Temperature tracking method for mesh network-based cold chain system using clustering

Abstract

A temperature tracking method for a mesh network-based cold chain system using clustering according to various embodiments of the present invention is disclosed. The method may include: a step of forming a cluster among a plurality of packaging devices according to a predefined rule and selecting a cluster representative device; a step of automatically selecting a relay device among the cluster representative devices based on at least one of a battery level, a link score, and a recent connection history with a driver terminal; and a step of establishing a communication connection with the relay device when a new driver terminal enters, and the relay device transmitting sensor data collected from the plurality of packaging devices through each cluster representative device to the driver terminal.

Description

Technology Field

[0001] The present invention relates to a temperature tracking method for a mesh network-based cold chain system using clustering, and specifically, to a temperature tracking method for a cold chain system in which clustering is formed in a mesh network considering scalability and stability. Background Technology

[0003] Cold chain logistics is an essential technology for the safe transportation and storage of temperature-sensitive products such as pharmaceuticals, vaccines, and food. Failure to maintain a consistent low-temperature environment can lead to product quality degradation, safety issues, and financial losses. For this reason, there is a continuously increasing demand for technologies that enable real-time temperature monitoring during transportation, as well as the prediction and response to temperature excursions.

[0004] Conventionally, methods were mainly used to simply record temperatures by attaching data loggers to each transport container, or to collect and transmit data through a single gateway for some vehicles. However, these methods made it difficult to accurately reflect environmental changes at specific locations within individual packaging units, resulted in data loss when communication was interrupted, and had limitations in tracking or predicting temperature anomalies in real time caused by various external factors (weather conditions, vehicle driving conditions, deterioration of cooling device performance, etc.) that may occur during transportation.

[0005] In particular, if a Bluetooth-based data logger is used merely as a recording device, it is difficult to fuse and precisely analyze relative position information between packaging units or environmental data. Additionally, since the distribution of cold air inside the vehicle varies depending on the loading location, surrounding airflow, and vehicle vibration, a single-point measurement method is insufficient to detect the possibility of temperature deviation in advance.

[0006] In conventional technology, short-range communication technologies such as Bluetooth or Wi-Fi were partially applied for location-based monitoring, but they had limitations such as low network stability due to reliance on a simple transmission and reception structure and unstable signal quality due to interference in the vehicle environment. In addition, a structure concentrated on a single gateway contains a single point of failure problem in which the entire monitoring function is paralyzed if the gateway fails.

[0007] Prediction technology for temperature anomalies was also limited. Existing systems were merely capable of providing simple warnings when sensor values ​​exceeded acceptable ranges and failed to account for complex factors such as the rate of temperature change, external weather conditions, and vehicle driving situations. Consequently, it was difficult to predict the likelihood of actual temperature deviations or to take proactive measures for hazardous areas.

[0008] Furthermore, there are limitations in terms of response. Traditional cold chain systems provide only simple warnings upon detecting anomalies and fail to suggest specific response measures. Since no immediate actionable steps for operators on-site are proposed, problem-solving becomes entirely dependent on the operator's experience and intuition. In such cases, consistency in response is compromised, and financial losses can be significant in the event of a response failure. Moreover, because the quantitative evaluation or cost estimation of these losses is not systematically conducted, risk management at the corporate level is also constrained.

[0009] To overcome these limitations, the present invention is characterized by utilizing a mesh network-based packaging device to collect and analyze location and sensor data for each packaging unit in real time, calculating the probability of temperature deviation in advance through an AI-based prediction model, and providing an automatic response scenario in the event of an anomaly. The problem to be solved

[0011] The present invention is devised in response to the aforementioned background technology and aims to provide a temperature tracking method for a mesh network-based cold chain system using clustering.

[0012] The technical problems of the present invention are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0014] According to an embodiment of the present invention for solving the problem described above, a temperature tracking method for a mesh network-based cold chain system using clustering is disclosed. The method may include: a step of forming a cluster among a plurality of packaging devices according to a predefined rule and selecting a cluster representative device; a step of automatically selecting a relay device among the cluster representative devices based on at least one of a battery remaining capacity, a link score, and a recent connection history with a driver terminal; and a step of establishing a communication connection with the relay device when a new driver terminal enters, and the relay device transmitting sensor data collected from the plurality of packaging devices through each cluster representative device to the driver terminal.

[0015] In one embodiment, the step of the plurality of packaging devices forming a cluster according to a predefined rule and electing a cluster representative device may include: each packaging device calculating a representative candidate score of a neighboring packaging device based on a link score and remaining battery capacity, and electing the packaging device having the highest representative candidate score as the cluster representative device, wherein in the case of a tie, the decision is made according to a predefined identifier rule; if there is a representative device that satisfies a join threshold based on the link score, cluster size, and workload of the plurality of cluster representative devices, each packaging device joining the cluster governed by that cluster representative device; and in the case of a packaging device where there is no cluster representative device that satisfies the join threshold, self-electing as a temporary cluster representative device.

[0016] In one embodiment, the step of the plurality of packaging devices forming a cluster according to a predefined rule and selecting a cluster representative device may further include a step of applying a rule to maintain the cluster representative device for a certain period of time, wherein if at least one of battery level deterioration, signal deterioration, and no continuous response exceeds a criterion, the device is automatically replaced with a next-ranked candidate.

[0017] In one embodiment, the step of the plurality of packaging devices forming a cluster according to a predefined rule and selecting a cluster representative device may further include the step of merging two clusters if the link score between different cluster representative devices exceeds a merging criterion for a predetermined period, and dividing by designating a next-ranked representative candidate as the cluster representative device of a new cluster if the cluster diameter or load indicator exceeds a splitting criterion.

[0018] In one embodiment, the step of the plurality of packaging devices forming a cluster according to a predefined rule and electing a cluster representative device may further include the step of setting different join thresholds and withdrawal thresholds when switching between clusters of packaging devices and applying a minimum dwell time.

[0019] In one embodiment, after the step of the plurality of packaging devices forming a cluster according to a predefined rule and electing a cluster representative device, if the location of a specific packaging device changes, the method may further include the step of each packaging device automatically joining the new cluster and re-electing a cluster representative device.

[0020] In one embodiment, the sensor data may include a packaging device identifier, a sequence number, and a timestamp.

[0021] In one embodiment, the method may further include: a step in which, when communication between the driver terminal and the relay device is interrupted, a plurality of packaging devices record sensor data in chronological order in a storage unit; a step in which, when communication is resumed, the driver terminal identifies and requests missing data after the last received point using a sequence number and a timestamp; a step in which a plurality of packaging devices retransmit the missing data in chronological order according to the request; and a step in which the driver terminal sorts the received data in chronological order according to the sequence number and timestamp and removes duplicate data.

[0022] Other specific details of the present invention are included in the detailed description and drawings. Effects of the invention

[0024] The present invention can provide stable data transmission and precise temperature tracking between packaging devices by forming a cluster and dynamically managing a representative device.

[0025] The effects of the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by a person skilled in the art from the description below. Brief explanation of the drawing

[0027] FIG. 1 is a drawing showing the entire mesh network-based packaging device cold chain system according to one embodiment of the present invention. FIG. 2 is a schematic diagram showing the configuration of a mesh network-based packaging device cold chain system according to a first embodiment of the present invention. FIG. 3 is a diagram illustrating the configuration of a packaging device according to a first embodiment of the present invention. FIG. 4 is a schematic diagram showing the configuration of a device according to a first embodiment of the present invention. FIG. 5 is a diagram exemplarily showing the installation location of an anchor according to the first embodiment of the present invention. FIG. 6 is a schematic diagram showing a server according to a first embodiment of the present invention. FIG. 7 is a schematic diagram showing the configuration of a mesh network-based cold chain system using clustering according to a second embodiment of the present invention. FIGS. 8 to 10 are drawings illustrating a temperature tracking method for a mesh network-based cold chain system using clustering according to a second embodiment of the present invention. FIG. 11 is a drawing illustrating a system according to one embodiment of the present invention. FIG. 12 is a hardware configuration diagram of a computing device according to one embodiment of the present invention. FIGS. 13 to 17 are drawings illustrating a method for predicting the probability of temperature deviation in a mesh network-based cold chain system using AI according to a third embodiment of the present invention. FIGS. 18 and 19 are drawings illustrating a method for responding to temperature deviation prediction in a mesh network-based cold chain system using AI according to the fourth embodiment of the present invention. FIG. 20 is a schematic diagram showing one or more network functions related to an embodiment of the present invention. Specific details for implementing the invention

[0028] Various embodiments are now described with reference to the drawings. In this specification, various descriptions are provided to facilitate an understanding of the invention. However, it is evident that these embodiments can be practiced without such specific descriptions.

[0029] As used herein, terms such as “component,” “module,” “system,” etc. refer to computer-related entities, hardware, firmware, software, combinations of software and hardware, or executions of software. For example, a component may be, but is not limited to, a procedure executed on a processor, a processor, an object, an execution thread, a program, and / or a computer. For example, both an application executed on a computing device and the computing device itself may be a component. One or more components may reside within a processor and / or an execution thread. A component may be localized within a single computer. A component may be distributed among two or more computers. Additionally, these components may be executed from various computer-readable media having various data structures stored therein. Components may communicate through local and / or remote processes, for example, according to signals having one or more data packets (e.g., data from a component interacting with another component in a local system or distributed system, and / or data transmitted through signals to other systems and networks such as the Internet).

[0030] Furthermore, the term "or" is intended to mean an implicit "or" rather than an exclusive "or." That is, unless otherwise specified or evident from the context, "X uses A or B" is intended to mean one of the natural implicit substitutions. In other words, if X uses A; if X uses B; or if X uses both A and B, "X uses A or B" may apply to any of these cases. Additionally, the term "and / or" as used herein should be understood to refer to and include all possible combinations of one or more of the enumerated related items.

[0031] Additionally, the terms “comprising” and / or “comprising” should be understood to mean that such features and / or components are present. However, the terms “comprising” and / or “comprising” should be understood not to exclude the presence or addition of one or more other features, components and / or groups thereof. Furthermore, unless otherwise specified or clearly evident from the context to indicate a singular form, the singular in this specification and claims should generally be interpreted to mean “one or more.”

[0032] Those skilled in the art should recognize that the various exemplary logical blocks, configurations, modules, circuits, means, logics, and algorithmic steps described in connection with the embodiments disclosed herein may be implemented in electronic hardware, computer software, or a combination of both. To clearly exemplify the interchangeability of hardware and software, various exemplary components, blocks, configurations, means, logics, modules, circuits, and steps have been generally described above in terms of their functionality. Whether such functionality is implemented in hardware or software depends on the specific application and design constraints imposed on the overall system. Skilled technicians may implement the described functionality in various ways for each specific application. However, such decisions regarding implementation should not be interpreted as moving out of the scope of the invention.

[0033] The description of the presented embodiments is provided to enable those skilled in the art to use or practice the present invention. Various modifications to these embodiments will be apparent to those skilled in the art. The general principles defined herein may be applied to other embodiments without departing from the scope of the present invention. Thus, the present invention is not limited to the embodiments presented herein. The present invention should be interpreted in the broadest possible scope consistent with the principles and novel features presented herein.

[0034] In this specification, the term "computer" refers to any type of hardware device comprising at least one processor, and may be understood to include software configurations operating on said hardware device according to the embodiments. For example, the term "computer" may be understood to include smartphones, tablet PCs, desktops, laptops, and user clients and applications running on each of these devices, but is not limited thereto.

[0035] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

[0036] Each step described in this specification is described as being performed by a computer, but the subject of each step is not limited thereto, and depending on the embodiment, at least some of each step may be performed on different devices.

[0038] FIG. 1 is a diagram showing an overall schematic diagram of a mesh network-based packaging device cold chain system according to the present invention.

[0039] Referring to Fig. 1, the entire logistics flow from the shipment of goods from the seller's warehouse to their arrival at a regional warehouse and final medical institution via transport vehicles is illustrated. During this process, maintaining a cold chain may be essential.

[0040] The present invention is a solution for logistics requiring such cold chain maintenance, and the main components constituting a mesh network-based packaging device cold chain system are illustrated in FIG. 1. First, each packaging device can measure temperature and humidity data generated during transportation in real time. In addition, the packaging device can receive advertising packets from adjacent packaging devices using Bluetooth communication and record the Received Signal Strength Indicator (RSSI), channel information, and Transmission Power (TX Power) to generate basic data for calculating the loading position. The packaging device itself is composed of a cold chain packaging container with thermal insulation performance and can respond to external fluctuations while maintaining the appropriate temperature of the internal contents.

[0041] Multiple packaging devices can be connected to each other via Bluetooth communication to form a mesh network within the vehicle. The driver terminal can collect sensor data and Bluetooth signal data from each packaging device from the mesh network and transmit them to a server.

[0042] Based on the received data, the server can calculate the loading position of each packaging device by estimating distances between devices and aligning coordinates. Subsequently, the calculated positions can be visualized in grid cell units and displayed on a monitoring dashboard. In particular, if a temperature anomaly exceeding the set range is detected in a specific packaging device, the location of that device is highlighted on the operator terminal and monitoring screen, enabling immediate action on-site.

[0043] Accordingly, the overall schematic diagram shown in FIG. 1 can clearly illustrate a series of structures in which the mesh network-based cold chain system provided by the present invention simultaneously tracks temperature and loading location throughout the entire transportation process and supports a rapid response in the event of an anomaly.

[0045] Referring to FIG. 2, a mesh network-based packaging device cold chain system according to a first embodiment of the present invention may include a plurality of packaging devices (10) forming a mesh network including at least one sensor and device, one or more anchors (30) installed at a predetermined location within a vehicle to provide a reference signal, a driver terminal (40) that receives sensor data and Bluetooth signal data of each of the plurality of packaging devices through communication with the mesh network and transmits the received sensor data and Bluetooth signal information to a server (50), and a server (50) that calculates the loading location of each of the plurality of packaging devices based on the sensor data and Bluetooth signal data collected from the plurality of packaging devices, performs temperature monitoring, and provides a notification when a temperature abnormality occurs.

[0046] Referring to FIG. 3, the plurality of packaging devices (10) may be composed of a dedicated transport container (Fig. 3 (a)) containing at least one sensor and a Bluetooth communication-based device (20) (Fig. 3 (b)) connected to the dedicated transport container. Additionally, the device (20) may be included in the dedicated transport container containing at least one sensor.

[0047] Here, according to the first embodiment of the present invention, the at least one sensor may include at least one of a temperature sensor, a humidity sensor, and a vibration sensor.

[0048] FIG. 4 is a drawing for explaining the structure of a device (20) connected to a packaging device (10).

[0049] According to one embodiment of the present disclosure, the device (20) may include at least one of a control unit (21), a communication unit (22), and a storage unit (23).

[0050] The control unit (21) is a central processing unit that controls the overall operation of the device (20) and can be implemented as any one of a microprocessor (MCU), a microcontroller, or an ARM-based system-on-chip (SoC).

[0051] The control unit (21) can process temperature data received from one or more sensors, control read / write operations for the storage unit (23), and manage data transmission and reception through the communication unit (22).

[0052] Additionally, the control unit (21) can monitor the communication connection status with the driver terminal (40) in real time and execute an algorithm to determine whether to back up data by analyzing the terminal's memory usage information. To this end, the control unit (21) may be equipped with a real-time operating system (RTOS) or embedded Linux, and can perform tasks such as temperature data collection, communication management, and storage control in parallel through multi-thread processing.

[0053] The communication unit (22) is a wireless communication module responsible for data communication with the driver terminal (40) and can support various communication methods such as Bluetooth, WiFi, LTE, and 5G.

[0054] The communication unit (22) may include a communication chipset, an RF circuit, and an antenna, and may include a V2X (Vehicle-to-Everything) communication module or a CAN (Controller Area Network) interface, which are vehicle communication standards, for stable communication in a vehicle environment.

[0055] Additionally, the communication unit (22) includes a signal strength measurement function and a packet loss rate detection function for continuously monitoring the communication connection status, and may be equipped with a heartbeat signal transmission and reception function to immediately detect a disconnection.

[0056] The storage unit (23) is a memory system for storing temperature data and can be divided into a first partition and a second partition having different physical characteristics.

[0057] The first partition can be implemented as non-volatile memory capable of large-capacity storage and providing high durability, and specifically, eMMC (embedded MultiMediaCard), UFS (Universal Flash Storage), or industrial SSD (Solid State Drive) can be used.

[0058] The first partition is responsible for the cumulative storage of general temperature data and supports tens of thousands of write / erase cycles, thereby ensuring data integrity even during long-term use. The second partition can be implemented as a high-speed NAND flash memory that provides fast data access speeds. The second partition provides a high write speed to enable rapid data storage in emergency backup situations and supports fast data retrieval and transmission during communication recovery. The storage unit (23) also includes a memory controller to efficiently perform data movement and management between the two partitions, and may be equipped with a super capacitor or backup battery for data protection even when power is cut off.

[0059] Each partition can be implemented physically independently. For example, the first partition and the second partition may be installed separately as different memory chips or separate storage modules, and may be connected to the control unit (21) through independent memory buses and control lines.

[0060] According to the first embodiment of the present invention, the device (20) can record data collected from a sensor, exchange data with an adjacent packaging device, and analyze an advertisement packet received from another packaging device to record at least one of a received signal strength (RSSI), channel information, and transmission power (TX Power) and store it as Bluetooth signal data. The above series of processes can be performed through a control unit (21) of the device (20).

[0061] A plurality of packaging devices (10) are loaded inside the cold chain transport container, and each packaging device (10) can scan and receive advertising packets transmitted from adjacent packaging devices (10). At this time, the received advertising packet includes the transmission power value of the transmitting device, the transmission channel, and the signal strength at the time the signal is received, and each packaging device (10) can generate Bluetooth signal data by recording this in the storage unit (23) of the device (20). In this way, each device continuously accumulates communication signals with surrounding devices, thereby providing basis data that allows the server (50) to determine the relative location and network structure between devices.

[0062] For example, a packaging device located at the front bottom of a transport box may receive relatively strong signal strengths from immediately adjacent devices to the left and right, while receiving relatively weak signal strengths from devices located at the rear of the transport box. This collected signal strength data is stored along with channel characteristics and transmission power values, enabling position calculations that consider not only simple distance estimation between devices but also multipath reflections and radio interference.

[0063] According to the first embodiment of the present invention, the device (20) is not limited to a simple recording device and can perform at least one of TX power multiplexing, channel weighted coupling, and robust estimation.

[0064] In the case of TX power multiplexing, if the control unit (21) of the device (20) repeatedly transmits the same packet with different transmission powers, the receiving device, which is a different packaging device (10), can obtain signals of various strengths, thereby reducing measurement deviations that may occur under single power conditions.

[0065] Specifically, when the control unit (21) of the device (20) continuously transmits the same packet in the order of low power, medium power, and high power, the receiving device can sequentially record the received signal strength at each power. If the two packaging devices (10) are very close in the cargo compartment, the high power sample may have reduced reliability as the receiving circuit becomes saturated, and conversely, if they are far apart or shielded, the low power sample may be buried in noise, making detection unstable. The receiving device can calculate the equivalent received strength at the reference power using the remaining value after excluding the value showing signs of saturation or non-detection among the three observations obtained in this way, or correct the relationship between power and signal strength to combine them into a single representative value. A selection rule can be applied such that the medium power observation is prioritized to avoid saturation in close situations, and the high power observation is prioritized to avoid non-detection in far situations. As a result, excessive fluctuations caused by using only one power in the same environment are reduced, and the input used for path loss estimation or distance estimation is stabilized, thereby reducing the overall error in temperature and location tracking.

[0066] In the case of channel weighted combination, the control unit (21) of the device (20) can improve the accuracy of the final position calculation by evaluating the reliability of signals received from different frequency channels and giving a higher weight to the relatively stable channel.

[0067] Specifically, the control unit (21) of the device (20) can calculate a reliability score based on the packet success rate per channel, the fluctuation range of the received signal strength (RSSI), and the frequency of recent drops for signals entering through different frequency channels within the same time window. For example, a channel with severe interference near the vehicle window can be given a low score because packets are frequently interrupted and the strength fluctuates, while a channel that receives relatively stable signals can be given a high score. Then, the distance estimate or location estimate vector calculated for each channel can be combined as a weighted sum in proportion to the corresponding score to calculate the final location. Even if the value of a specific channel is momentarily skewed due to interference near the window, the final result can be less erratic because the weight of the stable channel is greater. The control unit (21) smoothly updates these weights over time and temporarily excludes channels with rapidly declining reliability, then reflects them again when the condition recovers, thereby maintaining consistency in location calculation even in the event of environmental changes such as entering a tunnel or external wireless interference while driving.

[0068] In addition, in the case of a robust estimation technique, the control unit (21) of the device (20) can remove temporary signal distortion or outliers, thereby minimizing data distortion that may occur in a moving vehicle environment.

[0069] For example, even if the signal of a channel weakens rapidly or becomes excessively large due to reflection as the vehicle enters a tunnel, the control unit (21) determines the values ​​at that moment as abnormal values ​​to reduce their influence and can produce continuous results based on the remaining stable values ​​and the previous movement flow. When the signal returns to normal after exiting the tunnel, the reliability of the values ​​can be raised again and included in the combination. Values ​​that are excluded for reference are stored separately but are not reflected in real-time calculations, thereby minimizing distortion caused by environmental changes during driving and maintaining consistency in position tracking.

[0070] Accordingly, by each of the multiple packaging devices (10) receiving an advertising packet from a peripheral device and refining it using an advanced signal processing technique, the calculation of the loading position inside the cold chain transport container can be made more accurate. This process is not limited to sensor data of individual devices but strengthens the stable connectivity and position matching performance of the entire mesh network, and ultimately supports the accurate identification of the location of a specific packaging device (10) when a temperature anomaly occurs in that device, thereby enabling immediate response during the transport process.

[0071] Referring to FIG. 5, the anchor (30) may be installed on at least one of the floor, ceiling, wall, or corner inside the vehicle transport box, and such location may be a fixed point suitable for forming a reference coordinate system inside the vehicle. For example, an anchor (30) installed on the floor or ceiling may provide a reference in the vertical direction (Z-axis), and an anchor (30) installed on the wall or corner may be advantageous for forming a reference in the horizontal direction (X-axis, Y-axis). When multiple anchors (30) are placed at different locations, the entire space inside the vehicle transport box can be reliably covered, thereby improving the coordinate alignment accuracy of the packaging device (10).

[0072] The anchor (30) can periodically transmit a reference Bluetooth signal from a fixed position. This signal is received by a packaging device (10) and can be compared and analyzed with the received signal strength, channel response, and transmission power between packaging devices. The server (50) collects this signal data, first calculates the relative coordinates between packaging devices, and then converts these relative coordinates into an absolute coordinate system using the reference signal of the anchor (30).

[0073] The driver terminal (40) can be implemented as at least one of a vehicle driver's smartphone, tablet, or dedicated in-vehicle terminal, and can receive sensor data and Bluetooth signal data from a plurality of packaging devices (10) by communicating directly with a mesh network. The driver terminal (40) transmits the received data to a server (50) and can simultaneously provide the driver with visual monitoring results or notification information transmitted from the server (50). Through this, the driver can immediately recognize temperature abnormalities that may occur during transportation and perform on-site response.

[0074] Referring to FIG. 6, the server (50) may include a neighbor table generation unit (51) that collects Bluetooth signal data between packaging devices to form a neighbor table, a distance estimation unit (52) that performs paired distance estimation using the neighbor table, a relative coordinate estimation unit (53) that calculates relative coordinates through the distance estimation result, a matching unit (54) that matches relative coordinates to absolute coordinates by a reference anchor (30), and a grid unit (55) that snaps the absolute coordinates to a grid cell to indicate a loading position.

[0075] The above neighbor table generation unit (51) can form a neighbor table by collecting Bluetooth signals between multiple packaging devices inside the transport box. The neighbor table is a data structure indicating which other devices each packaging device can communicate with, and items such as the received signal strength between each device, signal stability per channel, and transmission power difference are recorded. For example, if the signal strength received by device A located at the front of the transport box from device B on the left and device C at the rear is -50 dBm and -70 dBm, respectively, information that the connection between A and B is stronger than the connection between A and B can be stored in the neighbor table.

[0076] The distance estimation unit (52) can perform pair distance estimation between devices based on the neighbor table configured in this way. This may be a process of converting the signal strength between two specific devices into a distance, or calculating a relative distance value by combining the average signal values ​​from multiple channels. For example, if the distance between AB is estimated to be about 0.5m and the distance between AC is about 1.5m, the relative spatial arrangement of devices A, B, and C can be derived based on this.

[0077] Subsequently, the relative coordinate estimation unit (53) can place the entire packaging device on a single coordinate plane to calculate relative coordinates that satisfy all dual distance constraints. Through this, the relative positional relationship of each device within the transport box can be diagrammed. However, this relative coordinate system may differ from the absolute position within the actual vehicle. Therefore, the alignment unit (54) can convert the relative coordinates into absolute coordinates by referring to the anchor (30) signal installed at a fixed position. For example, if the anchor (30) is installed at the front left corner of the transport box, the coordinates of the anchor (30) are set to (0,0), and all relative coordinates can be repositioned to align with this reference point.

[0078] Finally, the gridding unit (55) can snap the corrected absolute coordinates into predefined grid cell units to indicate which cell each packaging device is actually loaded into. For example, when a transport box is divided into a grid of 4 columns and 3 columns, if a specific device is located at coordinates (1.2, 2.7), it can be automatically assigned to grid (1,3).

[0079] According to the first embodiment of the present invention, when two or more packaging devices are assigned to the same grid cell, the grid unit can perform a local optimization operation by reflecting the estimated distance between the two or more packaging devices and the difference in distance from the surrounding anchor (30) so that each packaging device is assigned to a different grid cell.

[0080] When the server (50) performs a gridding process inside the transport box, the calculated absolute coordinates are converted into grid cell units, and two or more packaging devices may be assigned to the same cell. This may occur when relative coordinates are distorted due to instantaneous interference or multipath reflection of Bluetooth signals, or when two devices are actually adjacent but converge to the same cell when expressed in grid units. In such cases, the server (50) does not simply keep the two devices in the same cell, but can perform a local optimization operation by comprehensively considering the estimated distance between the devices and the difference in distance from the surrounding anchor (30).

[0081] For example, assuming that both device A and device B are assigned to the grid cell corresponding to coordinates (2.4, 1.6), the distance between A and B is calculated to be approximately 0.8m based on the neighbor table distance estimation, confirming that the two devices should be placed at a certain distance from each other. Additionally, device A receives a strong signal from the anchor closer to the left wall, while device B is calculated to have a shorter distance from the rear anchor. Reflecting these conditions, the server can modify the positions by relocating device A to cell (2,2) and device B to cell (3,2).

[0082] This local optimization operation can be performed to resolve collision situations that cannot be resolved by simple coordinate snapping and to derive results that correspond to the actual physical loading state. In particular, when multiple devices are assigned to the same cell, the server (50) can solve the optimization problem while maintaining consistency with the anchor (30) signal at the same time, while minimizing violation of the dual distance constraints of each device. As a result, each packaging device is assigned to a different grid cell and displayed so as to be visually distinguishable, and the location of a specific device can be clearly displayed to the driver terminal (40) when a temperature anomaly occurs.

[0083] According to the first embodiment of the present invention, the server (50) can visually highlight the grid cell coordinates of the packaging device to the driver terminal (40) for the packaging device in which a temperature anomaly is detected.

[0084] Specifically, the server (50) can visually display specific location information in grid units by mapping it to the floor plan of the vehicle transport box on the screen of the driver terminal (40). The driver terminal (40) provides a grid map view that simplifies the interior of the transport box into horizontal and vertical grid cells as a base, and can highlight the latest grid cell of the packaging device where an anomaly has been detected by filling it with a high-contrast color or flashing the border. The highlighting can be overlaid with the device's identifier (ID), current temperature, degree of temperature threshold exceedance, duration of exceedance, and last update time to shorten the on-site judgment time. The driver terminal (40) can ensure that the anomaly event is immediately recognized by combining a short vibration or alarm sound with the visual warning.

[0085] In addition, the highlighting can be guided by taking into account the reference point the driver actually needs to approach. For example, in the case of a vehicle with a rear door opening, a "door reference" icon is fixed at the bottom left of the screen, and an arrow indicating the approximate distance to the highlighted grid cell and the direction of movement is provided, allowing the driver to immediately determine the path inside the vehicle. If necessary, tapping the grid cell switches to zoom mode, which can additionally display useful on-site layout information such as the placement of adjacent cells, the location of surrounding anchors (30), and the width of the passageway.

[0087] Referring to FIG. 7, a mesh network-based cold chain system using clustering according to a second embodiment of the present invention may include a plurality of packaging devices (10) forming a mesh network including at least one sensor and device, a driver terminal (40) that receives sensor data and Bluetooth signal data from each of the plurality of packaging devices (10) through communication with the mesh network and transmits the received sensor data and Bluetooth signal information to a server (50), and a server (50) that performs temperature monitoring based on sensor data collected from the plurality of packaging devices (10) and provides a notification when a temperature abnormality occurs.

[0088] Here, the plurality of packaging devices (10) may be composed of a dedicated transport container including at least one sensor and a Bluetooth communication-based device (20) connected to the dedicated transport container (see FIG. 3). The device connected to the packaging device (10) may include at least one of a control unit, a communication unit, and a storage unit (see FIG. 4), and the control unit of each packaging device (10) may perform a temperature tracking method of a mesh network-based cold chain system using clustering.

[0089] Referring to FIG. 8, a temperature tracking method for a mesh network-based cold chain system using clustering according to a second embodiment of the present invention may include the steps of: forming a cluster of a plurality of packaging devices (10) according to a predefined rule and selecting a cluster representative device (S110); automatically selecting a relay device among the cluster representative devices based on at least one of a battery level, a link score, and a recent connection history with a driver terminal (40) (S120); and when a new driver terminal (40) enters, establishing a communication connection with the relay device and transmitting sensor data collected from the plurality of packaging devices (10) through each cluster representative device to the driver terminal (40) (S130).

[0090] Referring to FIG. 7, for a specific example, a plurality of packaging devices forming a mesh network may form clusters A, B, and C according to predefined rules, and a packaging device indicated by a dotted circle in each cluster A, B, and C may be selected as the representative device of each cluster. The selected cluster representative device may manage sensor data collected from other packaging devices belonging to the cluster and perform the role of relaying connections with the outside of the cluster.

[0091] Next, a relay device may be selected from among the representative devices of clusters A, B, and C based on at least one of the remaining battery level, link score, and recent connection history with the driver terminal. For example, if the representative device of cluster C maintains a relatively high remaining battery level compared to other representative devices, has a stable signal strength and packet success rate with the driver terminal, and has a history of being normally connected to the driver terminal in the past, the representative device of cluster C may be designated as the relay device that establishes a direct communication connection with the driver terminal. The relay device is designated as a single unit within the network and serves as the primary point of contact with external driver terminals, thereby controlling the data flow of the entire cluster.

[0092] When a new driver terminal enters the network, a communication connection can first be established with the relay device. Once the connection is established, the relay device can aggregate sensor data collected from each cluster representative device and transmit it to the driver terminal. During this process, each cluster representative device provides the relay device with sensor data periodically collected from individual packaging devices within the cluster it manages, and the relay device organizes this data in chronological order and transmits it to the driver terminal. The driver terminal transmits the received data to a server to enable centralized temperature monitoring, and can support response during the transportation process by promptly generating an alert if a temperature anomaly is detected.

[0093] This series of processes utilizes the characteristics of a clustering-based mesh network to simultaneously ensure network stability and scalability, and enables the efficient collection and transmission of all data through cluster representative devices and relay devices, regardless of when or where a driver terminal joins the network.

[0094] Referring to FIG. 9, the step of the plurality of packaging devices (10) forming a cluster according to a predefined rule and electing a cluster representative device may include: each packaging device (10) calculating a representative candidate score of surrounding packaging devices (10) based on a link score and remaining battery capacity, and electing the packaging device (10) having the highest representative candidate score as the cluster representative device, and in the case of a tie, determining according to a predefined identifier rule (S111); if there is a representative device that satisfies a membership threshold based on the link score, cluster size, and workload of the plurality of cluster representative devices, each packaging device (10) joins the cluster governed by that cluster representative device (S112); and if there is no cluster representative device that satisfies the membership threshold, the packaging device is self-elected as a temporary cluster representative device (S113).

[0095] Specifically, each packaging device (10) can calculate a link score based on the signal strength and packet success rate received from surrounding packaging devices, and calculate a representative candidate score for surrounding packaging devices (10) along with its own battery level. Each packaging device (10) can declare itself as a representative device candidate if it does not receive a representative declaration signal from a device with a higher score while waiting for a randomly assigned backoff time. If multiple packaging devices (10) are declared as candidates simultaneously, the device with the highest score is finally selected as the cluster representative device after a preset competition time has elapsed, and if there are candidates with the same score, the cluster representative device can be determined according to a predefined identifier rule, such as a rule regarding the size of the device identification number.

[0096] When a cluster representative device is elected in this manner, each packaging device (10) can select a cluster to join by evaluating the link score, the size of the cluster, and the current load for multiple cluster representative devices. For example, if a specific packaging device (10) is located at the boundary between two cluster representative devices, it can compare the link score with the two cluster representative devices, the size of each cluster, and the data processing load, and then select a cluster representative device that satisfies the join threshold and automatically join that cluster. If it does not satisfy the join threshold with any cluster representative device, the packaging device (10) can temporarily self-elect itself as the cluster representative device and re-evaluate the link score with neighboring cluster representative devices at regular intervals. Subsequently, when a suitable cluster representative device that satisfies the join threshold is found, it can join that cluster and release its temporary cluster representative device status.

[0097] According to a second embodiment of the present invention, the step of the plurality of packaging devices (10) forming a cluster according to a predefined rule and selecting a cluster representative device may further include a step of applying a rule to maintain the cluster representative device for a certain period of time, and automatically replacing it with a next-ranked candidate if at least one of battery level deterioration, signal deterioration, and no continuous response exceeds a standard.

[0098] To reiterate, the cluster representative device can be configured to maintain its status for a certain period once elected. This prevents network instability caused by frequent replacement of the representative device and ensures that data management and delivery procedures within the cluster remain stable. However, if the elected cluster representative device is determined to have exceeded a predefined threshold for at least one of the following: low battery level, deteriorating signal quality, or continuous no response, it may be automatically replaced by the next-in-line candidate.

[0099] For example, if the battery level of the cluster representative device drops below 20%, the device can no longer reliably perform the roles of cluster management and data relay, so a replacement procedure for the next-ranked candidate may be initiated. As another example, if the cluster representative device's link score with surrounding packaging devices consistently decreases, resulting in weakened signal strength or a packet success rate falling below a threshold, it may be determined that communication quality has deteriorated and replacement may occur. Additionally, if the representative device fails to respond for a certain number of consecutive times, it may be considered to be in a state of failure or inability to communicate, and the representative status may be automatically delegated to the next-ranked candidate.

[0101] The next-ranked candidate is predefined based on the previously calculated representative candidate scores; for example, a packaging device with the second-highest score within the cluster can be designated as the next-ranked candidate. Once the replacement occurs, the packaging devices within the cluster resynchronize around the new cluster representative device, allowing normal data collection and transmission to continue. This procedure ensures that the replacement of the cluster representative device proceeds naturally without causing network disconnection, thereby securing the stability of the clustering-based mesh network even in long-duration transportation environments.

[0102] According to a second embodiment of the present invention, the step of the plurality of packaging devices (10) forming a cluster according to a predefined rule and selecting a cluster representative device may further include the step of merging two clusters when the link score between different cluster representative devices exceeds a merging criterion for a predetermined period, and dividing by designating a next-ranked representative candidate as the cluster representative device of a new cluster when the cluster diameter or load indicator exceeds a dividing criterion.

[0103] To reiterate, clusters can be resized in response to changes in the transportation environment or fluctuations in network conditions, and accordingly, procedures for merging or splitting clusters may be performed. Specifically, if the link score between representative devices of different clusters exceeds the merger criteria for a certain period, the two clusters may be merged into a single cluster. For instance, if both the signal strength and packet success rate between the representative devices of Cluster A and Cluster B remain high, and the communication quality between the two clusters is deemed stable, the two clusters may be merged and operated under a single representative device. In this case, the device with the higher representative candidate score among the existing cluster representative devices may be designated as the representative device of the newly merged cluster, while the remaining representative devices may be converted into general nodes.

[0104] Conversely, if a specific cluster becomes excessively large or the data transmission load exceeds a threshold, the cluster may be split. For example, if the diameter of cluster C increases, causing a deterioration in communication quality between the cluster representative device and the remote packaging device, or if the amount of sensor data transmitted simultaneously within the cluster becomes excessive and exceeds the processing capacity of the representative device, the system may determine that the splitting criteria have been satisfied. In such cases, the packaging device designated as the next-ranked representative candidate within the existing cluster may be elected as the new cluster representative device to divide the cluster into two.

[0105] Through such merging and splitting procedures, the cluster can be flexibly reconfigured according to network conditions, and as a result, the cluster size and load can be maintained within an appropriate range, enabling stable data collection and transmission.

[0106] According to a second embodiment of the present invention, the step of the plurality of packaging devices (10) forming a cluster according to a predefined rule and electing a cluster representative device may further include the step of setting a different join threshold and a different withdrawal threshold when switching between clusters and applying a minimum dwell time.

[0107] Since frequent switching between clusters can lead to network instability and reduced data transmission efficiency, to control cluster switching of packaging devices, the join threshold and the unsubscribe threshold can be set differently, and a constant minimum dwell time can be applied.

[0108] Specifically, each packaging device (10) determines whether to withdraw when the link score with the representative device of the cluster to which it currently belongs falls below a certain level, and at the same time determines whether to join when the link score with the representative device of another cluster satisfies a level above a certain level. At this time, by setting the withdrawal threshold relatively low and the joining threshold relatively high, control can be made so that the cluster is not unnecessarily moved due to mere momentary signal strength fluctuations or temporary data loss.

[0109] For example, even if the link score with the Cluster A representative device temporarily decreases while the packaging device belongs to Cluster A, it may remain without withdrawing if the value remains higher than the withdrawal threshold. Conversely, a cluster switch may occur only when the link score with the Cluster B representative device is stably maintained above a certain level and satisfies the join threshold.

[0110] In addition, to prevent excessively frequent switching between clusters, a minimum dwell time can be established to restrict the packaging device from moving to another cluster until a certain period has elapsed after joining a cluster. For example, even if the signal strength of Cluster A momentarily increases again immediately after the packaging device joins Cluster B, unnecessary switching can be blocked by ensuring that the device does not leave the cluster before the minimum dwell time has elapsed.

[0111] By distinguishing and applying the subscription threshold and withdrawal threshold in this way and assigning a minimum dwell time, the cluster switching of the packaging device is not influenced by simple temporary environmental changes but is carried out according to stable standards, and as a result, network stability and data transmission reliability can be improved.

[0112] According to a second embodiment of the present invention, after the step of the plurality of packaging devices (10) forming a cluster according to a predefined rule and electing a cluster representative device, if the location of a specific packaging device (10) changes, each packaging device (10) may further include the step of automatically joining to a new cluster and re-electing a cluster representative device.

[0113] Even after multiple packaging devices (10) form a cluster according to predefined rules and elect a cluster representative device, a situation may occur in which the position of the packaging devices (10) changes during the transportation process.

[0114] For example, assuming that a specific packaging device located at the front of the vehicle's cargo compartment is moved to the rear during transport, the signal strength between that device and the representative device of the existing Cluster A weakens, making stable data transmission and reception difficult; conversely, the link score and packet success rate may increase with the representative device of the adjacent Cluster B. In this case, the packaging device may automatically join Cluster B by determining that it satisfies the join threshold of Cluster B.

[0115] Furthermore, since the addition of packaging devices to a new cluster may alter the load or network structure within the cluster, a procedure to re-elect a representative device across the entire cluster may be performed. During this re-election process, the representative candidate scores of the existing representative device and the newly joined devices are compared; considering factors such as battery level, link quality, and workload, the most suitable device may be selected as the representative device. For example, if the existing representative device of Cluster B still possesses sufficient battery level and maintains stable communication, it will remain as the representative device; however, if a newly joined device records a higher score, its representative status may be replaced.

[0116] Such automatic joining and representative device re-election procedures enable the stable maintenance of network connectivity even if the locations of packaging devices frequently change in the transportation environment, and consequently ensure that sensor data generated from multiple packaging devices is collected without interruption and transmitted to the operator terminal and server.

[0117] According to a second embodiment of the present invention, the sensor data may include a packaging device identifier, a sequence number, and a timestamp.

[0118] Specifically, the sensor data collected from each packaging device (10) may include not only simple temperature values ​​but also additional information such as a packaging device identifier, sequence number, and timestamp to clarify the source and order of the data.

[0119] For example, when multiple packaging devices within the same vehicle simultaneously measure and transmit temperature data, the receiving end needs to clearly distinguish which packaging device the data originated from. To this end, the sensor data includes a unique packaging device identifier, allowing the receiving terminal or server to individually track the status of a specific device.

[0120] In addition, in a mesh network environment, since data may be transmitted redundantly or out of order through multiple paths, each piece of data may be assigned a sequence number. For example, if a packaging device generates data every minute, the data may sequentially have sequence numbers such as 001, 002, and 003. Through these assigned sequence numbers, the receiving end can check for missing data and, even if duplicate data is received, can organize it based on the correct order.

[0121] Finally, a timestamp is recorded for each data point, allowing the exact time at which the data was generated to be identified. For example, temperature data measured by a packaging device at 10:05:00 is stored and transmitted, including the timestamp at that point in time. Through this, the server can analyze temperature changes along the time axis based on when the data occurred and can quickly track abnormal phenomena that occurred in specific intervals.

[0122] Referring to FIG. 10, the temperature tracking method of a mesh network-based cold chain system using clustering according to a second embodiment of the present invention may further include the steps of: when communication between the driver terminal and the relay device is interrupted, a plurality of packaging devices (10) recording sensor data in a storage unit in chronological order (S140); when communication is resumed, the driver terminal identifying and requesting missing data after the last received point using a sequence number and a timestamp (S150); the plurality of packaging devices (10) retransmitting the missing data in chronological order according to the request (S160); and the driver terminal sorting the received data in chronological order according to the sequence number and timestamp and removing duplicate data (S170).

[0123] For example, assuming a situation where communication is interrupted for 5 minutes due to a communication failure while driving, packaging devices can sequentially accumulate collected temperature data in a local storage unit during this period. When communication resumes, the driver terminal can identify the missing data based on the sequence number and timestamp it last received and request the data for that period.

[0124] Accordingly, packaging devices can retransmit only the items from the stored data after the requested point in time in chronological order. For example, if the operator terminal has received data up to sequence number 100, the packaging devices extract and transmit only the data after sequence number 101. In this process, duplicate data may be transmitted, but the operator terminal can ultimately secure a continuous and consistent data sequence by utilizing sequence numbers and timestamps to sort the entire data in chronological order and remove duplicate items.

[0125] Through this procedure, even if communication failures occur during transportation, sensor data is reliably restored and transmitted without loss, and the server can accurately monitor temperature changes based on complete time-series data.

[0127] Referring to FIG. 13, a method for predicting the probability of temperature deviation in a mesh network-based cold chain system using AI according to a third embodiment of the present invention may include the step (S210) of a machine learning-based prediction model analyzing sensor data and loading location information collected from a plurality of packaging devices forming a mesh network in a time-series and spatial manner to calculate the probability of temperature deviation of a plurality of packaging devices; the step (S220) of first correcting the probability of temperature deviation by reflecting external weather information and GPS location information of a vehicle in the calculated probability of temperature deviation; and the step (S230) of secondarily correcting the probability of temperature deviation by reflecting seasonal weather changes and cooling performance of a vehicle in the first corrected probability of temperature deviation.

[0128] The above-mentioned method for predicting the probability of temperature deviation in a mesh network-based cold chain system using AI is a technology that calculates the probability of temperature deviation more accurately by stepwise analyzing and correcting various data collected from multiple packaging devices connected by a mesh network.

[0129] Specifically, by analyzing temperature, humidity, and vibration data measured by each packaging device along with Bluetooth signal-based loading location information based on the time and spatial axes, a primary probability of temperature deviation can be calculated that considers not only individual sensor values ​​but also relationships with surrounding devices and repeating patterns by section. Subsequently, by combining this value with real-time external weather information and the vehicle's GPS location information, the probability of temperature deviation can be primarily corrected to reflect contexts where cooling efficiency may decrease even with the same sensor data, such as heatwaves or tunnel passages. Finally, by considering long-term transportation history, seasonal climate characteristics, and changes in cooling performance by vehicle, the probability of temperature deviation can be secondarily corrected to reflect long-term factors, such as high temperatures in summer, freezing sections in winter, or performance degradation of a specific vehicle's cooling system.

[0130] Referring again to Fig. 14, the first temperature deviation probability calculated by analyzing sensor data collected by a mesh network and loading location information in a time-series and spatial manner reflects the values ​​of individual devices and spatial context, but may still contain short-term variability due to external factors. Therefore, in the first correction step, external weather information and vehicle GPS location information are added to eliminate short-term biases caused by external environments and driving conditions such as heatwaves, high humidity, tunnel sections, and congested sections, thereby reflecting the context of short-term and situational factors and increasing the reliability of the prediction results. Subsequently, in the second correction step, long-term and structural factors such as seasonal weather changes and vehicle-specific cooling performance are reflected to correct for risk sensitivity that may vary depending on the season or vehicle characteristics even under the same conditions.

[0131] This dual correction structure suppresses overfitting and promotes the generalization of the prediction model by separating and reflecting short-term and long-term factors, and through stepwise input validation, the overall prediction can be maintained stably even if some data is missing or of low quality. In addition, the cause of the increased probability of temperature deviation can be explained by distinguishing between external environmental factors and vehicle / seasonal factors, thereby enhancing interpretability; from an operational perspective, it enables asynchronous operation where real-time alerts are provided with only the first correction, and the second correction can be applied during idle time.

[0132] Specific examples of prediction models that can be used in the present invention include recurrent neural network-based models such as LSTM (Long Short-Term Memory) or GRU (Gated Recurrent Unit) for handling time series data, and a CNN (Convolutional Neural Network) structure can be applied for spatial pattern analysis. In addition, Temporal Convolutional Networks (TCN) and Transformer-based time series models can be utilized to simultaneously reflect time series and spatial features. Furthermore, by applying ensemble models (Random Forest, Gradient Boosting, XGBoost, LightGBM, etc.) that combine the advantages of multiple models, prediction accuracy can be improved by comprehensively reflecting various input variables such as sensor data, weather data, and vehicle data.

[0133] Here, according to a third embodiment of the present invention, the sensor data may include at least one of temperature, humidity, and vibration data.

[0134] In addition, according to a third embodiment of the present invention, the loading location information can be obtained by estimating the distance between devices using the received signal strength (RSSI) between each packaging device and calculating coordinates within the vehicle based on the estimated distance.

[0135] Here, a detailed explanation of the method for calculating the loading position information of the packaging device was given with reference to FIGS. 2 to 6.

[0136] Referring to FIG. 15, the step of calculating the probability of temperature deviation of a plurality of packaging devices by analyzing sensor data collected from a plurality of packaging devices forming a mesh network and loading location information in a time-series and spatial manner may include: a step of analyzing at least one of the absolute value of sensor data of each of the plurality of packaging devices, the rate of change over time, variability, and deviation from adjacent packaging devices (S211); a step of synchronizing the analyzed sensor data of each of the plurality of packaging devices in a time-series manner to compare and analyze them (S212); and a step of modeling and reflecting a repetitive temperature rise pattern appearing in a specific loading location area (S213).

[0137] The step (S211) of analyzing at least one of the absolute value of sensor data of each of multiple packaging devices, the rate of change over time, variability, and deviation from adjacent packaging devices means evaluating not only individual values ​​but also the pattern of change based on sensor data such as temperature, humidity, and vibration collected from each packaging device. For example, when the absolute temperature value of a specific packaging device is measured as 8°C, this can be used as an indicator to immediately determine whether it is within an acceptable range, and at the same time, if the same device rises from 5°C to 8°C over 10 minutes, a rate of change of 0.3°C per minute is calculated, which can detect the possibility of cooling instability. Furthermore, if the data for the same time period is stably distributed within a range of ±0.5°C centered on an average of 6°C, the variability is evaluated as low; however, if it fluctuates by more than ±2°C, it is judged as high variability and can be interpreted as a long-term risk signal. Furthermore, by comparing the difference in values ​​with adjacent packaging devices loaded in the same area, if a specific device records a temperature significantly higher or lower than the surroundings, the possibility of localized cooling imbalance, packaging defects, or sensor malfunction can be identified. For example, if three of the four devices located at the top rear of the vehicle maintain 5°C while only one records 9°C, this is immediately revealed through the analysis of deviations from adjacent devices and can be reflected as a risk factor in the process of calculating the probability of temperature deviation.

[0138] Next, the step (S212) of chronologically synchronizing and comparing / analyzing the sensor data of each of the multiple packaging devices analyzed above refers to the process of aligning the data collected individually from each device along the same time axis and comparing patterns that appear simultaneously. For example, although packaging devices inside a vehicle record temperatures at one-minute intervals, timestamps may be slightly misaligned due to data transmission delays or communication interference; however, this can be corrected by the server to synchronize the data so that it is aligned at the same time unit. By comparing the aligned data in this way, it is possible to specifically identify whether the temperature of multiple devices rises simultaneously at a specific point in time, or whether temperature changes are pronounced only in devices in certain areas. For instance, if, at the 30-minute mark while driving, the packaging devices in the front area all maintain a temperature of 5°C while the devices in the rear area simultaneously rise by 2 to 3°C, this can be interpreted as a localized cooling imbalance pattern. As another example, if a sudden vibration event is detected at the same time in multiple devices and the temperature temporarily rises immediately afterward, this can be classified as a pattern caused by external factors such as vehicle door opening or loading / unloading events. By comparing time-series synchronized data in this way, it is possible to clearly distinguish between isolated anomalies from individual sensors and actual cooling performance degradation occurring simultaneously across multiple devices, and to improve the accuracy of calculating temperature deviation probabilities by modeling repetitive risk factors in specific sections as patterns.

[0139] The step of modeling and reflecting a repetitive temperature rise pattern appearing in a specific loading location area (S213) means that when multiple packaging devices placed in a specific area inside a vehicle repeatedly exhibit similar temperature rise phenomena during various transportation processes, this is generalized into a statistical and learned pattern and reflected in a prediction model. For example, if it is confirmed in various driving records that the temperature of devices located in the upper rear area of ​​the vehicle repeatedly rises by 2 to 3°C after about one hour of driving, this may suggest that the area is a structurally cooling vulnerable zone due to poor air circulation or insufficient insulation. Such a repetitive pattern can be learned by the model as a continuous phenomenon caused by the characteristics of the area, rather than simply a one-time event occurring during individual driving. Therefore, in subsequent predictions, even if the packaging devices placed in the same area momentarily maintain a normal temperature, the probability of temperature deviation in that area may be calculated as higher because the previously observed repetitive temperature rise pattern is reflected. As another example, if a tendency is detected for the temperature of devices in the area near the vehicle's side door to repeatedly rise by 1 to 2°C immediately after loading or unloading events, this can be modeled as a pattern of external heat inflow upon door opening, and the risk level in the prediction can be automatically increased when the same event occurs subsequently. By patterning the phenomenon of repeated temperature rises in specific location areas and reflecting it in the model in this way, it is possible to calculate a more precise probability of temperature deviation by combining spatial context and past experiential data, rather than simply relying on current sensor data.

[0140] Referring to FIG. 16, the step of first correcting the temperature deviation probability by reflecting external weather information and the vehicle's GPS location information in the calculated temperature deviation probability may include: a step of analyzing the difference value and rate of change between the sensor data and the external weather information (S221); a step of determining at least one vehicle driving condition among whether the vehicle is stopped, driving speed, whether it is an uphill or downhill section, and whether it is a tunnel or an urban section from the vehicle's GPS location information (S222); and a step of correcting the temperature deviation probability by reflecting the change in cooling efficiency predicted according to the determined vehicle driving condition in the result of the difference value and rate of change between the analyzed sensor data and the external weather information (S223).

[0141] The step (S221) of analyzing the difference value and rate of change between the sensor data and external weather information means a process of comparing the temperature and humidity data of the packaging device measured inside the vehicle with the weather data outside the vehicle to quantify the difference between the internal and external environments and to track the change.

[0142] For example, if the internal temperature of a packaging device is 5°C and the external temperature is 30°C at a specific point in time, the temperature difference between the inside and outside is calculated as 25°C; this difference can indicate the level of burden that the cooling device must maintain. If the external temperature was 28°C 30 minutes ago and has now risen to 30°C, the rate of change in the external temperature is calculated as +2°C, which may imply the possibility that the cooling device will face a greater load in the future. In the case of humidity, if the internal relative humidity is 70% and the external relative humidity is 90%, the difference is 20%p; if the external humidity is trending rapidly upward, it can be determined that the risk of condensation on the packaging material or an increase in internal humidity is rising. The internal and external difference values ​​and rates of change calculated in this way clearly reveal the impact of the external environment on the internal state rather than simply looking at absolute values, and can be utilized as key indicators to reflect the burden on cooling performance or factors increasing risk during the subsequent temperature deviation probability correction process.

[0143] The step (S222) of determining at least one vehicle driving situation among whether the vehicle is stopped, driving speed, whether it is an uphill or downhill section, and whether it is a tunnel or urban section from the vehicle's GPS location information means a process of deriving the vehicle's driving context by interpreting the collected location data and speed data along the time axis.

[0144] For example, if GPS coordinates remain unchanged for a certain period and the speed value stays at 0 km / h, the vehicle is identified as being stationary; this suggests that the cooling system receives less external air circulation, which may lead to reduced cooling efficiency. Conversely, if the speed value remains constant while the coordinates change linearly, it can be considered a normal driving situation. Additionally, analyzing GPS altitude values ​​and route data together allows for determining whether the vehicle is traveling uphill or downhill; on uphill sections, there is a high probability that the vehicle's internal temperature will rise due to engine load. In the case of tunnels, if the GPS signal weakens rapidly or cuts off, and the location on the map data corresponds to a tunnel, the vehicle may be recognized as passing through a tunnel; this implies a situation where external convection is blocked, leading to predicted reduced cooling efficiency. Finally, if GPS coordinates correspond to a dense urban area and the average speed remains low, it can be identified as urban driving due to traffic congestion or waiting at traffic lights; this increases the likelihood of internal temperature instability caused by external heat and frequent stops. In this way, by determining the detailed driving conditions of the vehicle based on GPS location information, the probability of temperature deviation can be more precisely corrected according to the situational context, even with the same internal sensor data.

[0145] The step (S223) of correcting the temperature deviation probability by reflecting the change in cooling efficiency predicted according to the determined vehicle driving situation in the difference value and rate of change result between the analyzed sensor data and external weather information means a process of adjusting the probability value by combining the risk indicator calculated from the difference between the internal and external environments with the vehicle's driving context to consider situations where the actual cooling performance may change.

[0146] For example, if a vehicle enters a tunnel while the internal temperature is maintained at 5°C but the external temperature is high at 35°C—with a temperature difference of 30°C—the reduced external airflow makes it highly likely that the cooling system will not operate as efficiently as usual. In this case, the predicted probability is calculated higher by adjusting for the decrease in cooling efficiency in the previously calculated probability of temperature deviation. As another example, if a vehicle is stopped for a long time in a congested urban area, even if the rate of change in external temperature is not significant, external heat and engine heat accumulate inside the vehicle, which can reduce cooling performance; thus, the probability of temperature deviation can be adjusted upward. Conversely, when a vehicle is traveling at high speed, even if the external temperature is high, the airflow actually improves the efficiency of the cooling system; therefore, the probability of temperature deviation can be adjusted downward under the same difference value and rate of change conditions. In this way, by combining sensor data and external weather indicators with vehicle driving conditions, it is possible to precisely reflect actual changes in cooling performance that are difficult to obtain through simple numerical comparisons.

[0147] According to a third embodiment of the present invention, the step of first correcting the temperature deviation probability by reflecting external weather information and the vehicle's GPS location information in the calculated temperature deviation probability may further include the step of detecting a vehicle door opening event by combining the vehicle's GPS location information and vibration data, and the step of correcting the temperature deviation probability by reflecting the possibility of cold air leakage predicted from the difference value and rate of change result between the analyzed sensor data and the external weather information.

[0148] The step of detecting a vehicle door opening event by combining the vehicle's GPS location information and vibration data refers to a process of recognizing a door opening situation occurring during loading, unloading, or inspection by simultaneously analyzing changes in the vehicle's location and vibration signal patterns.

[0149] For example, if a vehicle arrives at a designated coordinate at a logistics center and remains stationary with its GPS location unchanged for a certain period, while momentary strong vibrations and repetitive micro-vibrations are simultaneously detected by packaging equipment, this can be interpreted as signals related to door opening and worker movement of goods. As another example, if a vehicle briefly stops in front of a building during an urban delivery route and a vibration sensor captures a distinctive impact signal associated with the opening and closing of a door at that moment, that section can be detected as a door-opening event. In this way, while it is difficult to distinguish between a simple stop and an operational event using GPS information alone, considering vibration data alongside it allows for the determination of whether the vehicle's door has actually been opened with high reliability. Such door-opening event detection plays a crucial role in identifying moments when cold air is highly likely to escape to the outside, and can subsequently be utilized as foundational information to reflect the risk of cold air leakage during the temperature deviation probability correction process.

[0150] The step of correcting the temperature deviation probability by reflecting the possibility of cold air leakage predicted from the difference value and rate of change results between the analyzed sensor data and external weather information refers to a process of estimating the impact on the actual internal temperature by combining the vehicle door opening event and external environmental conditions, and reflecting this in the predicted probability.

[0151] For example, if a vehicle is stopped at a logistics center during the daytime in summer under conditions of an outside temperature of 35°C and 80% humidity, and it is detected that the door has been opened, it can be determined that there is a high probability of cold air rapidly escaping to the outside, even if the internal sensor data still maintains 5°C. In this case, the model can compensate for the risk of temperature deviation by weighting the probability that the internal temperature will rise within the next few minutes, taking into account the time the door was opened and the temperature difference between the outside and inside. Conversely, if the door is opened while the outside temperature is -5°C and the vehicle interior is maintained at 2°C in winter, there is a possibility that the internal temperature will drop rapidly due to the influx of cold outside air; therefore, this leads to a risk of freezing rather than cold air leakage, and a separate compensation may be applied.

[0152] As another example, considering GPS location data allows for the differentiation of cold air leakage correction levels by determining the nature of the vehicle's stopping location. When a vehicle arrives at a logistics center and remains stationary for an extended period, it can be recognized as a collection operation involving the simultaneous loading and unloading of multiple packages; in this scenario, the doors remain open for a long time, increasing the likelihood of cold air leakage due to the temperature difference between the inside and outside. Therefore, even if the internal temperature remains within the normal range, the prediction model can significantly increase the probability of temperature deviation by combining the condition of prolonged opening with the rate of change in external temperature. Conversely, when the GPS coordinates correspond to an individual delivery location and the vehicle stops briefly within a few minutes before moving on, it is recognized as an unloading operation for a single or small quantity of goods; since the door opening time is limited, the impact of cold air leakage may be relatively small. In this case, the prediction model can reduce the correction range under the same external environmental conditions to suppress excessive alarms and adjust the probability of temperature deviation to distinguish it from actual high-risk situations.

[0153] Referring to FIG. 17, the step of secondarily correcting the temperature deviation probability prediction by reflecting seasonal weather change patterns and vehicle-specific cooling performance characteristics in the firstly corrected temperature deviation probability may include: a step of obtaining seasonal weather change patterns by collecting past weather data from an external weather agency (S231); a step of obtaining vehicle-specific cooling performance characteristics by analyzing vehicle-specific sensor data and cooling device performance changes from the past transportation history of the dispatch solution platform (S232); and a step of correcting the temperature deviation probability by reflecting the results of modeling the effect on the temperature change of the packaging device by matching the seasonal weather change patterns and vehicle-specific cooling performance characteristics based on the vehicle's GPS location information and time axis (S233).

[0154] The step of collecting past weather data from an external weather agency to obtain seasonal weather change patterns (S231) refers to the process of securing long-term weather data corresponding to the region and period in which transportation takes place, processing it on a seasonal basis, and analyzing repetitive environmental changes.

[0155] For example, by collecting hourly temperature, humidity, solar radiation, and wind speed data from the past five years provided by sources such as the Korea Meteorological Administration, it can be confirmed that during the summer in a specific region, a pattern repeats where average temperatures rise above 30°C, daytime solar radiation is strong, and humidity increases rapidly. Conversely, during the winter, average temperatures drop below freezing, daylight hours shorten, and a trend of increased risk of freezing may appear at specific times. Deriving these seasonal weather change patterns allows for the incorporation of seasonal sensitivity into prediction models, going beyond simply reflecting current weather conditions. This enables statements such as, "The risk of temperature rise is greater in the afternoon during summer even under the same external temperature conditions," or "The risk of freezing is higher in the early morning during winter even under the same humidity conditions." Therefore, this step provides foundational information that allows sensor data from packaging devices to be interpreted differently according to the seasonal cycle, and plays a crucial role in correcting for long-term temperature deviation risks.

[0156] The step (S232) of obtaining vehicle-specific cooling performance characteristics by analyzing vehicle-specific sensor data and changes in cooling device performance from the past transportation history of the dispatch solution platform refers to the process of extracting a pattern of cooling maintenance capability based on transportation records that the same vehicle has repeatedly performed over a long period of time.

[0157] For example, if an analysis of a specific vehicle's transportation history over the past year reveals that the interior temperature was stably maintained at around 5°C during most of the operation when the outside temperature was 32°C during the summer, that vehicle can be evaluated as having excellent cooling performance. Conversely, if another vehicle frequently experienced interior temperatures rising to 8–10°C under the same conditions, it could be classified as a vulnerable vehicle due to cooling system degradation or insulation issues. Furthermore, if a specific vehicle repeatedly experiences temperature increases as the load increases, it can be interpreted as possessing characteristics sensitive to loading conditions. By analyzing sensor data such as temperature, humidity, and vibration alongside cooling system operation logs from transportation history data, it is possible to distinguish between vehicle types that react quickly, those that gradually degrade, and those that become vulnerable under specific conditions. These vehicle-specific cooling performance characteristics are subsequently reflected as individualized sensitivities when adjusting for temperature deviation probabilities, allowing for the specific consideration of vehicle-specific differences even under identical environmental conditions.

[0158] The step (S233) of correcting the temperature deviation probability by reflecting the results of modeling the impact on the temperature change of the packaging device by matching the above-mentioned seasonal weather change patterns and vehicle-specific cooling performance characteristics based on the vehicle's GPS location information and time axis means a process of calculating the degree of influence on the temperature change by combining external environmental factors and vehicle performance factors based on the actual transportation route and time point, and reflecting this in the prediction probability.

[0159] For example, if GPS data confirms a vehicle is passing through a hot and humid urban area during the afternoon in summer, and the vehicle has previously exhibited a pattern of recurring internal temperature increases of 2–3°C in this section, the probability of temperature deviation can be significantly adjusted upward by simultaneously reflecting the seasonal high-temperature pattern and the vehicle's performance vulnerability characteristics. Conversely, for a vehicle that maintains stable cooling performance even within the same urban section, the adjustment range for the probability of temperature deviation may be smaller under the same external conditions. As another example, if the GPS points to a mountainous region during the early morning hours in winter and the external temperature drops below -10°C, and the vehicle's past transportation history shows a risk of freezing due to excessively low internal temperatures under these conditions, the probability of temperature deviation can be adjusted higher toward the risk of freezing by matching the seasonal low-temperature pattern with the vehicle's performance data. In this way, by matching seasonal environmental patterns with vehicle-specific cooling performance data based on GPS and the time axis, it is possible to quantitatively reflect repetitive trends learned from past data beyond a simple interpretation of current sensor values; consequently, the probability of temperature deviation can be adjusted to more accurately reflect actual transportation conditions.

[0161] Referring to FIG. 18, a method for responding to a temperature deviation prediction in a mesh network-based cold chain system using AI according to a fourth embodiment of the present invention may include: a step of monitoring the temperature deviation probability of each of a plurality of packaging devices in real time (S310); a step of generating a response scenario based on the inferred cause of the temperature deviation and presenting it to an operator terminal when the temperature deviation probability exceeds a first threshold (S320); a step of terminating the monitoring result and saving the data when the operator accepts the response scenario and returns to below the first threshold (S330); a step of determining whether the second threshold is exceeded when the first threshold is not exceeded, and calculating the loss cost and notifying the administrator terminal when the second threshold is exceeded (S340); and a step of shortening the monitoring cycle and determining whether the first threshold is not exceeded and whether the second threshold is exceeded when the operator does not accept the response scenario (S350).

[0162] Here, according to the fourth embodiment of the present invention, the temperature deviation probability is calculated by a machine learning-based prediction model analyzing sensor data collected from a plurality of packaging devices forming a mesh network and loading location information in a time-series and spatial manner, and the sensor data may include at least one of temperature, humidity, and vibration data.

[0163] Here, a detailed explanation of the method for calculating the probability of temperature deviation of a packaging device was given with reference to FIGS. 13 to 17.

[0164] To explain in detail with reference to FIG. 19, in the step (S310) of monitoring the temperature deviation probability of each of the plurality of packaging devices in real time, sensor data such as temperature, humidity, and vibration collected periodically through a mesh network by the plurality of packaging devices loaded inside the vehicle can be transmitted to a server via a driver terminal. The transmitted data is stored synchronized with the time axis and combined with relative location information between devices to track abnormal patterns such as temperature rise or sudden humidity changes occurring in a specific loading section, and can be displayed on the user interface of the manager terminal or driver terminal.

[0165] In the step (S320) of generating a response scenario based on the inferred cause of temperature deviation and presenting it to the driver terminal when the probability of temperature deviation exceeds a first threshold, the prediction model can infer the cause of the temperature deviation by synthesizing collected sensor data and external context information. For example, if the vehicle's GPS location information and vibration sensor data change simultaneously in a specific section, suggesting that loading and unloading occurred, the system can determine that cold air leaked as the vehicle door opened. As another example, if the temperature difference between packaging devices increases while the external temperature rises rapidly, the cause can be inferred to be an increase in the load of the cooling device or a cooling imbalance occurring at a specific loading location. The response scenario may vary depending on the cause inferred in this way.

[0166] Here, according to the fourth embodiment of the present invention, the corresponding scenario may include a scenario that guides the use of an auxiliary cooling device in a loading location area with a high probability of temperature deviation.

[0167] In addition, according to the fourth embodiment of the present invention, the auxiliary cooling device may include at least one of a refrigerant pack, a small Peltier cooling unit, a battery-based small refrigeration unit, and a small fan unit.

[0168] When a specific loading location with a high probability of temperature deviation is identified, the system can present a response scenario to the driver terminal to mitigate the temperature rise in that area.

[0169] For example, if a rapid rise in temperature is detected in the rear interior of the vehicle due to high outside temperatures during the summer, the driver may be instructed to place an additional auxiliary cooling unit in that area. In this case, the auxiliary cooling unit can be selected in various forms depending on the situation.

[0170] In cases requiring a short-term response, stable cooling can be maintained for a certain period by immediately inserting pre-frozen refrigerant packs into the designated storage location. In environments where power is available, small Peltier cooling units can be installed to induce localized temperature reduction; for situations requiring longer-term support, small battery-based refrigeration units can be utilized to temporarily store high-value products or sensitive pharmaceuticals, thereby preventing quality degradation. Additionally, if the temperature rises locally simply due to poor air circulation, small fan units can be used to guide the even distribution of cold air.

[0171] As such, the selection of the auxiliary cooling device may vary depending on the inferred cause and field conditions, and the operator can minimize the possibility of temperature deviation by taking immediate action based on the scenarios presented on the terminal.

[0172] If the cause of the temperature deviation is presumed to be a vehicle door opening event, the system may display a warning message to the driver terminal to immediately close the door. In particular, in environments with high outside temperatures during loading and unloading operations, the cold air inside the vehicle can be rapidly lost even if the door is left open for only a short period of time, so the driver may be provided with specific instructions to quickly close the door to minimize cold air loss.

[0173] If sudden external weather changes are identified as the cause, a scenario may be presented to the driver terminal instructing the prediction model to adjust the vehicle's driving speed and route by considering weather information such as heatwaves, extreme cold, or heavy rain. For example, if cooling efficiency is reduced in congested urban areas due to a heatwave, it may be recommended to select an alternative route or reduce stopping time.

[0174] If performance degradation of the cooling system is identified as the cause, the system may provide a scenario in which it recommends the driver inspect the cooling system, in addition to instructing the operation of the auxiliary cooling system. If it detects that the vehicle's cooling system is failing to maintain the target temperature, the driver may be instructed to send a request to the control server to inspect the system.

[0175] In cases where air circulation is restricted due to unbalanced loading, a scenario may be presented to the driver to reposition the cargo. For example, if cargo concentrated at the front of the vehicle blocks the flow of cold air, moving some of the cargo to the rear can ensure air circulation, thereby mitigating localized temperature increases in specific areas.

[0176] As such, response scenarios are not limited to simply adding or supplementing cooling devices, but may include various alternatives such as blocking cold air loss, optimizing driving conditions, inspecting devices, and rearranging loads, which can be provided to the driver terminal in real time in accordance with the inferred cause of temperature deviation and on-site conditions.

[0177] In addition, according to the fourth embodiment of the present invention, when the probability of temperature deviation exceeds a first threshold, the step (S320) of generating a response scenario based on the inferred cause of temperature deviation and presenting it to the driver terminal may include the step of providing a plurality of response scenario candidates by assigning priority according to at least one of cost, effectiveness, and ease of execution.

[0178] In the process of providing the generated response scenarios to the driver terminal, multiple response measures may be proposed simultaneously for the same temperature deviation situation. At this time, the terminal may not simply list all measures, but may display them with priority based on evaluation factors such as cost, effectiveness, and ease of execution for each response scenario. For example, if a temperature rise occurs at a specific loading location, the system may simultaneously propose methods such as adding refrigerant packs, operating small fans, or relocating some cargo. In this case, adding refrigerant packs provides a significant immediate cooling effect but incurs costs due to the use of consumables, while relocating cargo is less costly but may be difficult to execute. On the other hand, operating small fans is easy to execute and incurs almost no cost, but the cooling effect may be limited. The computing device (100) synthesizes these factors to sort each scenario by priority and displays it on the driver terminal, allowing the driver to select the most suitable response measure depending on the situation. This method of providing solutions with priority considerations supports the driver in making the most rational decision within a limited time and can minimize unnecessary expenditure or the selection of unfeasible measures.

[0179] According to the fourth embodiment of the present invention, when the driver accepts the response scenario, the step (S330) of terminating the monitoring result and saving the data when it returns to below the first threshold may include the step of storing whether the response scenario was executed and the result thereof in a database to retrain a machine learning-based prediction model, and automatically recommending a countermeasure by referring to past temperature deviation response history data when the same condition is repeated.

[0180] At this time, the computing device (100) can not only store sensor data but also record whether a response scenario was executed and the result thereof. For example, if the operator selects a scenario in which additional refrigerant packs are placed when the probability of a temperature deviation in a specific loading area exceeds a first threshold, and the temperature of the area is actually stabilized within a certain period of time, this information can be stored in the database as a success history of "refrigerant pack use → temperature normalization." Conversely, if the operator does not take action in the same situation or only operates the fan but eventually exceeds the second threshold and incurs losses, this information can be stored as a history of "response failure."

[0181] The historical data accumulated in this way is utilized to retrain machine learning-based predictive models, enabling them to autonomously learn through machine learning which response scenarios were most effective under specific conditions. For example, if data accumulates showing that when the outside temperature exceeds 30°C during the summer and a vehicle is stuck in traffic, using a refrigerant pack had a high probability of success while operating only the fan had a high probability of failure under the same conditions in the past, the system can prioritize recommending the use of the refrigerant pack when the same or similar conditions are repeated. In this manner, the machine learning-based predictive model can present increasingly precise and situation-appropriate countermeasures over time, and drivers can receive recommendations that reflect accumulated experience.

[0182] According to the fourth embodiment of the present invention, the step (S340) of determining whether the second threshold is exceeded when the first threshold is not returned to or below, and calculating the loss cost and notifying the administrator terminal when the second threshold is exceeded, may include the step of calculating the monetary loss amount based on at least one of the predicted temperature deviation time, the actual unit price of the product, the rate of exceeding the allowable storage time per product, and the transportation distance.

[0183] If a situation occurs where the second threshold is exceeded, the computing device (100) can not only provide a risk notification but also quantitatively calculate the resulting financial loss and transmit it to the administrator terminal.

[0184] For example, assuming that the time for a packaging device at a specific loading location to exceed a secondary threshold in a vehicle transporting expensive vaccines requiring refrigerated distribution is predicted to be 2 hours, the computing device (100) can calculate the amount of loss by reflecting the unit price of the product and the rate of exceeding the allowed storage time. If the unit price of one box of vaccine is 1 million won and a 50% quality degradation occurs for every 1 hour the allowed storage time is exceeded, the computing device (100) can calculate a loss of 1 million won by considering the possibility of discarding the entire quantity when exceeding 2 hours. As another example, reflecting the characteristic that the risk of spoilage increases as the transportation distance increases in the case of fresh food, a higher amount of loss can be estimated for transportation over longer distances even under the same time-exceeding conditions. In this way, the calculation of loss costs is not limited to a simple time-exceeding status, but comprehensively considers factors such as the actual unit price of the product, the rate of exceeding the allowed storage time for each product, and the transportation distance, enabling the manager to immediately identify economic risks and decide on follow-up measures.

[0185] Subsequently, if the driver does not accept the response scenario, in the step (S350) of determining whether to return to below the first threshold and whether to exceed the second threshold by shortening the monitoring cycle, sensor data can be collected and analyzed at shorter intervals than before to track the temperature change trend more precisely.

[0186] For example, if data is collected at 5-minute intervals, it is possible to quickly recognize the moment when a temperature rise or a sudden change in humidity occurs without missing it. This method allows the computing device (100) to actively assess the severity of the dangerous situation and prepare for additional response, even if the driver does not take immediate action.

[0187] For example, if the probability of a temperature deviation at a specific loading location continues to increase but the operator fails to execute the suggested refrigerant pack usage scenario, the system can shorten the monitoring cycle to closely verify whether the temperature in that area actually stabilizes or continues to rise. If the temperature naturally returns below a certain level, the problem can be recorded as resolved without executing the response scenario; however, if the temperature continues to rise, it is determined that the likelihood of exceeding a secondary threshold has increased, allowing for an immediate transition to the loss cost calculation stage. This step of shortening the monitoring cycle serves as a safety net to minimize risks arising from a lack of response and assists managers in recognizing the situation in advance.

[0189] FIG. 20 is a schematic diagram showing one or more network functions related to one embodiment of the present invention.

[0190] Throughout this specification, artificial intelligence model, neural network model, neural network, network function, and neural network may be used interchangeably. A neural network may generally consist of a set of interconnected computational units that may be referred to as “nodes.” These “nodes” may also be referred to as “neurons.”

[0191] A deep neural network (DNN) may refer to a neural network that includes multiple hidden layers in addition to input and output layers. Using a deep neural network allows for the identification of latent structures in data. That is, it is possible to identify the latent structures of photos, text, videos, voice, and music (e.g., what objects are present in a photo, what the content and emotions of a text are, what the content and emotions of a voice are, etc.). Deep neural networks may include convolutional neural networks (CNN), recurrent neural networks (RNN), autoencoders, Generative Adversarial Networks (GAN), restricted Boltzmann machines (RBM), deep belief networks (DBN), Q networks, U networks, Siamese networks, etc. The description of deep neural networks described above is merely illustrative and the present invention is not limited thereto.

[0192] Neural networks can be trained using at least one of supervised learning, unsupervised learning, and semi-supervised learning. The purpose of training a neural network is to minimize the error in the output. The training process involves repeatedly inputting training data into the network, calculating the error between the network's output and the target for the training data, and updating the weights of each node in the network by backpropagating the error from the output layer to the input layer in a direction that reduces the error. In the case of supervised learning, training data is used where the correct answer is labeled for each data point (i.e., labeled training data), whereas in the case of unsupervised learning, the correct answer may not be labeled for each training data point. For instance, in the case of supervised learning regarding data classification, the training data may consist of data where each training data point is labeled with a category. Labeled training data is input into the neural network, and the error can be calculated by comparing the network's output (category) with the label of the training data. As another example, in the case of unsupervised learning for data classification, the error can be calculated by comparing the input training data with the neural network output. The calculated error is backpropagated within the neural network (i.e., from the output layer to the input layer), and this backpropagation can update the connection weights of each node in each layer of the neural network. The amount of change in the connection weights of each updated node can be determined by the learning rate. The neural network's calculation of the input data and the backpropagation of the error can constitute a learning cycle (epoch). The learning rate can be applied differently depending on the number of iterations of the neural network's learning cycle.For example, a high learning rate can be used during the early stages of neural network training to quickly achieve a certain level of performance and increase efficiency, while a low learning rate can be used during the later stages to improve accuracy.

[0193] In the training of neural networks, the training data is generally a subset of the real-world data (i.e., the data intended to be processed by the trained neural network). Consequently, a training cycle may exist where errors decrease on the training data but increase on the real-world data. Overfitting is a phenomenon where the network learns excessively on the training data, leading to increased errors on real-world data. For example, a neural network trained on yellow cats might fail to recognize cats when seeing anything other than yellow, which can be considered a type of overfitting. Overfitting can act as a cause for increased errors in machine learning algorithms. Various optimization methods can be used to prevent this overfitting. To prevent overfitting, methods such as increasing the training data, regularization, or dropout—which involves omitting some nodes from the network during the training process—can be applied.

[0194] The steps of the method or algorithm described in connection with embodiments of the present invention may be implemented directly in hardware, implemented as a software module executed by hardware, or implemented by a combination thereof. The software module may reside in RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), Flash Memory, a hard disk, a removable disk, a CD-ROM, or any form of computer-readable recording medium well known in the art to which the present invention belongs.

[0195] The components of the present invention may be implemented as a program (or application) and stored on a medium to be executed in combination with a computer, which is hardware. The components of the present invention may be implemented as software programming or software elements, and similarly, embodiments may be implemented in programming or scripting languages ​​such as C, C++, Java, assembler, etc., including various algorithms implemented as combinations of data structures, processes, routines, or other programming configurations. Functional aspects may be implemented as algorithms executed on one or more processors.

[0196] Those skilled in the art will understand that the various exemplary logic blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented by electronic hardware, various forms of programs or design code (referred to herein as “software”), or a combination of all such. To clearly illustrate this interoperability between hardware and software, various exemplary components, blocks, modules, circuits, and steps have been generally described above in relation to their functions. Whether such functions are implemented as hardware or software depends on the design constraints imposed on the specific application and the overall system. Those skilled in the art may implement the functions described in various ways for each specific application, but such implementation decisions should not be interpreted as being outside the scope of the invention.

[0197] The various embodiments presented herein may be implemented as methods, devices, or articles of manufacture using standard programming and / or engineering techniques. The term “article of manufacture” includes a computer program, carrier, or medium accessible from any computer-readable device. For example, computer-readable media include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical discs (e.g., CDs, DVDs, etc.), smart cards, and flash memory devices (e.g., EEPROMs, cards, sticks, key drives, etc.). Additionally, the various storage media presented herein include one or more devices and / or other machine-readable media for storing information. The term “machine-readable media” includes, but is not limited to, wireless channels and various other media capable of storing, holding, and / or transmitting command(s) and / or data.

[0198] It should be understood that the specific order or hierarchy of steps in the presented processes is merely an example of exemplary approaches. It should be understood that, based on design priorities, the specific order or hierarchy of steps in the processes may be rearranged within the scope of the invention. The appended method claims provide various step elements in a sample order, but do not imply limitation to the specific order or hierarchy presented.

[0200] Although embodiments of the present invention have been described above with reference to the attached drawings, those skilled in the art will understand that the present invention may be implemented in other specific forms without altering its technical concept or essential features. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive.

Claims

Claim 1 A temperature tracking method using a mesh network between multiple packaging devices comprises: a step in which the multiple packaging devices form a cluster according to a predefined rule and elect a cluster representative device; a step in which, whenever a new driver terminal enters, a relay device is selected among the cluster representative devices based on at least one of a battery level, a link score, and a recent connection history with the driver terminal, and a communication connection is established between the selected relay device and the driver terminal; and a step in which, when a new driver terminal enters, a communication connection is established with the relay device, and the relay device transmits sensor data collected from the multiple packaging devices through each cluster representative device to the driver terminal; wherein the step in which the multiple packaging devices form a cluster according to a predefined rule and elect a cluster representative device comprises: each packaging device calculating a representative candidate score of surrounding packaging devices based on a link score and a battery level; waiting for a randomly assigned backoff time and, if it does not receive a representative declaration signal from a device with a higher score than itself, declaring itself as a representative device candidate; finally electing the device with the highest score as the cluster representative device at the time when a predefined contest time has elapsed; and in the case of a tie, according to a predefined identifier rule A temperature tracking method for a mesh network-based cold chain system using clustering, comprising: a step of determining accordingly; a step of joining a cluster governed by a cluster representative device if there exists a representative device satisfying a join threshold based on link scores, cluster size, and workload for multiple cluster representative devices for each packaging device; and a step of self-electing as a temporary cluster representative device in the case of a packaging device where there is no cluster representative device satisfying the join threshold. Claim 2 delete Claim 3 A temperature tracking method for a mesh network-based cold chain system using clustering, wherein, in claim 1, the step of forming a cluster of multiple packaging devices according to a predefined rule and selecting a cluster representative device further comprises the step of applying a rule to maintain the cluster representative device for a certain period of time, wherein if at least one of battery level deterioration, signal deterioration, and no continuous response exceeds a standard, the cluster representative device is automatically replaced with a next-ranked candidate. Claim 4 A temperature tracking method for a mesh network-based cold chain system using clustering, wherein, in claim 1, the step of forming clusters and selecting cluster representative devices according to predefined rules comprises: merging two clusters when the link score between different cluster representative devices exceeds a merging criterion for a predetermined period, and dividing by designating a next-ranked representative candidate as the cluster representative device of a new cluster when the cluster diameter or load indicator exceeds a dividing criterion. Claim 5 A temperature tracking method for a mesh network-based cold chain system using clustering, wherein, in claim 1, the step of the plurality of packaging devices forming a cluster according to a predefined rule and electing a cluster representative device further comprises the step of setting a different join threshold and a different withdrawal threshold when switching between clusters of packaging devices and applying a minimum residence time. Claim 6 A temperature tracking method for a mesh network-based cold chain system using clustering, further comprising: a step in which, after the step of the plurality of packaging devices forming a cluster according to a predefined rule and electing a cluster representative device, when the location of a specific packaging device changes, each packaging device automatically joins the new cluster and re-elects the cluster representative device. Claim 7 A temperature tracking method for a mesh network-based cold chain system using clustering, wherein, in claim 1, the sensor data includes a packaging device identifier, a sequence number, and a timestamp. Claim 8 In claim 7, the method further comprises: a step in which, when communication between the driver terminal and the relay device is interrupted, a plurality of packaging devices record sensor data in chronological order in a storage unit; a step in which, when communication is resumed, the driver terminal identifies and requests missing data after the last received point using a sequence number and a timestamp; a step in which a plurality of packaging devices retransmit the missing data in chronological order according to the request; and a step in which the driver terminal sorts the received data in chronological order according to the sequence number and timestamp and removes duplicate data; a temperature tracking method for a mesh network-based cold chain system using clustering.

Images