subsequent highway project.

[0084] (iv) Application and promotion value This system is not only applicable to highway reconstruction and expansion projects, but can also be extended to various highway engineering scenarios such as rural road construction, municipal road construction, and bridge and tunnel engineering. By replacing the sensing devices and adjusting the model parameters, it can meet the needs of construction projects of different terrains and scales. Its closed-loop architecture of "perception-computation-decision-execution-feedback" and dynamic optimization technology based on reinforcement learning provide an intelligent solution for highway engineering construction resource scheduling, effectively solving the problems of low efficiency, high cost, and poor security of traditional scheduling methods, and has broad application prospects and promotional value.

Claims

1. A dynamic optimization scheduling system for highway engineering construction resources based on the Internet of Things, characterized in that... ； The system includes a multi-dimensional perception layer, an edge computing transmission layer, an intelligent decision-making and scheduling layer, and an execution feedback layer. Each layer achieves data interaction and command transmission through IoT protocols, forming a closed-loop scheduling system of perception-computation-decision-execution-feedback. The multi-dimensional perception layer is deployed in the highway construction area and on various construction resources to collect construction environment parameters, resource status data and construction progress information; The edge computing transmission layer performs real-time preprocessing and local computation on the multi-source data collected by the perception layer, filters key data and uploads it to the intelligent decision scheduling layer through an adaptive transmission protocol, and at the same time receives scheduling instructions and sends them to the execution feedback layer. The intelligent decision-making and scheduling layer has a built-in dynamic optimization scheduling model based on reinforcement learning. It combines preprocessed real-time data with construction plan baseline data to output resource scheduling optimization schemes. The execution feedback layer is used to drive the execution of scheduling instructions for construction resources and collect resource execution status data in real time and feed it back to the multi-dimensional perception layer to realize dynamic verification and adjustment of scheduling effect.

2. The system according to claim 1, characterized in that, The multi-dimensional perception layer includes environmental perception units, equipment perception units, material perception units, and personnel perception units; The environmental perception unit uses a combination of millimeter-wave radar and high-definition camera to collect terrain slope, road surface flatness, real-time meteorological data and temporary obstacle information of the construction area. The millimeter-wave radar is used to acquire three-dimensional terrain data, and the high-definition camera, together with image recognition algorithm, can identify obstacle types. The equipment sensing unit collects real-time location, workload, fuel / electricity remaining, fault diagnosis codes, and work efficiency data of the equipment through an IoT terminal installed on the construction machinery. The IoT terminal integrates a Beidou-3 positioning module and a CAN bus data acquisition module to achieve full-dimensional monitoring of equipment status. The material sensing unit uses a combination of RFID tags and ultra-wideband positioning. The RFID tags have built-in information on material type, specifications, quantity and storage location, while the ultra-wideband positioning module collects real-time location data of material transport vehicles and on-site material piles, enabling full traceability of material flow. The personnel sensing unit collects real-time data on the worker's location, heart rate, body temperature, and labor intensity through the positioning module and physiological state monitoring sensors built into the smart safety helmet. When physiological parameters exceed safety thresholds, an early warning is automatically triggered.

3. The system according to claim 1, characterized in that, The edge computing transport layer includes edge computing nodes and an adaptive transport module; Edge computing nodes are deployed around the construction area and adopt a heterogeneous computing architecture. They extract edge features from unstructured data collected by the perception layer, clean, denoise and standardize the structured data, remove redundant data and retain valid data with a confidence level of ≥95%, thereby reducing data transmission bandwidth consumption. The adaptive transmission module has a built-in transmission protocol switching mechanism, and at the same time, it reduces the amount of transmitted data by 40%-60% through data compression algorithms.

4. The system according to claim 1, characterized in that, The intelligent decision-making and scheduling layer includes a data fusion module, a dynamic optimization scheduling model, and a scheme output module; The data fusion module uses DS evidence theory to fuse multi-source data uploaded from the edge computing transmission layer. Combined with the baseline construction period, resource allocation standards and cost budget data in the construction plan database, it constructs a scheduling decision matrix that includes three-dimensional objectives of "time-cost-efficiency". The dynamic optimization scheduling model is based on the DQN algorithm in reinforcement learning and introduces a construction scenario adaptation factor to improve the reward function: the reduction in resource idle rate, the duration of early / delayed construction, the amount of cost savings, and the incidence of safety accidents are used as core reward parameters, with the incidence of safety accidents having the highest weight (0.4), followed by the construction period parameter (0.3), the cost parameter having a weight of 0.2, and the efficiency parameter having a weight of 0.

1. The model updates the state space through real-time data iteration to achieve online self-learning and dynamic optimization of the scheduling strategy. When the construction scenario changes suddenly, the model response time is ≤500ms. The solution output module transforms the optimized scheduling strategy into structured instructions, including equipment scheduling instructions, material scheduling instructions, and personnel scheduling instructions, and simultaneously generates a visual report of the scheduling solution.

5. The system according to claim 1, characterized in that, The execution feedback layer includes an instruction execution unit and a status feedback unit; The instruction execution unit connects with the construction equipment control system, material transportation management system, and personnel management system through an Internet of Things terminal, converting scheduling instructions into equipment control signals (such as excavator operating range and transport vehicle speed) and management instructions, wherein the execution accuracy error of the equipment control signals is ≤2%; The status feedback unit collects resource status data in real time after the command is executed, including actual equipment operating parameters, actual material transportation volume and actual personnel working status. It feeds back to the intelligent decision-making and scheduling layer through the edge computing transmission layer to form a scheduling closed loop. When the actual execution status deviates from the scheduling plan by ≥10%, the model is triggered to re-optimize the process.

6. An implementation method for a dynamic optimization scheduling system for highway engineering construction resources based on the Internet of Things as described in claim 1, characterized in that, Includes the following steps: Step 1: Multi-dimensional perception data collection; Deploy perception devices in the highway construction area and construction resources to simultaneously collect construction environment parameters, resource status data and construction progress information to form multi-source raw data; Step 2: Edge data processing and transmission; The edge computing nodes around the construction area perform real-time preprocessing on the multi-source raw data, filter out valid data with a confidence level of ≥95%, and then the adaptive transmission module dynamically selects the transmission protocol according to the network quality and uploads it to the intelligent decision-making terminal, while receiving the scheduling instructions issued by the intelligent decision-making terminal. Step 3: Intelligent decision-making generates scheduling schemes; the intelligent decision-making terminal processes the uploaded data through data fusion technology, combines the construction plan baseline data to construct a three-dimensional decision matrix of "time-cost-efficiency", and uses a dynamic optimization scheduling model based on reinforcement learning to output a structured resource scheduling optimization scheme; Step 4: Execution feedback and closed-loop optimization; drive the construction resource execution scheduling plan, collect resource execution status data in real time and feed it back to the perception link to verify the scheduling effect. When the actual execution deviates from the plan by ≥10%, the model is triggered to re-optimize, forming a closed-loop scheduling system of perception-computation-decision-execution-feedback.

7. The method according to claim 6, characterized in that, The multi-dimensional sensing data acquisition mentioned in step 1 specifically includes: 1) Environmental perception: A combination of millimeter-wave radar and high-definition camera is used. The millimeter-wave radar acquires three-dimensional terrain data of the construction area to extract terrain slope and road surface smoothness parameters. The high-definition camera, in conjunction with image recognition algorithm, collects real-time meteorological data and temporary obstacle information and identifies obstacle types. 2) Equipment sensing: Through an IoT terminal that integrates a Beidou-3 positioning module and a CAN bus data acquisition module, the real-time location, workload, remaining fuel / electricity, fault diagnosis codes, and work efficiency data of the construction equipment are collected; 3) Material sensing: The system combines RFID tags with ultra-wideband positioning. The RFID tags store information on material type, specifications, quantity, and storage location, while the ultra-wideband positioning module collects real-time location data of material transport vehicles and on-site material piles. 4) Personnel perception: Through the positioning module and physiological sensors built into the smart safety helmet, real-time location, heart rate, body temperature and labor intensity data of the workers are collected. When the physiological parameters exceed the safety threshold, a local warning is automatically triggered. The environmental perception also includes: collecting soil moisture content data in the roadbed operation area through a soil moisture sensor, and locating the underground pipeline position through an electromagnetic induction underground pipeline detection module with a positioning error of ≤5cm.

8. The method according to claim 6, characterized in that, The edge data processing described in step 2 specifically includes: The edge computing nodes adopt a heterogeneous computing architecture to extract edge features from unstructured data (images, videos) and clean, denoise, and standardize the format of structured data (device parameters, environmental values) to remove redundant data; the effective data screening is based on a data confidence level of ≥95%. The adaptive transmission described in step 2 specifically includes: The adaptive transmission module has a built-in protocol switching mechanism. When the 5G signal strength is ≥-85dBm, it uses the 5G protocol for transmission; when the 5G signal strength is -100dBm≤5G signal strength<-85dBm, it switches to industrial Ethernet; when the network signal strength is <-100dBm, it uses the LoRaWAN protocol for transmission; and before transmission, the data volume is reduced by 40%-60% through a data compression algorithm.

9. The method according to claim 1, characterized in that, The data fusion technology mentioned in step 3 is the DS evidence theory. This theory integrates edge transmission data with baseline schedule, resource allocation standards, and cost budget data from the construction plan database to construct a three-dimensional decision matrix. The dynamic optimization scheduling model described in step 3 is based on the reinforcement learning DQN algorithm and introduces a construction scenario adaptation factor to improve the reward function. The core parameters of the reward function include the reduction in resource idle rate, the duration of early / delayed construction, the amount of cost savings, and the accident rate. The weights of each parameter are as follows: accident rate 0.4, construction period parameter 0.3, cost parameter 0.2, and efficiency parameter 0.

1. The model updates the state space iteratively through real-time data, and the response time is ≤500ms when there are sudden changes in the construction scenario. Before deployment, the dynamic optimization scheduling model described in step 3 needs to be pre-trained using a historical data training module based on a database containing 500+ typical highway construction cases. The transfer learning algorithm is then used to transfer historical scheduling experience to new construction scenarios, thereby improving the initial scheduling optimization effect of new scenarios by more than 30%. The structured resource scheduling optimization scheme described in step 3 includes: equipment scheduling instructions, material scheduling instructions, and personnel scheduling instructions, and simultaneously generates a visual scheduling report.

10. The method according to claim 6, characterized in that, The execution of scheduling instructions in step 4 specifically includes: converting scheduling instructions into equipment control signals and management instructions through an IoT terminal, with the execution accuracy error of the equipment control signals being ≤2%; the status feedback is achieved by collecting actual operating parameters of the equipment, actual material transportation volume, and actual working status of personnel, and the feedback data is uploaded to the intelligent decision-making terminal via the edge computing transmission layer.

Images