Community intelligent express management method and system

Abstract

This invention relates to the field of intelligent express delivery management technology, and discloses a community intelligent express delivery management method and system. The method includes: acquiring historical travel behavior data, express delivery pickup behavior data, and address information of community residents; constructing a multi-express locker deployment spatial model; outputting a candidate set of delivery lockers; determining the target locker number; pushing inbound status data to residents; and using a long short-term memory network for temperature prediction and adjusting the temperature based on package temperature control requirements. Compared to existing express locker allocation methods that are mostly based on static addresses or fixed paths, especially in communities with multiple express lockers and diverse resident travel habits, this invention addresses the technical problem of difficulty in achieving efficient locker scheduling and environmental adaptation by using behavioral prediction modeling and a multi-target locker matching strategy. This achieves better express delivery decisions, thereby avoiding problems such as misdelivery, delayed delivery, or improper temperature control, and improving delivery accuracy and package security.

Description

Technical Field

[0001] This invention relates to the field of intelligent express delivery management technology, and in particular to a community intelligent express delivery management method and system. Background Technology

[0002] Currently, with the rapid development of last-mile logistics systems, smart parcel lockers are a key facility for improving delivery efficiency and alleviating manpower pressure. However, most mainstream parcel locker systems currently employ a "point-to-point distribution, static matching" deployment mechanism. Their scheduling strategies mainly rely on fixed address binding or courier experience judgment, making it difficult to accurately adapt to the personalized needs and dynamic pickup behaviors of different residents. This leads to problems such as high misdelivery rates, low delivery efficiency, and poor pickup experience. For example, in actual community scenarios, the following situations often occur: multiple parcel lockers are distributed at different entrances and exits, but the parcel system fails to intelligently select lockers based on residents' daily travel routes; different residents have significantly different pickup frequencies and time period preferences, and the system cannot dynamically predict residents' active times, resulting in mistimed delivery and delayed pickup; the usage rate of parcel lockers in different buildings within the same community is severely uneven, with some being full for extended periods and others idle and wasted. Furthermore, with the increasing demand for cold chain delivery and fresh food parcels, for special parcels requiring temperature or humidity control, most existing parcel locker systems only have passive temperature control functions and lack a predictive proactive adjustment mechanism. Therefore, there is an urgent need for an intelligent express delivery management method that integrates resident travel pattern modeling, express delivery behavior preference analysis, package attribute identification, spatial collaborative scheduling, and temperature control prediction and adjustment. This method should be able to achieve high responsiveness and low error rate, as well as intelligent temperature control throughout the entire express delivery management process, even under conditions of multi-source information perception and complex environments. This would improve the intelligence level, operational efficiency, and user satisfaction of the express delivery service system in the community environment. Summary of the Invention

[0003] To address the aforementioned technical shortcomings, the purpose of this invention is to propose a community-based intelligent express delivery management method. This method aims to solve the technical problem that existing express delivery locker allocation methods, which are mostly based on static addresses or fixed routes, are difficult to implement in terms of efficient locker scheduling and environmental adaptation, especially given the presence of multiple express delivery lockers in a community and the diverse travel habits of residents.

[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a community intelligent express delivery management method.

[0005] The community-based intelligent express delivery management method includes:

[0006] Step S10: Obtain historical travel behavior data, express delivery pickup behavior data, and community residents' address information of community residents; pre-build and train a multi-express delivery locker deployment spatial model; use historical travel behavior data, express delivery pickup behavior data, and community residents' address information as input to the multi-express delivery locker deployment spatial model; and output a candidate set of delivery lockers.

[0007] Step S20: Automatically collect basic attribute data and extended demand data of the incoming parcels through QR code recognition or RFID chip recognition, and determine the target cabinet number from the candidate set of delivery cabinets based on the basic attribute data and extended demand data;

[0008] Step S30: Send an opening control command to the target locker number and monitor the opening and closing status of the target locker number in real time. When the opening and closing status of the target locker number is detected to be closed, send the package entry status data to community residents in real time via email or mobile APP.

[0009] Step S40: Collect ambient temperature dataset in real time, and calculate the temperature error term by combining the ambient temperature dataset with the long short-term memory network method;

[0010] Step S50: Perform PID control of the temperature control system in the community smart cabinet based on the temperature error term and extended requirement data.

[0011] Preferably, in step S10, the historical travel behavior data includes residents' average daily travel time periods, residents' travel frequency, coordinates of residents' frequently used entrances and exits, time density distribution of residents entering and exiting the community, and characteristics of residents' transportation modes; the express delivery pickup behavior data includes residents' pickup frequency, residents' peak pickup time periods, residents' preferred pickup locker locations, residents' pickup delay time, and abnormal pickup behavior; the residents' express delivery address information includes the residents' building number, geographical relative location to the express delivery locker, and accessible route data.

[0012] Preferably, in step S10, the structure of the multi-delivery locker deployment space model specifically includes:

[0013] The feature fusion and input layer is used to receive and fuse historical travel behavior data, express delivery pickup behavior data and community residents' address information, and uniformly encode them to form a high-dimensional behavioral feature vector; among them, travel behavior and express delivery pickup behavior are encoded through time-series normalization, and address information is topologically connected to the cabinet layout through spatial coordinate mapping.

[0014] The spatial awareness coding layer is used to extract spatial accessibility feature vectors and walking path accessibility feature vectors between community building nodes and express delivery locker nodes from high-dimensional behavioral feature vectors.

[0015] The behavior-temporal aggregation layer is used to extract high-frequency active time window feature vectors from high-dimensional behavior feature vectors;

[0016] The cabinet preference matching layer is used to extract the resident preference feature vector from the high-dimensional behavioral feature vector;

[0017] The feature vector fusion layer is used to fuse spatial accessibility feature vectors, walking path accessibility feature vectors, high-frequency active time window feature vectors, and resident preference feature vectors to obtain an intermediate fused feature vector;

[0018] The candidate set output layer for delivery lockers is used to calculate the multi-target delivery score based on the intermediate fused feature vector using a multi-factor optimization method, and outputs the candidate set of delivery lockers based on the multi-target delivery score.

[0019] Preferably, in step S30, before sending the opening control command to the target cabinet number, the method further includes:

[0020] A preset facial database template is used to collect static facial images of users, and the collected static facial images are compared with the preset facial database template for the first layer of identity verification.

[0021] Real-time acquisition of dynamic facial images of users, combined with liveness detection method for second-level identity verification;

[0022] Once the first and second authentications are successful, an open control command is sent to the target cabinet number.

[0023] Preferably, in step S20, the basic attribute data includes package size information, package weight data, package barcode, RFID chip information, package delivery address information, and package type information; the extended requirement data includes temperature control requirements, fragility requirements, and special safety requirements.

[0024] Preferably, step S20, which involves determining the target cabinet number from the candidate set of delivery cabinets based on basic attribute data and extended demand data, specifically includes:

[0025] The package size and weight data are obtained from the basic attribute data. Based on the package size and weight data, the locker size is matched with the candidate locker set to obtain the first target locker number set.

[0026] The fragility requirement and special security requirement are obtained from the extended requirement data. Based on the fragility requirement and special security requirement, the safety of the container is matched to obtain the second target container number set.

[0027] Temperature control requirements are obtained from the expanded demand data. The temperature of cabinets in the second target cabinet number set is collected by temperature sensors. Temperature matching is performed based on the temperature control requirements and the cabinet temperature to obtain the target cabinet number.

[0028] Preferably, step S40, which involves real-time acquisition of an ambient temperature dataset and the calculation of the temperature error term based on the ambient temperature dataset using a long short-term memory network method, specifically includes:

[0029] A temperature long short-term memory prediction network is pre-set, and historical environmental temperature datasets are obtained. The temperature long short-term memory prediction network is pre-trained using the historical environmental temperature datasets.

[0030] The system collects ambient temperature datasets and current temperature data in real time, uses the ambient temperature datasets as input to a pre-trained temperature long short-term memory prediction network, and outputs predicted temperatures.

[0031] The temperature error term is calculated based on the predicted temperature and the current temperature data.

[0032] In the pre-training process of the Long Short-Term Memory (LSTM) prediction network, a mean squared error (MSE) term and a first-order derivative difference term are introduced as loss functions for training. The MSE term is used to measure the static deviation between the predicted value and the true value, while the first-order derivative difference term is used to suppress abrupt changes in the predicted value.

[0033] The present invention also provides a community intelligent express delivery management system, comprising:

[0034] The resident behavior modeling module is used to acquire historical travel behavior data, express delivery pickup behavior data, and community residents' address information of community residents. It pre-builds and trains a multi-express delivery locker deployment spatial model, and takes historical travel behavior data, express delivery pickup behavior data, and community residents' address information as input to the multi-express delivery locker deployment spatial model, and outputs a candidate set of delivery lockers.

[0035] The package information identification and cabinet matching module is used to automatically collect basic attribute data and extended demand data of inbound packages through QR code identification or RFID chip identification, and determine the target cabinet number from the candidate set of delivery cabinets based on the basic attribute data and extended demand data.

[0036] The delivery control and status feedback module is used to send open control commands to the target locker number and monitor the opening and closing status of the target locker number in real time. When the opening and closing status of the target locker number is detected to be closed, the module sends the package entry status data to community residents in real time via email or mobile APP.

[0037] The temperature prediction module is used to collect ambient temperature datasets in real time, and calculate the temperature error term based on the ambient temperature datasets using the long short-term memory network method.

[0038] The temperature control PID adjustment module is used to perform PID control of the temperature control system in the community smart cabinet based on the temperature error term and extended requirement data.

[0039] The present invention also provides a computer program product, including a community intelligent express delivery management program, which, when executed by a processor, implements the community intelligent express delivery management method.

[0040] The beneficial effects of this invention are as follows: Compared with the existing technology, which is mostly based on static addresses or fixed paths for parcel locker allocation, especially in communities with multiple parcel lockers and diverse residents' travel habits, it is difficult to achieve efficient locker scheduling and environmental adaptation. This application achieves better parcel delivery decisions through behavior prediction modeling and multi-objective locker matching strategies, thereby avoiding problems such as misdelivery, delayed delivery, or improper temperature control, and improving delivery accuracy and package security. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating the first embodiment of a community intelligent express delivery management method according to the present invention.

[0043] Figure 2 This is a schematic diagram of a community intelligent express delivery management method according to the present invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] Example 1: As Figure 1 The diagram shown is a flowchart of the first embodiment of the community intelligent express delivery management method of the present invention, which presents the first embodiment of the community intelligent express delivery management method of the present invention.

[0046] In the first embodiment, the community smart express delivery management method includes:

[0047] Step S10: Obtain historical travel behavior data, express delivery pickup behavior data, and community residents' address information of community residents; pre-build and train a multi-express delivery locker deployment spatial model; use historical travel behavior data, express delivery pickup behavior data, and community residents' address information as input to the multi-express delivery locker deployment spatial model; and output a candidate set of delivery lockers.

[0048] It should be noted that, in step S10, the specific structure of the multi-delivery locker deployment space model includes:

[0049] The feature fusion and input layer is used to receive and fuse historical travel behavior data, express delivery pickup behavior data and community residents' address information, and uniformly encode them to form a high-dimensional behavioral feature vector; among them, travel behavior and express delivery pickup behavior are encoded through time-series normalization, and address information is topologically connected to the cabinet layout through spatial coordinate mapping.

[0050] The spatial awareness coding layer is used to extract spatial accessibility feature vectors and walking path accessibility feature vectors between community building nodes and express delivery locker nodes from high-dimensional behavioral feature vectors.

[0051] It should be noted that the high-dimensional behavioral feature vector is a unified numerical vector generated by uniformly structuring and encoding the historical travel behavior data, package pickup behavior data, and community resident address information related to package pickup of an individual community resident. For example:

[0052] ;

[0053] in, It is a high-dimensional behavioral feature vector. Rate the frequency of package pickups over a 24-hour period. To score the historical accessibility of k parcel lockers, Residents score their historical preferences for k cabinet compartments. Let these be the geometric coordinates of the building. A binary code is used to represent special package pickup needs. The 24-hour package pickup frequency vector, historical accessibility scores with k parcel lockers, residents' historical preference scores for the k lockers, building coordinates, and the auxiliary binary code representing special pickup needs are all obtained directly or indirectly (using mathematical statistical analysis methods) from community residents' historical travel behavior data, parcel pickup behavior data, and community residents' address information.

[0054] Understandably, the historical accessibility score describes the accessibility characteristics of a resident from their building location to various parcel lockers during actual parcel pickups, primarily considering their travel route, walking time, and parcel pickup success rate. This historical accessibility score is obtained by extracting recorded resident path trajectories or entry / exit times from historical travel behavior data and parcel pickup behavior data, combined with locker location data, to analyze whether the resident has frequently used a particular route to reach a specific parcel locker. If a resident has successfully reached a locker multiple times historically using the same route, and the walking path is short and the time spent is low, then the accessibility score for that locker is higher.

[0055] For example, there is an accessible pedestrian passage between Resident A's Building 5 in the East Zone and the East Gate locker, and more than 70% of A's package pickup records in the past month came from this locker. The system counts the average travel time and number of times the path is recorded, generates the corresponding accessibility score, and encodes it into its high-dimensional features.

[0056] Historical preference scores reflect a resident's subjective preference for a specific parcel locker. They are automatically derived from historical travel and parcel pickup data, including pickup frequency, locker selection rate, and user feedback. Specifically, the data is obtained by statistically analyzing the percentage of times a resident picked up parcels from each locker over a given period (e.g., 30 days), and combining this with resident feedback records after each pickup (e.g., ratings, complaints, whether the pickup was timely), and then normalizing the data to obtain a preference score for each locker. For example, if resident B historically prefers locker B in the atrium, had 60% of their parcels delivered to that locker in the past month, and received an average "satisfied" rating across all pickups, then the historical preference score for that locker will be higher than for other lockers.

[0057] It should be noted that the spatial accessibility feature vector and the walking path accessibility feature vector are mainly obtained from the geometric coordinates of the buildings and the historical accessibility scores in the high-dimensional behavioral feature vector. Specifically, the spatial accessibility feature vector is mainly obtained by calculating the straight-line distance between the geometric coordinates of the building and the smart parcel locker, while the walking path accessibility feature vector is obtained by normalizing the historical accessibility scores.

[0058] The behavior-temporal aggregation layer is used to extract high-frequency active time window feature vectors from high-dimensional behavior feature vectors;

[0059] It should be noted that the high-frequency active time window feature vector is mainly obtained by using the periodic heat sliding window aggregation method based on the pickup activity frequency score of 24 hours a day in the high-dimensional behavioral feature vector. Specifically, it includes using a 24-dimensional time vector of 30 consecutive days. The windows were divided into 12 sliding windows, each lasting 2 hours. The average activity frequency in each window was calculated to obtain the pickup activity score. The system extracts the time period with the highest score and assesses its volatility by combining it with the standard deviation score of the most recent week, forming a time window stability feature. The output is a high-frequency active time window encoding vector, where each element represents the resident's tendency to pick up the package during that time period. The higher the value, the more likely the resident is to pick up the package during that time period.

[0060] The cabinet preference matching layer is used to extract the resident preference feature vector from the high-dimensional behavioral feature vector;

[0061] It should be noted that the resident preference feature vector is obtained by using a behavioral backtracking scoring method based on residents' historical preference scores for k cabinets from the high-dimensional behavioral feature vector. The calculation of the historical preference scores specifically includes:

[0062] Establish a historical interaction matrix between residents and service units If residents In the cabinet Pick up the item, then Otherwise, it is 0; an interaction quality factor is introduced. Each interaction is evaluated based on its time delay, whether an appeal is filed, and the level of satisfaction feedback, which are considered as interaction quality factors. Assign values ​​and calculate residents' information according to the following formula. For the Historical preference rating for each cabinet compartment: ,in, Rate your historical preferences. For interactive historical time points, This is a stabilization regularization term. Among them, the interaction quality factor... The following methods can be used to determine the following: "whether the package was picked up in time (e.g., pick-up within 24 hours is recorded as 1, overtime is recorded as 0.5 or 0), whether the locker was complained about (complaints are recorded as negative weights such as -1), and the package condition / resident satisfaction score (values ​​are recorded as 0 to 5)". The principle of taking positive values ​​is to take positive values, and the range is generally [0, 1] or [-1, 1]. The higher the value, the higher the resident's satisfaction with the locker. The specific values ​​can be obtained by weighted average or discrete scoring. To stabilize the regularization term, a constant value, such as 1 or 2, is used to avoid the occurrence of certain variables. When the value is very small, the score is too extreme or unstable, and the value range is usually [0.5, 5]. A set of interactive historical time points used to represent residents. Historically, in the cabinet section The number of times items have been retrieved from each locker can be obtained directly from system logs, access control card swipes, and locker barcode scanning records.

[0063] This formula is a typical weighted average and regularization strategy, belonging to the experience-based scoring method in behavioral modeling. The numerator is the sum of the quality of all interactions between residents and the cabinet, reflecting the "intensity" of preference, and the denominator is the number of interactions plus the regularization term, reflecting the "credibility" of preference. The overall result is a stable, interpretable, and updatable personalized scoring index used to indicate whether residents tend to use a certain cabinet and the quality of their user experience.

[0064] The feature vector fusion layer is used to fuse the spatial accessibility feature vector, the walking path accessibility feature vector, the high-frequency active time window feature vector, and the resident preference feature vector to obtain an intermediate fused feature vector; the candidate set output layer for delivery lockers is used to calculate the multi-objective delivery score based on the intermediate fused feature vector using a multi-factor optimization method, and output the candidate set of delivery lockers based on the multi-objective delivery score.

[0065] It should be noted that the intermediate fusion feature vector uses a multi-channel attention mechanism to weight different types of input features. Each input feature is first unified in dimensionality through a fully connected layer, and then the attention mechanism is used to calculate the weight of each feature. Finally, all features are weighted and fused to obtain the intermediate fusion feature vector. Each component feature vector in the intermediate fusion feature vector is automatically matched with the relevant information of the parcel locker to generate a multi-target delivery score. A multi-target delivery score threshold is preset, and the set of parcel locker numbers whose multi-target delivery scores are greater than the multi-target delivery score threshold is output as the candidate set of delivery lockers.

[0066] Understandably, by deeply integrating behavioral characteristics, spatial layout, and time preferences, this spatial model not only possesses static geographic computing capabilities but can also dynamically adapt to changes in users' individual behaviors. This enables a multi-factor integrated grid recommendation mechanism that combines behavior-driven, geographic perception, and time synchronization, significantly outperforming traditional single-target matching methods that rely on address or distance.

[0067] It should be understood that traditional systems typically only place parcels into a single locker based on the resident's address or the courier's proximity, making it difficult to fully adapt to the diverse behaviors of residents within a community and the current usage situation of multiple lockers coexisting. In contrast, the multi-locker deployment spatial model designed in this invention has the following technical advantages: it can automatically match high-hit-rate lockers based on the user's actual usage path and daily active time; it can dynamically select multiple most suitable alternative lockers in a community with multiple locker deployments, improving the hit rate and load distribution capacity; and it supports subsequent algorithm modules for highly available delivery path planning and scheduling decisions.

[0068] For example, in a simulation test in a large residential community, resident A lives in the east unit of building 5. They typically leave home at 8 AM and return at 9 PM, with their historical package pickup times concentrated around 8 PM on Fridays, indicating a preference for the east gate parcel lockers. By constructing a multi-locker deployment spatial model, the system identified the lockers with the shortest spatial distance to the east gate lockers, smoothest access paths, high overlap in active times, and higher preference scores. In a single package delivery, the model prioritized the east gate lockers as candidate lockers. When the east gate lockers were full, the model recommended the north gate lockers, with their similar paths, as a secondary option. This resulted in an actual test hit rate increase of approximately 31.4% and a reduction in the average delivery path by 23.2%.

[0069] Step S20: Automatically collect basic attribute data and extended demand data of the incoming parcels through QR code recognition or RFID chip recognition, and determine the target cabinet number from the candidate set of delivery cabinets based on the basic attribute data and extended demand data;

[0070] It should be noted that step S20, which involves determining the target cabinet number from the candidate set of delivery cabinets based on basic attribute data and extended requirement data, specifically includes: obtaining package size information and package weight data from the basic attribute data; matching the cabinet size in the candidate set of delivery cabinets based on the package size information and package weight data to obtain a first target cabinet number set; obtaining fragileity requirements and special safety requirements from the extended requirement data; matching the cabinet safety based on the fragileity requirements and special safety requirements to obtain a second target cabinet number set; and obtaining temperature control requirements from the extended requirement data; collecting the cabinet temperature in the second target cabinet number set using a temperature sensor; and matching the temperature based on the temperature control requirements and the cabinet temperature to obtain the target cabinet number.

[0071] In step S30, by matching multi-dimensional data (size, weight, safety, temperature control, etc.), the most suitable cabinet compartment can be accurately selected for package storage. This ensures that packages receive appropriate storage space based on their size and weight, and also provides additional protection for expanded needs (such as temperature control and safety), avoiding risks during package storage. Furthermore, through real-time temperature monitoring, the system can ensure that packages with special needs (such as temperature-controlled storage) receive the optimal storage environment.

[0072] For example, suppose there is a package measuring 40×30×20 cm, weighing 3 kg, and marked as fragile, requiring cold chain storage. First, by matching size and weight, suitable cabinet compartments are selected, resulting in a first set of target cabinet compartment numbers. Next, based on the package's fragility requirements, cabinets with reinforced designs are selected, resulting in a second set of target cabinet compartment numbers. Finally, the temperature of the second target cabinet compartment is detected using a temperature sensor. Assuming the current compartment temperature is 4°C, and the package's temperature control requirement is 2°C to 8°C, the compartment is considered to meet the temperature control requirements, and this compartment is ultimately selected as the target cabinet compartment number, ensuring the package's safe storage.

[0073] Step S30: Send an opening control command to the target locker number and monitor the opening and closing status of the target locker number in real time. When the opening and closing status of the target locker number is detected to be closed, send the package entry status data to community residents in real time via email or mobile APP.

[0074] It should be understood that this method allows users to receive timely and accurate notifications of the package's storage status after it has been stored, improving user experience and system transparency. Instant notifications via email or the app ensure real-time feedback on the package's storage status, preventing inconvenience caused by delayed information.

[0075] For example, suppose a user's package is stored in locker 101 by a courier. A switch sensor detects that the locker's open / closed state changes within 3 seconds. At this point, inbound status data is generated containing the following information: Package ID: 12345; Locker Number: 101; Inbound Time: April 28, 2023, 14:30:00. Next, using the smptlib library, an email is sent to the user with the following content: Subject: Package Inbound Notification; Body: Your package (ID: 12345) has been successfully stored in locker 101 on April 28, 2023, at 14:30:00. You can pick it up at any time. Simultaneously, the same inbound status information is pushed to the user via the app, ensuring accurate and timely delivery of information.

[0076] Step S40: Collect ambient temperature dataset in real time, and calculate the temperature error term by combining the ambient temperature dataset with the long short-term memory network method;

[0077] It should be noted that step S40, which involves real-time acquisition of the ambient temperature dataset and temperature prediction calculation based on the ambient temperature dataset using a long short-term memory (LSM) network, specifically includes: pre-setting a temperature LSM prediction network; acquiring historical ambient temperature datasets; pre-training the temperature LSM prediction network using the historical ambient temperature datasets; real-time acquisition of the ambient temperature dataset and current temperature data; using the ambient temperature dataset as input to the pre-trained temperature LSM prediction network; and outputting the predicted temperature; calculating the temperature error term based on the predicted temperature and current temperature data. During the pre-training of the LSM prediction network, a mean squared error term and a first-order derivative difference term are introduced as loss functions for training. The mean squared error term measures the static deviation between the predicted value and the true value, while the first-order derivative difference term suppresses abrupt changes in the predicted value.

[0078] Understandably, by combining long short-term memory networks with environmental sensor data, it is possible to sense and predict temperature change trends inside community express delivery lockers, thereby identifying potential temperature control anomalies in advance and improving the foresight and accuracy of regulation. Unlike traditional control methods that rely solely on the current temperature value for regulation, this method introduces a temperature trend prediction mechanism, which can effectively avoid problems of "over-regulation" or "lagging regulation."

[0079] It should be understood that, compared to traditional single-point PID feedback control, this embodiment, by introducing a predictive drive mechanism and a robust optimized loss function, can effectively reduce control errors caused by sudden environmental changes (such as sunlight, outdoor high temperatures, and cabinet door opening), thus enhancing the temperature control system's adaptability and control accuracy in highly variable environments. Furthermore, by introducing a first-order derivative control term, the model can more smoothly fit the actual temperature change curve, avoiding oscillations or overcompensation at the "inflection point."

[0080] For example, in a simulation experiment, the system collected temperature data from a parcel locker over the past 48 hours as training samples. Under strong summer sunlight, the temperature inside the locker fluctuated drastically. Traditional MSE training resulted in frequent fluctuations and large errors in the output of the LSTM model. After introducing a first-order derivative difference term, the model's prediction curve became smoother and closer to the actual temperature change trend. In 24-hour rolling prediction, the average prediction error decreased from ±1.7°C to ±0.8°C, and the prediction response delay was reduced by approximately 34%. This temperature error term was used in subsequent PID control, further optimizing the response efficiency and energy consumption stability of the locker temperature control system.

[0081] Step S50: Perform PID control of the temperature control system in the community smart cabinet based on the temperature error term and extended requirement data.

[0082] Understandably, by introducing a PID control mechanism driven by a predictive temperature error term, the system can perform adjustments in advance based on predicted trends, avoiding the "passive response" problem of traditional systems based on current temperature, thereby improving adjustment response speed and energy efficiency. Furthermore, extending demand data to participate in control parameter selection enables temperature control behavior to possess "differentiated" and "precise" capabilities, meeting the requirements of diverse package storage environments.

[0083] It should be understood that, compared to the "constant temperature dead zone control" or "single PID adjustment" methods used in conventional temperature control cabinets, this embodiment, by integrating prediction errors and personalized packaging requirements, transforms PID control from a "single objective, fixed parameter" system into a dynamic parameter-adjusting closed-loop system under multi-objective conditions. Especially in actual deployments facing interference from factors such as large day-night temperature differences and frequent cabinet door openings, this solution can effectively reduce temperature control instability caused by adjustment delays or overreactions.

[0084] For example, in a community smart locker environment simulation experiment, a batch of fresh produce packages required the locker compartment temperature to be maintained between 2℃ and 8℃. Preset using a "steady-state PID parameter template," the predicted temperature under strong midday sunlight was 9.2℃, the current temperature was 8.3℃, and the system identification error was 0.9℃. The PID controller calculated the cooling power command, driving the cooling module to start in advance to prevent further temperature increases. The temperature difference remained stable within ±0.5℃ during the control process. Compared to traditional PID control without prediction, the temperature control response was approximately 11 minutes earlier, and the cold start frequency was reduced by approximately 27%, effectively improving temperature control accuracy and energy consumption control.

[0085] Example 2: Furthermore, the present invention provides a community intelligent express delivery management system, which employs a community intelligent express delivery management method as described in the above embodiments, and can solve a technical problem in community intelligent express delivery management. Compared with the prior art, the beneficial effects of the community intelligent express delivery management system provided by the present invention are the same as the beneficial effects of the community intelligent express delivery management method provided in the above embodiments, and other technical features in the community intelligent express delivery management system are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0086] Example 3: This invention provides a community intelligent express delivery management device. Please refer to... Figure 2A community smart express delivery management device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to perform a community smart express delivery management method as described in Embodiment 1 above. The community smart express delivery management device in this embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle terminals (e.g., vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. This community smart express delivery management device is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this invention. A community smart express delivery management device may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. RAM 1004 also stores various programs and data required for the operation of a community smart express delivery management device. Processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows a community smart express delivery management device to communicate wirelessly or wiredly with other devices to exchange data. Although a community smart express delivery management device with various systems is shown in the figure, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems may be implemented alternatively.

[0087] Example 4: This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the community intelligent express delivery management method described above. The computer program product provided by this invention can solve a technical problem in community intelligent express delivery management. Compared with the prior art, the beneficial effects of the computer program product provided by this invention are the same as the beneficial effects of the community intelligent express delivery management method provided in the above embodiments, and will not be repeated here.

[0088] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this invention.

[0089] It should be understood that the various parts disclosed in this invention can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0090] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A community-based intelligent express delivery management method, characterized in that, The methods include: Step S10: Obtain historical travel behavior data, parcel pickup behavior data, and community resident address information of community residents; pre-build and train a multi-parcel locker deployment spatial model; use the historical travel behavior data, parcel pickup behavior data, and community resident address information as input to the multi-parcel locker deployment spatial model; the output is a candidate set of delivery lockers; the specific structure of the multi-parcel locker deployment spatial model includes: The feature fusion and input layer is used to receive and fuse historical travel behavior data, express delivery pickup behavior data and community residents' address information, and uniformly encode them to form a high-dimensional behavioral feature vector; among them, travel behavior and express delivery pickup behavior are encoded through time-series normalization, and address information is topologically connected to the cabinet layout through spatial coordinate mapping. The spatial awareness coding layer is used to extract spatial accessibility feature vectors and walking path accessibility feature vectors between community building nodes and express delivery locker nodes from high-dimensional behavioral feature vectors. The behavior-temporal aggregation layer is used to extract high-frequency active time window feature vectors from high-dimensional behavior feature vectors; The cabinet preference matching layer is used to extract the resident preference feature vector from the high-dimensional behavioral feature vector; The feature vector fusion layer is used to fuse spatial accessibility feature vectors, walking path accessibility feature vectors, high-frequency active time window feature vectors, and resident preference feature vectors to obtain intermediate fused feature vectors. The candidate set output layer for delivery lockers is used to calculate the multi-target delivery score based on the intermediate fused feature vector using a multi-factor optimization method, and outputs the candidate set of delivery lockers based on the multi-target delivery score. Step S20: Automatically collect basic attribute data and extended demand data of the incoming parcels through QR code recognition or RFID chip recognition, and determine the target cabinet number from the candidate set of delivery cabinets based on the basic attribute data and extended demand data; Step S30: Send an opening control command to the target locker number and monitor the opening and closing status of the target locker number in real time. When the opening and closing status of the target locker number is detected to be closed, send the package entry status data to community residents in real time via email or mobile APP. Step S40: Collect ambient temperature dataset in real time, and calculate the temperature error term by combining the ambient temperature dataset with the long short-term memory network method. Step S50: Perform PID control of the temperature control system in the community smart cabinet based on the temperature error term and extended requirement data.

2. The community intelligent express delivery management method as described in claim 1, characterized in that, In step S10, historical travel behavior data includes residents' average daily travel time periods, residents' travel frequency, coordinates of residents' frequently used entrances and exits, time density distribution of residents entering and exiting the community, and characteristics of residents' transportation modes; express delivery pickup behavior data includes residents' pickup frequency, residents' peak pickup time periods, residents' preferred pickup locker locations, residents' pickup delay time and abnormal pickup behavior; residents' express delivery address information includes the resident's building number, geographical relative location to the express delivery locker, and accessible route data.

3. The community intelligent express delivery management method as described in claim 1, characterized in that, Before sending the opening control command to the target cabinet number in step S30, the following steps are also included: A preset facial database template is used to collect static facial images of users, and the collected static facial images are compared with the preset facial database template for the first layer of identity verification. Real-time acquisition of dynamic facial images of users, combined with liveness detection method for second-level identity verification; Once the first and second authentications are successful, an open control command is sent to the target cabinet number.

4. The community intelligent express delivery management method as described in claim 1, characterized in that, In step S20, the basic attribute data includes package size information, package weight data, package barcode, RFID chip information, package delivery address information, and package type information; the extended requirement data includes temperature control requirements, fragility requirements, and special safety requirements.

5. A community intelligent express delivery management method as described in claim 1, characterized in that, Step S20, which involves determining the target container number from the candidate set of delivery containers based on basic attribute data and extended demand data, specifically includes: The package size and weight data are obtained from the basic attribute data. Based on the package size and weight data, the locker size is matched with the candidate locker set to obtain the first target locker number set. The fragility requirement and special security requirement are obtained from the extended requirement data. Based on the fragility requirement and special security requirement, the safety of the container is matched to obtain the second target container number set. Temperature control requirements are obtained from the expanded demand data. The temperature of cabinets in the second target cabinet number set is collected by temperature sensors. Temperature matching is performed based on the temperature control requirements and the cabinet temperature to obtain the target cabinet number.

6. The community intelligent express delivery management method as described in claim 1, characterized in that, Step S40 involves real-time acquisition of the ambient temperature dataset, followed by the calculation of the temperature error term using a long short-term memory network method based on the ambient temperature dataset. This step specifically includes: A temperature long short-term memory prediction network is pre-set, and historical environmental temperature datasets are obtained. The temperature long short-term memory prediction network is pre-trained using the historical environmental temperature datasets. The system collects ambient temperature datasets and current temperature data in real time, uses the ambient temperature datasets as input to a pre-trained temperature long short-term memory prediction network, and outputs predicted temperatures. The temperature error term is calculated based on the predicted temperature and the current temperature data. In the pre-training process of the Long Short-Term Memory (LSTM) prediction network, a mean squared error (MSE) term and a first derivative difference term are introduced as loss functions for training. The MSE term is used to measure the static deviation between the predicted value and the true value, while the first derivative difference term is used to suppress abrupt changes in the predicted value.

7. A community intelligent express delivery management system, applied to the community intelligent express delivery management method according to any one of claims 1-6, characterized in that, The community intelligent express delivery management system includes: The resident behavior modeling module is used to acquire historical travel behavior data, parcel pickup behavior data, and resident address information of community residents. It pre-builds and trains a multi-parcel locker deployment spatial model, using the historical travel behavior data, parcel pickup behavior data, and resident address information as inputs to the model, and outputting a candidate set of delivery lockers. The specific structure of the multi-parcel locker deployment spatial model includes: The feature fusion and input layer is used to receive and fuse historical travel behavior data, express delivery pickup behavior data and community residents' address information, and uniformly encode them to form a high-dimensional behavioral feature vector; among them, travel behavior and express delivery pickup behavior are encoded through time-series normalization, and address information is topologically connected to the cabinet layout through spatial coordinate mapping. The spatial awareness coding layer is used to extract spatial accessibility feature vectors and walking path accessibility feature vectors between community building nodes and express delivery locker nodes from high-dimensional behavioral feature vectors. The behavior-temporal aggregation layer is used to extract high-frequency active time window feature vectors from high-dimensional behavior feature vectors; The cabinet preference matching layer is used to extract the resident preference feature vector from the high-dimensional behavioral feature vector; The feature vector fusion layer is used to fuse spatial accessibility feature vectors, walking path accessibility feature vectors, high-frequency active time window feature vectors, and resident preference feature vectors to obtain intermediate fused feature vectors. The candidate set output layer for delivery lockers is used to calculate the multi-target delivery score based on the intermediate fused feature vector using a multi-factor optimization method, and outputs the candidate set of delivery lockers based on the multi-target delivery score. The package information identification and cabinet matching module is used to automatically collect basic attribute data and extended demand data of inbound packages through QR code identification or RFID chip identification, and determine the target cabinet number from the candidate set of delivery cabinets based on the basic attribute data and extended demand data. The delivery control and status feedback module is used to send open control commands to the target locker number and monitor the opening and closing status of the target locker number in real time. When the opening and closing status of the target locker number is detected to be closed, the module sends the package entry status data to community residents in real time via email or mobile APP. The temperature prediction module is used to collect ambient temperature datasets in real time, and calculate the temperature error term based on the ambient temperature datasets using the long short-term memory network method. The temperature control PID adjustment module is used to perform PID control of the temperature control system in the community smart cabinet based on the temperature error term and extended requirement data.

8. A community intelligent express delivery management device, characterized in that, The community smart express delivery management device includes: a memory, a processor, and a community smart express delivery management program stored in the memory and executable on the processor. When the community smart express delivery management program is executed by the processor, it implements a community smart express delivery management method according to any one of claims 1 to 6.

9. A computer program product, characterized in that, The computer program product includes a community intelligent express delivery management program, which, when executed by a processor, implements a community intelligent express delivery management method according to any one of claims 1 to 6.

Images