Internet of Things logistics information management platform system and management method

Abstract

The invention relates to the technical field of Internet of Things and logistics management, and discloses an Internet of Things logistics information management platform system and management method. Comprising cooperative work of an Internet of Things data acquisition terminal, a multi-dimensional heterogeneous data fusion module, a dynamic risk quantification engine, a self-adaptive steady-state regulation and control module, an intervention cost deduction module and an instruction issuing and man-machine interaction module, and realizes the following processes: fusing internal and external multi-source heterogeneous data in real time, and dynamically calculating a comprehensive risk score; generating an active intervention plan based on the risk score, and quantifying an expected risk reduction amount; the resource consumption, the secondary risk and the operation complexity of the plan are deduced in parallel, and a comprehensive intervention cost score is calculated; and finally, through comparing the expected income with the intervention cost, realizing automatic execution of a regulation and control instruction or pushing decision-making auxiliary information to a manager, and completing closed-loop regulation and control of man-machine cooperation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet of Things and logistics management, and particularly to an Internet of Things logistics information management platform system and a management method. BACKGROUND

[0002] With the rapid development of global trade and e-commerce, the modern logistics industry, especially the cold chain logistics and precision instrument transportation involving high-value and environmentally sensitive goods, has put forward unprecedentedly high requirements for the fine and intelligent management of the transportation process. In order to protect the quality and safety of goods during transportation, Internet of Things (IoT) technology has been widely used. By deploying various sensors (such as temperature, humidity, and vibration sensors) and global positioning systems (GPS) on vehicles or goods packaging, managers can remotely and real-time obtain basic state data in the logistics process.

[0003] The application of these technologies greatly enhances the "visibility" of the transportation process. When a key indicator (such as the temperature inside the refrigerated truck) exceeds the pre-set static threshold, the existing monitoring system can trigger an alarm to notify the relevant management personnel. This indeed improves the timeliness of problem discovery to some extent and constitutes the basis of current logistics risk management.

[0004] However, the existing technical practice still has significant limitations in deep-level risk assessment and decision support. Current risk judgment is often based on simple threshold comparison of single dimension or isolated data, which ignores the dynamic cumulative effect of risk and the complex coupling relationship of multiple factors (such as external weather, road conditions, driving behavior, and equipment working conditions). Therefore, the system cannot accurately identify potential high risks caused by multiple "sub-health" state indicators or produce unnecessary over-reactions to short-term and harmless data fluctuations.

[0005] Furthermore, when the system detects risks and issues an alarm, it usually cannot provide a quantitatively evaluated and cost-effective solution. The responsibility of decision-making falls entirely on the remote management personnel, who often lack sufficient information to weigh the pros and cons of different intervention measures. A seemingly direct intervention instruction may lead to a sharp rise in fuel costs, secondary damage to goods, or new operational complexity risks.

[0006] The existing technology generally lacks a "cost-benefit" analysis mechanism for intervention measures, i.e., it cannot quantify the risk reduction benefits brought by the execution of an operation and the comprehensive cost required for it. This blindness in decision-making leads to the fact that logistics management still remains in the stage of passive response and reliance on experience-based judgment, making it difficult to achieve optimal allocation of resources and active, efficient, and economic control of risks. SUMMARY

[0007] The application aims to provide an Internet of Things logistics information management platform system and a management method, and solves the problem of inaccurate risk identification caused by single risk assessment dimension and lack of foresight in the existing logistics monitoring system.

[0008] To achieve the above purpose, the application is implemented by the following technical solutions:

[0009] In a first aspect, the application provides an Internet of Things logistics information management platform system, comprising:

[0010] One or more Internet of Things data acquisition terminals configured to acquire internal state data and carrier telemetry data related to logistics tasks;

[0011] A multi-dimensional heterogeneous data fusion module in communication with the one or more Internet of Things data acquisition terminals and configured to fuse the internal state data, the carrier telemetry data, and external environment data to form a unified data stream;

[0012] A dynamic risk quantification engine connected to the multi-dimensional heterogeneous data fusion module and configured to calculate a comprehensive risk score in real time based on the unified data stream;

[0013] An adaptive steady-state control module connected to the dynamic risk quantification engine and configured to generate at least one proactive intervention plan aimed at reducing expected risk based on the comprehensive risk score;

[0014] An intervention cost deduction module connected to the adaptive steady-state control module and configured to receive the proactive intervention plan and calculate an intervention cost score for the proactive intervention plan; and

[0015] An instruction issuing and human-computer interaction module configured to determine and execute the final control instruction based on the comparison result of the expected risk reduction amount of the proactive intervention plan and the intervention cost score.

[0016] Preferably, in the multi-dimensional heterogeneous data fusion module, the internal state data includes temperature or humidity data of the cargo environment, the carrier telemetry data includes real-time position or refrigeration unit energy consumption rate data of the carrier, and the external environment data includes future weather forecast or real-time traffic incident data of the transportation path.

[0017] In a specific embodiment, the dynamic risk quantification engine is configured to calculate the comprehensive risk score by the following formula

[0018] ;

[0019] wherein, is a comprehensive risk score; is an internal state risk function based on an internal state data vector and its time rate of change is calculated; is an external environment risk function based on an external environment data vector is calculated; is a vehicle operating condition risk function based on a vehicle telemetry data vector is calculated; , , are weight coefficients corresponding to the internal state risk function, the external environment risk function and the vehicle operating condition risk function, respectively.

[0020] Preferably, the adaptive steady state control module is configured to generate the active intervention plan when the comprehensive risk score exceeds a preset alert threshold, or a rate of change of the comprehensive risk score exceeds a preset slope threshold.

[0021] In one specific embodiment, the intervention cost deduction module is configured to calculate the intervention cost score by the following formula

[0022] ;

[0023] wherein, is an intervention cost score for an active intervention plan ; is a resource consumption cost function for evaluating additional resource consumption required for executing the active intervention plan ; is a secondary risk cost function for evaluating an increment of secondary failure probability possibly caused by executing the active intervention plan ; , are cost weight coefficients corresponding to the resource consumption cost function and the secondary risk cost function, respectively.

[0024] In one embodiment, the additional resource consumption evaluated by the resource consumption cost function includes an additional fuel consumption or an additional electric power consumption generated by executing the active intervention plan.

[0025] In one embodiment, the increment of secondary failure probability evaluated by the secondary risk cost function An increased failure probability value caused by an overload operation of a refrigeration unit of the vehicle due to the execution of the active intervention plan.

[0026] Further, the instruction issuing and human-computer interaction module is specifically configured to: when the expected risk reduction amount is greater than the intervention cost score, execute the regulation instruction corresponding to the active intervention plan.

[0027] Correspondingly, the instruction issuing and human-computer interaction module is further configured to: when the expected risk reduction amount is not greater than the intervention cost score, veto the active intervention plan, and send decision assistance information including the comprehensive risk score and the intervention cost score to the artificial administrator.

[0028] In a second aspect, the present application provides an Internet of Things logistics information management method, which is executed by the system of any of the above-mentioned embodiments, and includes the following steps:

[0029] S1. Collect and fuse internal state data, vehicle telemetry data and external environment data related to the logistics task to form a unified data stream;

[0030] S2. Based on the unified data stream formed in step S1, a comprehensive risk score is calculated in real time;

[0031] S3. According to the comprehensive risk score calculated in step S2, at least one active intervention plan aimed at reducing the expected risk is generated;

[0032] S4. For the active intervention plan generated in step S3, an intervention cost score is calculated, which is used to represent the resource consumption and possible secondary risks caused by the execution of the active intervention plan;

[0033] S5. Based on the comparison result of the expected risk reduction amount of the active intervention plan in step S3 and the intervention cost score calculated in step S4, a final regulation instruction is determined;

[0034] S6. The final regulation instruction determined in step S5 is executed.

[0035] In summary, the present application includes at least one of the following beneficial technical effects:

[0036] 1.The present application sets up a multi-dimensional heterogeneous data fusion module and a dynamic risk quantification engine, which not only collects the real-time internal state of goods and vehicles, but also actively fuses external environmental data such as future weather forecasts and traffic events on the transportation path. This enables the system to quantify potential risks in advance based on prediction information, generate response plans before abnormal events actually occur, and change the traditional "after-the-fact remediation" mode to a "before-the-fact prevention" active management mode, achieving predictive management of risks and improving active defense capabilities.

[0037] 2.The present application sets up an intervention cost deduction module, which quantifies the additional resource consumption and secondary risks that may be caused by any active intervention plan before the plan is executed, forming an intervention cost score. By comparing the expected risk reduction with the cost score, the system can avoid executing "unprofitable" control instructions, ensuring that each automated intervention is carried out under the premise of optimal overall cost-benefit, thereby improving the economy and reliability of overall operations, ensuring the economy and rationality of intervention decisions, and avoiding negative effects caused by inappropriate interventions.

[0038] 3.The system architecture of the present application constitutes a complete closed loop of "perception-analysis-decision-review-execution". The continuously changing comprehensive risk score output by the dynamic risk quantification engine enables the adaptive steady-state control module to perform hierarchical strategy adjustment that is precisely matched with the risk level, rather than simple binary threshold triggering. This refined control capability enables the system to automatically reduce monitoring consumption when the risk is low and concentrate resources to respond when the risk increases, achieving dynamic and efficient allocation of monitoring and intervention resources, achieving refined and adaptive closed-loop control, and optimizing resource allocation. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 The structural block diagram of an Internet of Things logistics information management platform system according to an embodiment of the present application;

[0040] Figure 2 The internal structural block diagram of an Internet of Things data acquisition terminal according to an embodiment of the present application;

[0041] Figure 3 The functional block diagram of a multi-dimensional heterogeneous data fusion module according to an embodiment of the present application;

[0042] Figure 4 The internal logic flowchart of a dynamic risk quantification engine according to an embodiment of the present application;

[0043] Figure 5 The internal logic block diagram of an adaptive steady-state control module according to an embodiment of the present application;

[0044] Figure 6Internal computing flow chart of intervention cost deduction module of one embodiment of the present application;

[0045] Figure 7 Internal function block diagram of instruction issuing and human-computer interaction module of one embodiment of the present application;

[0046] Figure 8 Flow chart of Internet of Things logistics information management method of one embodiment of the present application.

[0047] Among them, 10, Internet of Things data acquisition terminal; 11, microcontroller unit; 12, sensor array; 13, communication module; 14, power management unit; 20, multi-dimensional heterogeneous data fusion module; 21, internal data receiving unit; 22, external data acquisition unit; 23, data preprocessing pipeline; 30, dynamic risk quantification engine; 40, adaptive steady-state regulation module; 50, intervention cost deduction module; 60, instruction issuing and human-computer interaction module; 61, decision logic unit; 62, automatic instruction interface; 63, decision assistance presentation unit. DETAILED DESCRIPTION

[0048] The following will be described in detail in combination with the accompanying Figure 1 -Appendix Figure 8 The present application will be further described in detail.

[0049] The present application provides an Internet of Things logistics information management platform system.

[0050] Referring to the accompanying Figure 1 The present application provides an Internet of Things logistics information management platform system, which can include: an Internet of Things data acquisition terminal 10, a multi-dimensional heterogeneous data fusion module 20, a dynamic risk quantification engine 30, an adaptive steady-state regulation module 40, an intervention cost deduction module 50, and an instruction issuing and human-computer interaction module 60.

[0051] The Internet of Things data acquisition terminal 10 is configured to acquire internal state data and carrier telemetry data related to logistics tasks, the internal state data including but not limited to temperature, humidity, vibration, and gas composition data of the cargo environment; the carrier telemetry data including but not limited to real-time geographic position, speed, refrigeration unit working condition, and fuel or power consumption rate data of the carrier. The Internet of Things data acquisition terminal 10 sends the acquired data to the multi-dimensional heterogeneous data fusion module 20.

[0052] The multi-dimensional heterogeneous data fusion module 20 communicates with the IoT data collection terminal 10. The multi-dimensional heterogeneous data fusion module 20 receives data from the IoT data collection terminal 10, and also acquires external environment data, such as weather forecast data and real-time traffic event data, through a network interface. The multi-dimensional heterogeneous data fusion module 20 fuses the received and acquired internal state data, vehicle telemetry data, and external environment data, forms a unified data stream, and provides the unified data stream to the dynamic risk quantification engine 30.

[0053] The dynamic risk quantification engine 30 is connected to the multi-dimensional heterogeneous data fusion module 20. The dynamic risk quantification engine 30 calculates a comprehensive risk score in real time based on the unified data stream. In a specific embodiment, the comprehensive risk score is calculated by the following formula:

[0054]

[0055] Wherein:

[0056] is the comprehensive risk score at time point is the internal state data vector is the time rate of change of the internal state data vector is the external environment data vector is the vehicle telemetry data vector is the internal state risk function calculated based on and is the external environment risk function calculated based on is the vehicle operating condition risk function calculated based on is the preset weight coefficient corresponding to each of the three risk functions. The calculated comprehensive risk score is sent to the adaptive steady-state control module 40. The adaptive steady-state control module 40 is connected to the dynamic risk quantification engine 30. The adaptive steady-state control module 40 generates at least one active intervention plan aimed at reducing the expected risk based on the received comprehensive risk score. For example, when the comprehensive risk score exceeds a preset alert threshold, the adaptive steady-state control module 40 generates one or more active intervention plans. The active intervention plan is sent to the intervention cost deduction module 50 and the instruction issuing and human-machine interaction module 60.

[0057] The adaptive steady-state control module 40 is connected to the dynamic risk quantification engine 30. The adaptive steady-state control module 40 generates at least one active intervention plan aimed at reducing the expected risk based on the received comprehensive risk score. For example, when the comprehensive risk score exceeds a preset alert threshold, the adaptive steady-state control module 40 generates one or more active intervention plans. The active intervention plan is sent to the intervention cost deduction module 50 and the instruction issuing and human-machine interaction module 60.

[0058] The adaptive steady-state control module 40 is connected to the dynamic risk quantification engine 30. The adaptive steady-state control module 40 generates at least one active intervention plan aimed at reducing the expected risk based on the received comprehensive risk score. For example, when the comprehensive risk score exceeds a preset alert threshold, the adaptive steady-state control module 40 generates one or more active intervention plans. The active intervention plan is sent to the intervention cost deduction module 50 and the instruction issuing and human-machine interaction module 60.

[0059] ​​​The intervention cost estimation module 50 is connected to the adaptive steady-state control module 40. This intervention cost estimation module 50 is configured to receive an active intervention plan and calculate an intervention cost score for that plan. In one specific embodiment, this intervention cost score... The calculation is performed using the following formula:

[0060] ;

[0061] in:

[0062] : For proactive intervention plans Intervention cost score; To implement the contingency plan Additional resource consumption required; To implement the contingency plan The potential increase in the probability of secondary failures; To implement the contingency plan Additional operating load required; : is the resource consumption cost function, which is based on Perform calculations; : is the secondary risk cost function, which is based on Perform calculations; : is the cost function for operational complexity, which is based on Perform calculations; : These are the preset cost weight coefficients corresponding to the three cost functions.

[0063] The calculated intervention cost score is sent to the instruction delivery and human-computer interaction module 60.

[0064] The instruction issuance and human-machine interaction module 60 is connected to the adaptive steady-state control module 40 and the intervention cost inference module 50. This module 60 is configured to receive the expected risk reduction amount of the proactive intervention plan and the intervention cost score. Based on the comparison of these two values, the module 60 determines and executes the final control instruction. For example, if the expected risk reduction amount is greater than the intervention cost score, the module 60 executes the proactive intervention plan; otherwise, it rejects the plan.

[0065] In one embodiment, System 1 of the present invention can be deployed in a completely centralized manner. Specifically, the multi-dimensional heterogeneous data fusion module 20, the dynamic risk quantification engine 30, the adaptive steady-state control module 40, the intervention cost deduction module 50, and the instruction issuance and human-computer interaction module 60 are all deployed on a remote cloud server cluster. The Internet of Things data acquisition terminal 10 communicates with the cloud server via the public network.

[0066] In another embodiment, the system 1 of the present application can be deployed in an edge-cloud collaborative manner. Specifically, to reduce communication latency and support offline decision making, the multi-dimensional heterogeneous data fusion module 20 and the dynamic risk quantification engine 30 are deployed on an edge computing gateway inside or near the vehicle. The edge computing gateway directly processes data with high real-time requirements and performs preliminary risk assessment. The adaptive steady-state regulation module 40, the intervention cost deduction module 50, and the instruction issuance and human-computer interaction module 60, which require more powerful computing support and centralized management and human interaction, are deployed on the cloud server. The edge computing gateway periodically synchronizes the risk score and necessary data to the cloud.

[0067] Referring to the drawings Figure 2 In one specific embodiment, the IoT data acquisition terminal 10 includes a microcontroller unit 11, a sensor array 12, a communication module 13, and a power management unit 14.

[0068] The microcontroller unit 11, as the control core of the IoT data acquisition terminal 10, is responsible for executing the preset data acquisition program, processing sensor data, constructing data packets, and communicating with the communication module 13.

[0069] The sensor array 12, connected to the microcontroller unit 11, is used to obtain specific internal state data and vehicle telemetry data. In one embodiment, the sensor array 12 includes one or more semiconductor temperature sensors or platinum resistance temperature sensors (RTD) for measuring the temperature of the cargo storage space with a measurement accuracy of ±0.2 degrees Celsius; a capacitive humidity sensor for measuring the relative humidity of the cargo storage space; a three-axis accelerometer based on micro-electro-mechanical systems (MEMS) for monitoring vibration and impact events during transportation with a range of ±8g; an electrochemical ethylene gas sensor for monitoring the concentration of ethylene released during the ripening process of fruit and vegetable cargo; a global positioning system (GPS) module for obtaining real-time latitude, longitude, altitude, and speed information of the vehicle.

[0070] In addition, the IoT data acquisition terminal 10 is directly connected to the electronic control unit (ECU) of the vehicle through a controller area network (CAN) bus interface or an on-board diagnostic (OBD-II) interface. Through this connection, the microcontroller unit 11 can directly read vehicle telemetry data such as engine speed, actual power of the refrigeration unit, instantaneous fuel consumption rate, and remaining fuel / electricity percentage.

[0071] Regarding the execution parameters of data acquisition, the microcontroller unit 11 is configured to execute a configurable acquisition strategy. The default data acquisition cycle of this strategy is once every 60 seconds. However, this acquisition cycle can be dynamically adjusted by the upper-level system (e.g., the command issuing and human-machine interaction module 60) through issuing commands, with an adjustable range from 5 seconds to 600 seconds. Simultaneously, the microcontroller unit 11 is also configured with event-triggered acquisition logic: when the impact value detected by the triaxial accelerometer exceeds a preset threshold (e.g., 4g), the microcontroller unit 11 immediately performs a high-frequency data acquisition, that is, acquiring data from all sensors once per second for the following 10 seconds, and prioritizing the transmission of this group of data containing the event context.

[0072] Regarding communication with the upper-layer module, the communication module 13, under the control of the microcontroller unit 11, is responsible for establishing a remote data link with the multidimensional heterogeneous data fusion module 20. In one embodiment, the communication module 13 is a cellular communication module supporting 4G LTE networks. Data transmission uses the Message Queuing Telemetry Transport (MQTT) protocol. The microcontroller unit 11 packages the collected and processed data into a JSON-formatted data packet and publishes it via the MQTT protocol. This communication link is bidirectional; in addition to sending data, the communication module 13 is also responsible for receiving control commands from the multidimensional heterogeneous data fusion module 20 or from the instruction setter module 60, such as commands for adjusting the data acquisition cycle.

[0073] See attached document Figure 3 In one specific implementation, the multidimensional heterogeneous data fusion module 20 is deployed on a cloud server or edge computing node, and includes an internal data receiving unit 21, an external data acquisition unit 22, and a data preprocessing pipeline 23.

[0074] The internal data receiving unit 21 is responsible for processing data streams from one or more IoT data acquisition terminals 10. In one embodiment, the internal data receiving unit 21 has a built-in MQTT broker server for subscribing to data published by the IoT data acquisition terminals 10. Upon receiving a JSON-formatted data packet, the internal data receiving unit 21 parses it and extracts various data fields, such as sensor type, value, unit, and the acquisition timestamp recorded by the terminal.

[0075] External data acquisition unit 22 is responsible for proactively acquiring external environmental data related to the transportation task from third-party services. This external data acquisition unit 22 is configured to periodically (e.g., every 15 minutes) send requests to the application programming interface (API) of external service providers based on the latest vehicle geographic location information obtained from the internal data receiving unit 21. For example, it may send a request to a weather service API carrying key geographic coordinates along the future path to obtain hourly temperature, precipitation probability, and wind speed forecasts for the next 24 hours; and send a request to a map service API to obtain real-time traffic congestion indices and published traffic accident or road closure events along the transportation route. The acquired data is in XML or JSON format.

[0076] The data preprocessing pipeline 23 receives data from the internal data receiving unit 21 and the external data acquisition unit 22, and performs a series of processing steps to form a unified data stream. This pipeline includes the following serial operations:

[0077] First, the data format is standardized. All data from all sources, whether internal sensor readings or external API returns, are converted into a standardized internal data structure. This structure contains the following fields: source_id (data source identifier), timestamp_utc (uniform timestamp), data_type (data type, such as temperature), value (numerical value), and unit (unit, such as celsius).

[0078] Second, data cleaning. After standardizing the data format, validity processing is performed. For sensor data, a sliding window (e.g., containing the 10 most recent data points) is used for outlier detection. If a data point deviates from the standard deviation of the mean within the window by more than three times, it is marked as invalid and discarded. For sporadic missing data points due to network issues, linear interpolation is used to fill in the gaps using the two nearest valid data points before and after the missing data point, but the interpolation time span cannot exceed 120 seconds.

[0079] Third, timestamp alignment. To address the delays and discrepancies in timestamps from different data sources (sensors, vehicle CAN bus, external APIs), this step aligns all data to a common timeline. The system generates a series of time bins at fixed time intervals (e.g., 10 seconds). Each valid data entry is assigned to the nearest time bin based on its timestamp_utc field. This process ensures that at any given point in time, the system obtains a snapshot of the status of all relevant data sources at similar times.

[0080] After completing the above steps, the data preprocessing pipeline 23 combines the data from each time bay into a structured fused data frame. This fused data frame is a unit of the unified data stream, containing aligned internal state data vectors, vehicle telemetry data vectors, and external environment data vectors, and is immediately transmitted to the dynamic risk quantification engine 30 for the next step of risk calculation.

[0081] See attached document Figure 4 The dynamic risk quantification engine 30 receives fused data frames from the multi-dimensional heterogeneous data fusion module 20, performs risk calculations based on the data frames, and finally outputs a normalized comprehensive risk score within the range of [insert range here]. .

[0082] The core calculation model of the Dynamic Risk Quantification Engine 30 is:

[0083] ;

[0084] in, These are the preset weighting coefficients, and their sum is 1. The following provides a detailed explanation of each function in the formula.

[0085] Regarding the internal state risk function Implementation:

[0086] This function quantifies the risks arising from sensor data directly related to the cargo. It is calculated by weighted summation of the risks of each internal state sub-risk.

[0087] In one embodiment, for temperature-sensitive goods, the transport contract specifies a temperature range of [temperature range to be specified]. Then the temperature risk The calculation method is as follows:

[0088] ;

[0089] in, The absolute deviation risk is calculated as follows:

[0090] like ,but ;

[0091] like ,but ;

[0092] like ,but ;

[0093] Here, The current temperature. This is a preset critical temperature difference (e.g., 5 degrees Celsius).

[0094] The risk of rate of change is calculated as follows:

[0095] ;

[0096] Here, For the rate of temperature change, This is a preset critical temperature change rate (e.g., 0.5 degrees Celsius / minute). and These are the corresponding weights. Similarly, sub-risks for other internal states such as humidity and vibration can be calculated. Ultimately, It is the weighted sum of all sub-risks.

[0097] Regarding the external environment risk function Implementation:

[0098] This function is used to quantify the risks caused by future external environmental factors along the transportation route.

[0099] In one embodiment, external data vector Includes the predicted temperature of the geographical area the vehicle will pass through within the next 2 hours. If the specified upper temperature limit for the goods is Then external temperature risk The calculation method is as follows:

[0100] like ,but ;

[0101] like ,but ,in It is a proportionality coefficient.

[0102] Similarly, other external sub-risks can be calculated based on predicted precipitation probabilities, road congestion indices, and other factors. Ultimately, This is the weighted sum of all external sub-risks.

[0103] Regarding the risk function of vehicle operating conditions Implementation:

[0104] This function is used to quantify the risks caused by the vehicle's own state.

[0105] In one embodiment, vehicle telemetry data vector Includes the vehicle's remaining fuel percentage Set a fuel warning threshold. (For example, 20%). Then the risk of fuel e-fuel. The calculation method is as follows:

[0106] like ,but ;

[0107] like ,but .

[0108] Another example is the operating condition of the cooling unit. Based on the current ambient temperature and the set target temperature, the system has a pre-defined baseline value for the cooling unit's energy consumption. If the actual energy consumption read from the CAN bus is... Then the sub-risk of abnormal operating conditions The calculation method is as follows:

[0109] ,in It is an acceptable tolerance for deviation (e.g., 30%).

[0110] final, This is a weighted sum of the risks for all vehicle operating conditions.

[0111] Weighting coefficient It is configured based on the specific attributes of the transportation task and stored in a configuration table. This configuration table is indexed by the corresponding weight set according to the cargo type and service level agreement (SLA).

[0112] For example, when transporting high-value, highly sensitive biological agents, the internal cargo condition is crucial, and the weight can be set as follows: .

[0113] For transporting routine, environmentally insensitive industrial components, the primary concern is the reliability of the vehicle itself; therefore, the weighting can be set as follows: .

[0114] Before the transportation task begins, the system automatically loads the corresponding set of weight coefficients from the configuration table based on the task order information for subsequent risk calculation.

[0115] In a further embodiment, the weighting coefficients are... and cost weighting coefficient ( This is not solely achieved through manual pre-setting. The invention may also include an offline model training module. This module is responsible for collecting and storing a large amount of complete data from historical transportation tasks, including full-cycle sensor data, external environmental data, executed intervention instructions, and final transportation results (e.g., whether the goods are intact, whether they are delivered on time, total cost, etc.).

[0116] The model training module utilizes historical data and employs machine learning algorithms (e.g., logistic regression or gradient boosting decision trees) for training. Using transportation outcomes as labels and individual risks or costs as features, it automatically optimizes and generates a set of weights that maximizes risk prediction accuracy and cost assessment under different cargo types and transportation conditions. When starting a new task, the system can load these optimal weights generated by machine learning, enabling adaptive parameter configuration.

[0117] See attached document Figure 5 The adaptive steady-state control module 40 is connected to the dynamic risk quantification engine 30, and receives the comprehensive risk score output by the latter. and its rate of change over time The adaptive steady-state control module 40 internally includes a risk interval determination unit 41, a contingency plan generation rule base 42, and a contingency plan output interface 43.

[0118] The risk interval determination unit 41 is configured to map the current risk state to a risk interval based on a preset threshold. In one embodiment, two risk thresholds are defined: a safety threshold and a security threshold. (e.g., a normalized score of 0.4) and danger threshold (For example, a normalized score of 0.7). The risk range is divided into three:

[0119] Safe zone: when ;

[0120] Warning zone: When ;

[0121] Danger zone: When .

[0122] In addition, this unit 41 also assesses the rate of change of risk scores over time. And define a slope threshold. (For example, 0.05 / minute). When Even if the current risk score is in the safe zone, the system status is still judged to require attention.

[0123] The contingency plan generation rule base 42 stores a series of "condition-action" rule pairs. The judgment result of the risk interval determination unit 41 serves as an input condition, used to search for and trigger the corresponding action in this rule base, that is, to generate one or more contingency plans. Contingency plans are divided into two types: monitoring strategy contingency plans and proactive intervention contingency plans.

[0124] When the risk zone determination unit 41 determines the risk zone as "safe zone" and the risk score change rate is stable, the contingency plan generation rule base 42 generates a monitoring strategy contingency plan to maintain baseline operation. For example, this contingency plan instructs the IoT data acquisition terminal 10 to use a regular data acquisition cycle of 60 seconds.

[0125] When the judgment result is "warning zone", or the judgment result is "safe zone" but the risk score change rate At that time, the contingency plan generation rule base 42 is triggered to generate a contingency plan of an early warning nature.

[0126] First, an enhanced monitoring strategy plan is generated, which instructs the IoT data acquisition terminal 10 to shorten the data acquisition cycle to 15 seconds in order to obtain higher density status data.

[0127] Secondly, based on the primary sources of risk, one or more preventative proactive intervention plans are generated. For example, if the main reason for the increase in risk score is the external environmental risk function in the dynamic risk quantification engine 30... If the output value increases (e.g., if high temperature is predicted on the path ahead), the generated active intervention plan is to "lower the target temperature setpoint of the vehicle cooling unit by 1.5 degrees Celsius for pre-cooling".

[0128] When the determination result is "danger zone", the contingency plan generation rule base 42 is triggered to generate an emergency response plan.

[0129] First, a high-priority monitoring strategy plan is generated, which instructs the IoT data acquisition terminal 10 to shorten the data acquisition cycle to the minimum value supported by its hardware (e.g., 5 seconds).

[0130] Secondly, generate one or more corrective proactive intervention plans. For example, if the main reason for the risk score exceeding the limit is the internal state risk function... The output value exceeds the standard (e.g., the cargo temperature has exceeded the limit). If the risk primarily originates from the vehicle's operating condition risk function, then the generated proactive intervention plan is to "set the operating mode of the on-board cooling unit to 'forced maximum power'". If the output value exceeds the standard (such as detecting abnormal energy efficiency of the main cooling unit), the generated contingency plan is to "immediately start the backup cooling unit and send the fault code of the main cooling unit to the human-machine interaction module".

[0131] The contingency plan output interface 43 packages the generated contingency plans into structured data objects and sends them out. Each contingency plan object contains the following fields: plan_id (unique identifier for the contingency plan), plan_type (monitoring strategy / active intervention), target_module (target execution module, such as IoT data acquisition terminal 10 or vehicle ECU), action_details (action parameters, such as {"cycle_time": 15} or {"mode": "max_power"}), and an expected_risk_reduction (estimated value of expected risk reduction). This estimate is derived based on a preset model in the rule base; for example, executing the "forced maximum power" contingency plan is expected to reduce internal state risk by 50% within 10 minutes. The contingency plan object is simultaneously sent to the intervention cost deduction module 50 and the instruction issuance and human-machine interaction module 60.

[0132] See attached document Figure 6 The intervention cost estimation module 50, connected to the adaptive steady-state control module 40, is configured to receive each active intervention plan generated by the latter. The core task of this module is to quantify the inherent, multi-dimensional costs of each received plan before its actual execution, thereby outputting a normalized intervention cost score. .

[0133] In one specific implementation, the intervention cost score The calculation follows the formula:

[0134] ;

[0135] in, These are the cost weighting coefficients, and their sum is 1. The following provides a detailed explanation of the implementation of each cost function in the formula.

[0136] Regarding the resource consumption cost function Implementation:

[0137] This function is used to quantify the execution plan. Additional tangible resources required beyond the baseline operational level.

[0138] Taking a "forced pre-cooling" plan as an example, this plan instructs the cooling unit to operate at 80% power for 30 minutes. First, the intervention cost simulation module 50 obtains the baseline energy consumption from the vehicle's operating condition model. (That is, the normal power required to maintain the current temperature under the same external environment), and then calculate the additional power. Additional resource consumption That is, the product of additional power and time. Hours. If it is a fuel-fired refrigeration unit, then it will be based on its fuel efficiency (liters / kWh). Converted to additional fuel consumption (liters). The function normalizes this physical consumption by dividing it by a preset "maximum allowable additional fuel consumption per task" (e.g., 5 liters) to obtain a score between 0 and 1.

[0139] Regarding the cost function of secondary risk Implementation:

[0140] This function is used to quantify the execution plan. This results in additional wear and tear on system components or an increased probability of failure.

[0141] Continuing with the example of the aforementioned "forced pre-cooling" plan, the intervention cost estimation module 50 internally stores a component lifespan degradation model based on the component's operating history and performance curves provided by the manufacturer. This model indicates that for every hour the refrigeration compressor operates continuously at 80% power, its mean time between failures (MTBF) decreases by an additional 1.5 hours. Therefore, the secondary risks resulting from implementing this plan... This is reflected in a 0.75-hour reduction in MTBF.

[0142] The function normalizes this loss value. For example, dividing the reduced MTBF value (0.75 hours) by the total expected MTBF of the component (e.g., 10,000 hours) yields a very small, but non-zero, cost score. For more complex scenarios, such as "increasing tire pressure on bumpy roads," additional tire wear costs are calculated based on a tire wear model.

[0143] Regarding the cost function of operation complexity Implementation:

[0144] This function is used to quantify the contingency plan. Interference and load on the driver or other manual operations. This function is typically implemented using a pre-defined rule lookup table. The intervention cost estimation module 50 analyzes the type of contingency plan and the required level of human interaction, assigning it a fixed cost score. For example:

[0145] For pre-planned actions that are executed automatically by the system in the background (such as adjusting the power of the chiller). To be "interference-free" The output is 0.

[0146] For pre-planned scenarios that require the driver to confirm via a click on the in-vehicle screen (such as "Do you agree to the optimized route?"), For "low interference", The output is 0.2.

[0147] For contingency plans that require drivers to perform complex operations or significantly change their driving behavior (such as "immediately exit the highway and proceed to the designated repair shop")... "High interference" The output is 0.9.

[0148] Regarding the cost weighting coefficient The basis for the setting:

[0149] The setting of these weighting coefficients is closely related to business objectives and operational strategies, reflecting the tolerance for different types of costs in different scenarios.

[0150] In one embodiment, for general cargo that is highly sensitive to transportation costs, the operator pays more attention to direct costs such as fuel, and therefore the weight can be set as follows: .

[0151] For transporting high-value, irreplaceable goods with extremely high time sensitivity (such as live organs), ensuring the absolute safety of equipment is the primary task. In this case, secondary risks have the highest weight and can be set as follows: .

[0152] For long-haul transportation tasks with strict delivery time windows and drivers already at full workload, it is even more important to avoid causing additional interference to the drivers. The following settings can be configured: .

[0153] After the calculation is completed, the final intervention cost score is... It is transmitted to the instruction issuing and human-computer interaction module 60 for final decision comparison.

[0154] See attached document Figure 7 The instruction issuance and human-computer interaction module 60 is the final execution and supervision link of the system decision, which includes a decision logic unit 61, an automatic instruction interface 62, and a decision assistance presentation unit 63.

[0155] The decision logic unit 61 is connected to the adaptive steady-state control module 40 and the intervention cost estimation module 50. It receives the proactive intervention plan from the adaptive steady-state control module 40. (This contingency plan includes a quantified expected risk reduction amount) (and simultaneously receive data from the intervention cost projection module 50 for the same contingency plan) Calculated intervention cost score The core function of this decision logic unit 61 is to perform a direct numerical comparison:

[0156] like If the plan is deemed to have positive benefits, the execution instructions for the plan will be passed to the automatic instruction interface 62.

[0157] like If the cost of implementing the plan exceeds its risk avoidance benefits, the plan and related analysis data will be transmitted to the decision support presentation unit 63.

[0158] The automatic command interface 62 is responsible for converting the decided plan into specific electronic commands that can be executed by the on-board equipment. In one embodiment, when an execution command is received, the automatic command interface 62 generates a JSON command data packet conforming to a predetermined format. For example, for a "forced maximum power" plan, the generated command packet is: {"command_id": "CMD001", "target_device": "refrigeration_ecu_main", "action": "set_mode", "parameter": "max_power"}. This command packet is sent to the IoT data acquisition terminal 10 on the target vehicle via the system's MQTT communication link. After receiving and verifying the command packet, the microcontroller unit 11 of the IoT data acquisition terminal 10 translates it into one or more CAN messages conforming to the Controller Area Network (CAN) bus protocol, and sends them to the designated cooling unit electronic control unit (ECU) via the CAN bus interface, thereby completing the direct control of the device status.

[0159] The decision support presentation unit 63 is responsible for generating and pushing data to the human-computer interaction interface. When the decision logic unit 61 determines that a plan should not be executed automatically, the decision support presentation unit 63 is responsible for pushing the complete decision context information to the user terminal of the back-end administrator (e.g., a web management platform or a mobile app).

[0160] On the web management platform, this data object is parsed and visualized by the front-end program: the interface displays the current risk score of 0.72 in the form of a dashboard, and a highlighted bar chart indicates that "internal status" is the main source of risk. A dialog box pops up in the center of the interface, clearly listing the "activate backup cooling unit" contingency plan that the system recommends "rejecting," and using two side-by-side bar charts to indicate its expected benefit (0.40) and execution cost (0.45). Below the cost bar is a pie chart showing that "secondary risks" account for 70% of the cost composition. At the bottom of the dialog box, there are two buttons: "Manual Approval" and "Confirm Rejection," for managers to make the final decision.

[0161] The IoT logistics information management method described below and the IoT logistics information management platform system described above can be referred to and correspond to each other.

[0162] This invention also provides an Internet of Things (IoT) logistics information management method, which can be operated on as shown in the appendix. Figure 1 The IoT logistics information management platform shown may be deployed in a distributed system consisting of cloud servers and vehicle terminals. See the attached document for details. Figure 8 The flowchart shown, in conjunction with the attached... Figure 1 To be continued Figure 7 The system structure diagram and the application scenario of "cold chain transportation of high-value temperature-sensitive drugs" in Part 3 provide a detailed explanation of the specific implementation steps of the method of the present invention.

[0163] S1. Collect and integrate internal status data, vehicle telemetry data and external environment data related to logistics tasks to form a unified data flow;

[0164] This step is completed collaboratively by the IoT data acquisition terminal 10 and the multidimensional heterogeneous data fusion module 20.

[0165] In a preferred embodiment, namely the scenario of "cold chain transportation of high-value temperature-sensitive pharmaceuticals":

[0166] The Internet of Things (IoT) data acquisition terminal 10 is deployed inside the refrigerated transport vehicle. It collects internal status data at a preset frequency (e.g., every 10 seconds) through built-in or external sensors, specifically including temperature and humidity sensor readings at multiple locations inside the cargo compartment, as well as vibration accelerometer readings on the vaccine packaging boxes.

[0167] Meanwhile, the terminal acquires vehicle telemetry data via the vehicle's CAN bus or OBD interface, including: the vehicle's real-time GPS geographic coordinates, driving speed, engine speed, instantaneous fuel consumption, and the current actual operating power of the cooling unit.

[0168] The collected raw data is sent to the multidimensional heterogeneous data fusion module 20 via a wireless communication network (such as 5G or NB-IoT).

[0169] While receiving the aforementioned internal data and telemetry data, the multidimensional heterogeneous data fusion module 20 actively retrieves external environmental data from third-party service platforms. Specifically, based on the vehicle's current GPS location and planned driving route, it requests hourly and kilometerly weather forecasts (including temperature and precipitation probability) for the next 24 hours from the meteorological service API; and requests real-time traffic events (such as congestion and accidents) for the road ahead from the traffic information service API.

[0170] Finally, the multidimensional heterogeneous data fusion module 20 performs fusion processing on these three types of data from different sources and with different formats: first, it cleans the data and removes outliers; then, it performs spatiotemporal alignment to unify all data to the same timestamp and geographic coordinates; finally, it integrates the processed data into a structured unified data stream containing information from all dimensions (such as a JSON data object) for use in subsequent steps.

[0171] S2. Based on the unified data stream formed in step S1, a comprehensive risk score is calculated in real time;

[0172] This step is performed by the Dynamic Risk Quantification Engine 30.

[0173] After receiving the unified data stream output from step S1, the dynamic risk quantification engine 30 uses its internal risk assessment model to calculate the comprehensive risk score in real time. The model is preferably a weighted fusion algorithm or a machine learning model. In this embodiment, the dynamic risk quantification engine 30 detects external environmental data in the unified data stream indicating that "the vehicle will enter a 38°C high-temperature zone in the next hour." Although the internal temperature state S(t) is perfectly normal at this moment (4.5°C), the risk assessment model... Based on thermodynamic principles and historical data (The vehicle has a history of insulation failure under similar temperature variations), predictively and significantly increasing the level of external environmental threats. The score is calculated based on the overall risk score. The score jumped from the normal 15 points (safe zone) to 65 points (alert zone), and the main source of risk was marked as "anticipated dramatic changes in the external environment".

[0174] S3. Based on the comprehensive risk score calculated in step S2, generate at least one proactive intervention plan aimed at reducing expected risks;

[0175] This step is executed by the adaptive steady-state control module 40.

[0176] The adaptive steady-state control module 40 continuously monitors the comprehensive risk score. The numerical value and its rate of change When detected Once the preset "alert zone" (e.g., score > 60) has been entered, the adaptive steady-state control module 40 is triggered.

[0177] Strategy Matching: The adaptive steady-state control module 40 first identifies the main driver of the risk as "expected external high temperature". Then, it queries and matches in its internal "strategy knowledge base" to find the best practice strategy for this type of risk applicable to the cargo type "temperature-sensitive medicines", namely the "pre-cooling" strategy.

[0178] Contingency Plan Instantiation: The adaptive steady-state control module 40 instantiates the abstract strategy into a concrete, executable proactive intervention plan. The plan includes specific execution parameters, such as: "30 minutes before the expected entry into the high-temperature zone, increase the power of the cooling unit from the current 70% to 95%."

[0179] Quantification of expected returns: The simulation unit within the adaptive steady-state control module 40 will perform a simulation based on the plan to predict the future trend of the comprehensive risk score after the plan is implemented. In the embodiment, the simulation results show that implementing the plan can reduce the predicted risk peak from 90 points to 20 points, therefore the expected risk reduction is calculated to be 70 points.

[0180] S4. Calculate an intervention cost score for the proactive intervention plan generated in step S3;

[0181] This step is executed in parallel with step S3 by the intervention cost deduction module 50.

[0182] After receiving the plan PK, the intervention cost simulation module 50 quantitatively assesses its costs and negative impacts from three dimensions:

[0183] Resource consumption cost ( ): Calculate the additional fuel consumption required to increase the refrigeration unit's power from 70% to 95% and maintain it for a specific period. Based on the vehicle's fuel consumption model and current fuel prices, this is monetized or quantified into a standard score. In this embodiment, the calculated score is 40.

[0184] Secondary risk costs ( This assesses the additional wear and tear on the refrigeration compressor's lifespan due to prolonged high-power operation, as well as the potential risk of "cold injury" (excessively low temperatures) to certain extremely sensitive pharmaceuticals. Based on the equipment depreciation model and cargo characteristics, a standard score is quantified. In this embodiment, a score of 15 is calculated.

[0185] Operational complexity cost ( ): Assess whether driver intervention or complex operations are required to execute this contingency plan. Since the plan can be executed fully automatically via electronic commands, its complexity is 0.

[0186] Finally, by weighted summation = The overall intervention cost score for this plan is 55 points.

[0187] S5. Based on the comparison results of the expected risk reduction amount and the intervention cost score, determine the final regulatory instructions;

[0188] This step is completed in the decision logic unit 61 of the instruction issuance and human-computer interaction module 60.

[0189] The decision logic unit 61 receives the "expected risk reduction" (70 points) from the adaptive steady-state control module 40 and the "intervention cost score" (55 points) from the intervention cost deduction module 50. It executes a core comparison logic: (Expected risk reduction > intervention cost score)

[0190] In this embodiment, 70 > 55, so the condition is met. Therefore, the decision logic unit 61 determines that the proactive intervention plan is "worth executing" and determines the final control instruction as "approving and executing the plan". ".

[0191] If the comparison result is "no" (for example, the cost is 75 points, which is higher than the benefit of 70 points), the decision-making unit will determine the control instruction as "reject automatic execution and switch to manual decision-making" and package all the analysis data.

[0192] S6. Execute the final control command determined in step S5.

[0193] The execution path of this step depends on the decision result of step S5 and is handled by different sub-units of the instruction issuance and human-computer interaction module 60.

[0194] In this embodiment, since the instruction is "Approve and execute plan Pk", the process will be handled by the automatic instruction interface 62:

[0195] The automatic command interface 62 translates the logic command "increase the power of the cooling unit to 95%" into one or more CAN messages that conform to the vehicle CAN bus protocol standard.

[0196] These messages are sent directly to the ECU (Electronic Control Unit) that controls the cooling unit via the vehicle's communication gateway.

[0197] The ECU receives the command and executes it faithfully, increasing the cooling power at the predetermined time.

[0198] Meanwhile, through continuous data collection from S1, the system monitored that the cooling power had indeed been increased and the cargo compartment temperature had begun to drop as expected, thus forming a complete and automated closed-loop control system of "perception-analysis-decision-execution-feedback".

[0199] If the decision is determined to be transferred to a human decision-maker in S5, the process will be handled by the decision support presentation unit 63: this unit will push complete information, including risk analysis, suggested plans, and benefit-cost comparison, to the manager's web backend or mobile app via API, and the final decision will be made by a human.

[0200] This concludes a complete implementation process of the method of this invention. Through this process, the system successfully implemented predictive and cost-effective automated intervention before risks occurred, ensuring the safe transportation of high-value goods.

Claims

1. An Internet of Things (IoT) logistics information management platform system, characterized in that, include: One or more IoT data acquisition terminals are configured to collect internal status data and vehicle telemetry data related to logistics tasks; The multidimensional heterogeneous data fusion module integrates the internal state data, the vehicle telemetry data, and the external environment data to form a unified data stream. The dynamic risk quantification engine calculates a comprehensive risk score in real time based on the unified data stream. The adaptive steady-state control module generates at least one proactive intervention plan aimed at reducing expected risks based on the comprehensive risk score. The intervention cost estimation module receives the proactive intervention plan and calculates an intervention cost score for the proactive intervention plan. The instruction issuance and human-computer interaction module determines and executes the final control instruction based on the comparison result between the expected risk reduction of the proactive intervention plan and the intervention cost score.

2. The IoT logistics information management platform system according to claim 1, characterized in that, In the multidimensional heterogeneous data fusion module, the internal state data includes temperature or humidity data of the cargo environment, the vehicle telemetry data includes real-time location of the vehicle or energy consumption rate of the refrigeration unit, and the external environment data includes future weather forecasts or real-time traffic event data of the transportation route.

3. The Internet of Things (IoT) logistics information management platform system according to claim 1, characterized in that, The dynamic risk quantification engine calculates the comprehensive risk score CRS(t) using the following formula: Wherein, CRS(t) is the comprehensive risk score; f internal Based on the internal state data vector S and its rate of change The internal state risk function; f external Let f be the external environment risk function based on the external environment data vector E; vehicle The vehicle condition risk function is based on the vehicle telemetry data vector V; w i w e and w v These are the corresponding weighting coefficients.

4. The Internet of Things (IoT) logistics information management platform system according to claim 1, characterized in that, The adaptive steady-state control module is configured to generate the proactive intervention plan when the comprehensive risk score exceeds a preset warning threshold or the rate of change of the comprehensive risk score exceeds a preset slope threshold.

5. The Internet of Things (IoT) logistics information management platform system according to claim 1, characterized in that, The intervention cost estimation module is configured to calculate the cost using the following formula. Intervention cost score (ICS) k ): ICS(P k )=c r ·g resource (ΔR k )+c s ·g secondary (ΔP fail ); Among them, ICS(P k ) for proactive intervention plan P k Intervention cost score; g resource To assess the additional resource consumption ΔR k Resource consumption cost function; g secondary To assess the probability increment of secondary failures ΔP fail Secondary risk cost function; c r c s This represents the corresponding cost weighting coefficient.

6. The Internet of Things (IoT) logistics information management platform system according to claim 1, characterized in that, The instruction issuance and human-computer interaction module is specifically configured as follows: When the expected risk reduction is greater than the intervention cost score, the control instruction corresponding to the proactive intervention plan is executed.

7. The Internet of Things (IoT) logistics information management platform system according to claim 1, characterized in that, The instruction issuance and human-computer interaction module is also configured to: when the expected risk reduction is not greater than the intervention cost score, reject the proactive intervention plan and send decision support information containing the comprehensive risk score and the intervention cost score to the human administrator.

8. The Internet of Things (IoT) logistics information management platform system according to claim 5, characterized in that, The resource consumption cost function g resource The assessed additional resource consumption ΔR k This includes additional fuel or electricity consumption resulting from implementing the aforementioned proactive intervention plan.

9. The Internet of Things (IoT) logistics information management platform system according to claim 5, characterized in that, The secondary risk cost function g secondary The assessed secondary failure probability increment ΔP fail This includes the increased probability of failure caused by the overload operation of the vehicle's refrigeration unit due to the implementation of the aforementioned proactive intervention plan.

10. A method for managing logistics information via the Internet of Things, characterized in that, Includes the following steps: S1. Collect and integrate internal status data, vehicle telemetry data and external environment data related to logistics tasks to form a unified data flow; S2. Based on the unified data stream formed in step S1, a comprehensive risk score is calculated in real time; S3. Based on the comprehensive risk score calculated in step S2, generate at least one proactive intervention plan aimed at reducing expected risks; S4. For the proactive intervention plan generated in step S3, calculate an intervention cost score, which is used to characterize the resource consumption required to execute the proactive intervention plan and the potential secondary risks. S5. Based on the comparison between the expected risk reduction of the proactive intervention plan described in step S3 and the intervention cost score calculated in step S4, the final control instruction is determined; S6. Execute the final control command determined in step S5.

Images