Multi-Mode Interactive Garbage Classification Guidance and Recycling System

Abstract

The present invention discloses a multi-mode interactive garbage classification guidance and recycling system, which specifically relates to the field of garbage classification and recycling. It is used to solve the problem of mismatch between equipment capacity planning and the spatio-temporal distribution of user delivery behaviors. Through spatio-temporal joint modeling, combining user movement trajectories and equipment load data, it captures behavior migration characteristics and realizes global coordination of equipment full-load prediction and path optimization. Relying on multi-modal interaction data, it can perceive the user's location and garbage category in real time, fuse heterogeneous indices and path topology constraints, and dynamically generate a diversion decision that takes into account equipment capacity, regional load balance, and user path preferences. At the same time, through user behavior feedback, it continuously optimizes the load prediction and path screening rules, forming a closed-loop link of behavior recognition, load prediction, diversion strategy, and system iteration. It improves the accuracy of classification guidance, equipment load balance, and user delivery convenience, effectively alleviates equipment overload and resource waste, and improves the operation efficiency of urban garbage classification and recycling.

Description

Technical Field

[0001] The present invention relates to the field of garbage classification and recycling, and more specifically, to a multi-mode interactive garbage classification guidance and recycling system. Background Art

[0002] Currently, urban intelligent garbage classification recycling bins are mostly concentrated in fixed scenarios such as office areas and residential areas, but there is a deep gap between the equipment capacity planning and the spatio-temporal distribution of users' delivery behaviors. Taking a typical office area as an example, during the lunch break, a large number of users concentrate on discarding disposable lunch boxes, causing the recycling bins to be quickly filled to capacity, while similar equipment in adjacent commercial areas is in a low-load state for a long time due to sparse traffic; the peak of kitchen waste delivery at night in residential areas often causes equipment overflow, but before the early morning cleaning the next day, the equipment is idling, resulting in wasted resources. Although existing systems can predict the capacity of a single device based on historical data, they cannot capture the migration characteristics of users' behaviors in the spatio-temporal dimension, such as the dynamic deviation of commuting routes and the temporary delivery tides caused by regional activities, leading to inaccurate equipment load prediction and ineffective guidance strategies.

[0003] The essence of the problem lies in: the periodic volatility of users' delivery behaviors in time and the path dependence in space form a high-dimensional non-linear coupling relationship with the equipment capacity consumption rate. Traditional static scheduling models cannot simultaneously interpret two types of contradictory features: on the one hand, the strong spatio-temporal inertia of users' delivery paths causes local sudden accumulations of equipment loads; on the other hand, global load balancing requires cross-regional dynamic diversion. However, limited by users' natural resistance to route detours, the system needs to construct a spatio-temporal joint optimization model under the constraints of users' behaviors. A deeper contradiction is that there is a modeling conflict between the time cycle law in historical data and the spatial load mutation triggered by real-time events, resulting in difficulties in coordinating the prediction of equipment full-load time and the diversion path planning, and ultimately forming a negative cycle in which users' behaviors exacerbate equipment overload - overloaded equipment reversely distorts users' behaviors.

[0004] To solve the above problems, a technical solution is provided now. Summary of the Invention

[0005] To overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a multi-mode interactive garbage classification guidance and recycling system. A spatio-temporal coupled closed-loop guidance mechanism is constructed to optimize the garbage classification path and resource scheduling in view of the contradiction between the spatio-temporal distribution of user delivery behaviors and the imbalance of equipment loads. Based on the spatio-temporal modeling of user trajectories and equipment loads, the capacity consumption characteristics of scenarios such as peak hours in office areas and tidal tides in residential areas are quantified; multi-modal data is used to real-time sense the location and garbage categories, and a diversion decision that takes into account capacity, load differences and user preferences is generated; the model is continuously optimized through behavior feedback to form a closed-loop of behavior recognition - load prediction - diversion generation - adaptive iteration. This method breaks through the limitations of traditional static scheduling, reduces the risks of overload and overflow while avoiding detours, reduces resource vacancies, and realizes multi-objective optimization of guidance accuracy, load balance and delivery convenience to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A multi-mode interactive garbage classification guidance and recycling system, comprising:

[0008] A trajectory clustering module, a load prediction module, a path optimization module and a feedback adjustment module;

[0009] The trajectory clustering module identifies high-frequency delivery paths through spatio-temporal clustering of user movement trajectories, and extracts dynamic prediction parameters for equipment full-load time by modeling the spatio-temporal correlation of equipment load data;

[0010] The load prediction module quantifies the maximum detour range based on the user's current location and path preference, and uses the spatio-temporal prediction results of the equipment capacity consumption rate to generate an index reflecting the matching degree of load and path through dynamic calculation of the equipment load urgency and path selection resistance, and dynamically screens out a set of candidate equipment that meets the remaining capacity requirements;

[0011] The path optimization module takes the user's detour distance as a hard constraint condition, takes the remaining capacity of the equipment and the regional load difference as optimization objectives, and solves the optimal delivery path through a multi-objective decision-making algorithm;

[0012] The feedback adjustment module adjusts the prediction parameters of the equipment load in the spatio-temporal association model based on the user behavior feedback data, synchronously optimizes the path screening threshold, and realizes the closed-loop optimization of prediction and scheduling.

[0013] In a preferred embodiment, the trajectory clustering module includes the following:

[0014] Identify high-frequency delivery paths through spatio-temporal clustering of legal mobile trajectories actively provided by users, specifically including collecting a set of trajectory points uploaded by users, performing clustering processing on the set of trajectory points in the spatio-temporal dimension using a density-based spatio-temporal clustering algorithm to generate multiple spatio-temporal clusters, and screening out high-frequency delivery paths by statistically analyzing the occurrence frequency of trajectory points within each spatio-temporal cluster within a specific time period; at the same time, combining spatio-temporal correlation modeling of device load data, specifically including constructing a vector autoregressive model to analyze the dynamic changes of device load in time and space, calculating the device load change rate, and then extracting dynamic prediction parameters for the device full-load time.

[0015] In a preferred embodiment, the dynamic prediction parameters include the capacity consumption rate and the predicted full-load time, and the predicted full-load time is calculated by the ratio of the remaining capacity to the capacity consumption rate.

[0016] In a preferred embodiment, the load prediction module includes the following:

[0017] Obtain the user's current location in real time and extract the preference feature vector by analyzing the user's historical delivery behavior data, calculate the maximum detour distance threshold acceptable to the user to determine the detour range; then obtain the dynamic prediction parameters of the device from the trajectory clustering module and calculate the coupling coefficient using the device capacity consumption rate; finally, dynamically generate a candidate device set through multiple screenings including the user's maximum detour range, the device remaining capacity being greater than the remaining capacity threshold, and the coupling coefficient being greater than or equal to the preset coupling coefficient threshold.

[0018] In a preferred embodiment, the coupling coefficient is obtained by adjusting the ratio of the spatio-temporal inertia decay index to the path topology resistance index and the regional device load difference degree.

[0019] In a preferred embodiment, the spatio-temporal inertia decay index is calculated by weighting the instantaneous change gradient of the device capacity consumption rate and the information entropy of the user's historical delivery time distribution by the reciprocal of the regional population flow density.

[0020] In a preferred embodiment, the path topology resistance index is obtained by multiplying the path network topology depth difference from the user's current location to the device location by the decay memory factor of the user's historical detour behavior.

[0021] In a preferred embodiment, the path optimization module includes the following:

[0022] Taking the user's detour distance as a hard constraint condition, and the remaining capacity of the device and the regional load difference as optimization objectives, first calculate the geographical distance from the user's current location to the candidate device location to verify whether it meets the hard constraint of the user's detour distance; then define the remaining capacity of the device as the full-load capacity of the device minus the current load, and at the same time define the regional load difference as the sum of the deviation indices of all device loads from the average load; then use an improved non-dominated sorting method to construct an initial solution set, evaluate each candidate device for its remaining capacity and the regional load difference after delivery, and generate a Pareto front solution set through non-dominated sorting; subsequently, introduce a target preference factor to calculate the comprehensive score of the solutions in the Pareto front solution set, and select the device with the highest comprehensive score as the optimal device; finally, use the Dijkstra algorithm to generate the shortest path from the user's current location to the optimal device location, so as to solve the optimal delivery path.

[0023] In a preferred embodiment, the feedback adjustment module includes the following:

[0024] First, collect multiple pieces of feedback data on the actual delivery of the user, compare them with the recommended results generated by the path optimization module, and analyze to obtain the device selection deviation, delivery time deviation, and delivery volume deviation; then, by comparing the difference between the actual capacity consumption rate and the predicted capacity consumption rate, use a non-linear fusion method to update the capacity consumption rate of the device, and recalculate the full-load time of the device according to the updated capacity consumption rate; at the same time, adjust the maximum detour distance threshold, remaining capacity threshold, and coupling coefficient threshold respectively according to the user's actual detour distance, the remaining capacity of the device after delivery, and the coupling coefficient of the actual path; finally, feedback the updated predicted parameters and path screening thresholds to the trajectory clustering module and the load prediction module.

[0025] The technical effects and advantages of the multi-mode interactive garbage classification guidance and recycling system of the present invention:

[0026] By constructing a closed-loop optimized multi-mode interactive garbage classification guidance and recycling system, it effectively addresses the challenge of the mismatch between equipment capacity planning and the spatio-temporal distribution of users' delivery behaviors in urban intelligent garbage classification and recycling; relying on the spatio-temporal joint modeling of users' mobile trajectories and equipment load data, it accurately captures the characteristics of behavior migration, and realizes the global coordination of equipment full-load prediction and path optimization; with the help of multi-modal interaction data to sense the user's location and garbage category in real time, and through the dynamic calculation of the fusion of heterogeneous indices and path topology constraints, it generates a diversion decision that takes into account equipment capacity, regional load balance and users' path preferences; at the same time, through the continuous optimization of the load prediction and path screening rules by the user behavior feedback, it forms a closed-loop link of behavior recognition, load prediction, diversion strategy and system iteration, thus significantly improving the accuracy of classification guidance, the balance of equipment load and the convenience of users' delivery, effectively alleviating the problems of equipment overload and resource waste, and comprehensively improving the operation efficiency and user experience of urban garbage classification and recycling. Brief Description of the Drawings

[0027] Figure 1 It is a schematic structural diagram of the multi-mode interactive garbage classification guidance and recycling system of the present invention.

[0028] Figure 2 It is a schematic diagram of the sub-step process of the path optimization module of the multi-mode interactive garbage classification guidance and recycling system of the present invention. Detailed Embodiments

[0029] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0030] Embodiment 1: Figure 1 The multi-mode interactive garbage classification guidance and recycling system of the present invention is given, including: a trajectory clustering module, a load prediction module, a path optimization module and a feedback adjustment module.

[0031] The trajectory clustering module identifies high-frequency delivery paths through the spatio-temporal clustering of users' mobile trajectories, and combines the spatio-temporal correlation modeling of equipment load data to extract dynamic prediction parameters for the equipment full-load time.

[0032] The load prediction module quantifies the maximum detour range based on the user's current location and path preference, and uses the spatio-temporal prediction results of the equipment capacity consumption rate to generate an index reflecting the matching degree of load and path through the dynamic calculation of the equipment load urgency and path selection resistance, and dynamically screens out a set of candidate equipment that meets the remaining capacity requirements.

[0033] The path optimization module takes the user's detour distance as a hard constraint condition, aims at the remaining capacity of the device and the regional load difference, and solves the optimal delivery path through a multi-objective decision-making algorithm.

[0034] The feedback adjustment module adjusts the prediction parameters of the device load in the spatio-temporal correlation model based on the user behavior feedback data, synchronously optimizes the path screening threshold, and realizes the closed-loop optimization of prediction and scheduling.

[0035] In the context of urban intelligent waste sorting and recycling, the problem of unbalanced device load significantly restricts the resource utilization efficiency and the user's delivery experience. There are significant differences in the spatio-temporal distribution of delivery behaviors between office areas and residential areas. For example, the centralized delivery of disposable lunch boxes in the office area at noon causes the device to be fully loaded, while the peak of kitchen waste in the residential area at night is prone to overflow. This spatio-temporal unevenness makes it difficult for traditional capacity prediction based on the historical data of a single device to adapt to the dynamic migration characteristics of user behavior. Therefore, the present invention proposes to construct an accurate prediction mechanism for the device full-load time through the spatio-temporal coupling analysis of user movement trajectories and device load data, so as to optimize the waste sorting path and resource scheduling.

[0036] The trajectory clustering module includes the following:

[0037] S1.1, spatio-temporal clustering of user movement trajectories:

[0038] In the spatio-temporal clustering of user movement trajectories, first, collect the legal movement trajectory data actively uploaded by users. These data are recorded as a set of trajectory points. Each trajectory point in the set contains the position information of the user at a specific time, namely the timestamp and the geographical coordinates (longitude and latitude). Next, use a density-based spatio-temporal clustering algorithm to process these trajectory points. Specifically, an extended version of the DBSCAN algorithm is used, which jointly considers the time and space dimensions, that is, analyzes the three dimensions of timestamp, longitude, and latitude simultaneously. During the clustering process, set a neighborhood distance threshold and a minimum number of points threshold. The neighborhood distance threshold is dynamically adjusted according to the density of user delivery behaviors, and the minimum number of points threshold is set as the minimum number of trajectory points that must be included in each cluster. Through the clustering algorithm, the trajectory points are divided into several spatio-temporal clusters, and each spatio-temporal cluster represents a group of trajectory points that are close to each other in time and space. Subsequently, count each spatio-temporal cluster, calculate the number of occurrences of trajectory points within a specific time period (such as every hour), and define it as the occurrence frequency. Compare the occurrence frequency with a preset frequency threshold, and filter out the spatio-temporal clusters with an occurrence frequency higher than the threshold. From these filtered spatio-temporal clusters, further extract the user's high-frequency delivery paths, which reflect the user's delivery habits within a specific time period.

[0039] S1.2, spatio-temporal correlation modeling of device load data:

[0040] In the spatio-temporal correlation modeling of device load data, first, the load data of the device are collected, which include the device ID, time stamp, and the current load of the device at that time stamp. Then, a vector autoregressive model is constructed to analyze the dynamic relationship of the device load in time and space. Specifically, a load vector is defined, which represents the set of loads of all devices at a specific moment. In this model, the change of the load vector is affected by both time factors and space factors. The time factor is represented by introducing time lag terms, that is, considering the influence of the loads at several past moments on the current load. The space factor is represented by constructing a spatial adjacency matrix, which describes the spatial proximity relationship between devices. The coefficients in the model are estimated by the least squares method. Specifically, the load vector is fitted with the weighted combination of its time lag terms and the spatial adjacency matrix to calculate the value of each coefficient, so as to capture the change law of the device load in time and space. Based on the prediction results of the model, the load change rate of each device is further calculated, which is defined as the increase in the load per unit time, reflecting the growth trend of the device load.

[0041] S1.3, Extraction of dynamic prediction parameters for device full-load time:

[0042] In the extraction of dynamic prediction parameters for device full-load time, first, based on the load change rate obtained from the aforementioned vector autoregressive model, the capacity consumption rate of each device is defined, which represents the increase in the device load per unit time. Then, the full-load time of each device is predicted. Assuming that the full-load capacity of the device is a known value and the current load is the load at the current moment. The calculation process of the full-load time is as follows: First, calculate the difference between the current load and the full-load capacity, that is, the remaining capacity is obtained by subtracting the current load from the full-load capacity; then divide the remaining capacity by the capacity consumption rate to get the time required to increase from the current load to the full-load capacity; finally, add this time to the current time to obtain the specific time point when the device is expected to reach full load. If the capacity consumption rate is greater than zero, indicating that the load is increasing, the full-load time is calculated according to the above method; if the capacity consumption rate is less than or equal to zero, indicating that there is no increasing trend in the load, it is considered that the device will not be full-load in the short term, and the full-load time is set to a very large value, indicating that the full-load situation will not occur. Finally, dynamic prediction parameters are generated for each device, which include the capacity consumption rate and the predicted full-load time, fully recording the dynamic change characteristics of the device load.

[0043] The trajectory clustering module accurately identifies the user's high-frequency delivery paths through spatio-temporal clustering of the user's movement trajectories, providing a behavioral basis for path optimization; and calculates the device capacity consumption rate and full-load time through spatio-temporal correlation modeling of device load data, forming dynamic prediction parameters. These parameters accurately reflect the spatio-temporal dynamic characteristics of the device load, providing reliable data support for the generation of the candidate device set in the load prediction module and ensuring that subsequent scheduling can adapt to the real-time changes of user behavior and device status.

[0044] In the trajectory clustering module, dynamic prediction parameters of the device full-load time are extracted through spatio-temporal clustering of the user's movement trajectories and spatio-temporal correlation modeling of device load data. These parameters depict the dynamic change characteristics of the device load, providing basic data support for the load prediction module. The goal of the load prediction module is to dynamically generate a candidate device set that meets the remaining capacity requirements based on the user's current location and path preference, combined with the spatio-temporal prediction results of the trajectory clustering module, laying a foundation for solving the optimal delivery path in the subsequent path optimization module, and at the same time achieving user delivery path optimization and device load balancing.

[0045] The load prediction module includes the following:

[0046] S2.1, Analysis and modeling of the user's current location and path preference:

[0047] In the analysis and modeling of the user's current location and path preference, first, the user's current location is obtained through real-time positioning technology, which is represented by longitude and latitude in the geographic coordinate system. Then, using the user's historical delivery behavior data, analyze the types of paths that the user tends to choose, such as the shortest path, the fastest path, or the habitual path, and extract a preference feature vector by statistically analyzing the characteristics of the user's historical delivery paths. The dimension of this vector is determined according to the specific scenario, for example, including elements such as distance, time, or detour frequency. Then, combining the user's current location and the preference feature vector, calculate the maximum detour distance that the user can accept. The specific calculation method is as follows: first calculate the average value of the user's historical detour distance; then calculate the modulus length of the preference feature vector, that is, the square root of the sum of the squares of each dimension value; then divide the average value of the historical detour distance by the natural logarithm of the modulus length of the preference feature vector plus 1 to obtain the maximum detour distance. Finally, with the user's current location as the center and the maximum detour distance as the radius, delimit the detour range that the user can accept.

[0048] S2.2, Application of spatio-temporal prediction results of the device capacity consumption rate:

[0049] In the application of the spatio-temporal prediction results of the device capacity consumption rate, first obtain the dynamic prediction parameters of the device from the trajectory clustering module, including the capacity consumption rate of the device and the predicted full-load time. Then, introduce a coupling coefficient to measure the matching degree between the load state of the device and the user path selection. This coefficient is calculated by combining the spatio-temporal inertia decay index and the path topology resistance index.

[0050] Spatio-temporal inertia decay index: This index combines the instantaneous change gradient of the device capacity consumption rate and the information entropy of the user's historical delivery time distribution, and is weighted by the reciprocal of the regional pedestrian flow density. The specific calculation method is as follows: First, calculate the rate of change of the capacity consumption rate with respect to time, that is, the time derivative; then calculate the information entropy of the user's historical delivery time distribution to measure the regularity of the time distribution; then multiply the time derivative by the information entropy and divide by the reciprocal of the regional pedestrian flow density to obtain the spatio-temporal inertia decay index. The spatio-temporal inertia decay index quantifies the urgency of the device load in the time and space dimensions, reflecting the load growth trend of the device at a specific moment and location and its priority for meeting demands. The larger the value, the higher the spatio-temporal urgency of the device load, indicating that the device is currently facing strong load growth pressure and needs to prioritize delivery guidance or resource scheduling to alleviate the situation.

[0051] Path topology resistance index: This index is based on the difference in the topological depth of the path network from the user's current position to the device position, that is, the shortest path length in the path network, and superimposes the decay memory factor of the user's historical detour behavior. The specific calculation method is as follows: First, determine the topological depth difference; then calculate the decay memory factor, which is obtained by taking the exponential function of the product of a negative decay coefficient and the time interval of the historical detour behavior; finally, multiply the topological depth difference by the result of this exponential function to obtain the path topology resistance index. The path topology resistance index measures the potential resistance when the user selects a certain path, reflecting the deviation degree between the path selection and the user's habits. The larger the value, the greater the resistance for the user to select this path, that is, the higher the degree of deviation of the path from the user's habitual trajectory, resulting in a lower possibility for the user to select this path in actual delivery.

[0052] Coupling Coefficient: This coefficient is based on the spatio-temporal inertia decay exponent and is adjusted by the arctangent function of the path topology resistance exponent and the regional device load difference degree. The specific calculation method is as follows: First, calculate the ratio of the spatio-temporal inertia decay exponent to the path topology resistance exponent, and map this ratio to between 0 and 1 through the hyperbolic tangent function; then calculate the arctangent function of the path topology resistance exponent, and use the regional device load difference degree (i.e., the entropy value of the remaining capacity of the devices in the region) as the power for adjustment; finally, multiply the result of the hyperbolic tangent function by the adjustment factor to obtain the coupling coefficient. The coupling coefficient measures the matching degree between the device load state and the user path selection. This coefficient comprehensively considers the urgency of the device load and the accessibility of the user path, reflecting the coordination between the two. The larger the value, the higher the matching degree between the device load state and the user path selection, which means that the device is both within the acceptable path range of the user and has a suitable load state for delivery, and is an ideal choice for user delivery.

[0053] S2.3, Dynamic Generation of the Candidate Device Set:

[0054] In the dynamic generation of the candidate device set, first set the remaining capacity constraint of the device, that is, after subtracting the current load from the full load capacity of the device, the remaining capacity should be greater than or equal to the preset remaining capacity threshold. Then, based on the user's maximum detour range, remaining capacity requirement, and coupling coefficient, screen the candidate device set. The screening conditions include the following three aspects:

[0055] Geographical Constraint: Calculate the geographical distance between the device location and the user's current location, which is determined by the spherical distance calculation method and is required to be less than or equal to the maximum detour distance threshold acceptable to the user.

[0056] Capacity Constraint: Ensure that the remaining capacity of the device is greater than or equal to the preset remaining capacity threshold to ensure that there is still available capacity when the user arrives.

[0057] Coupling Constraint: Require that the coupling coefficient of the device is greater than or equal to the preset coupling coefficient threshold, such as 0.5, to ensure that the matching degree between the device and the user path reaches the standard.

[0058] Finally, through comprehensively considering the above geographical constraint, capacity constraint, and coupling constraint, dynamically generate the candidate device set, which contains all devices that meet the conditions.

[0059] The load prediction module accurately delimits the maximum detour range by analyzing the user's current location and path preference feature vectors. Based on the dynamic prediction parameters of the device capacity consumption rate and the predicted full load time provided by the trajectory clustering module, combined with the dynamic calculation of the device load urgency and path selection resistance, it generates a coupling coefficient reflecting the matching degree between the load and the path. Finally, through multiple screenings of geographical distance, remaining capacity, and matching degree indicators, it dynamically forms a candidate device set that meets the remaining capacity requirements, providing a high-quality device selection basis for the multi-objective path optimization of the path optimization module, ensuring the efficiency of the user's delivery path and the balance of the device load.

[0060] In the load prediction module, the maximum detour range is delimited by the user's current location and path preference feature vectors, and a candidate device set is generated using the spatio-temporal prediction results of the device capacity consumption rate. Each device in this set meets the constraint conditions of geographical distance, remaining capacity, and coupling coefficient. These results provide an alternative device set for the path optimization module, aiming to solve the collaborative problem of user delivery path optimization and device load balance. The path optimization module takes the user's detour distance as a hard constraint, and the remaining capacity of the device and the regional load difference as optimization objectives, and solves the optimal delivery path through a multi-objective decision-making algorithm to ensure the balance between user convenience and device resource utilization efficiency.

[0061] As Figure 2 shown, the path optimization module includes the following:

[0062] S3.1, User detour distance:

[0063] The user detour distance refers to the geographical distance between the user's current location and the candidate device location, which is determined by the spherical distance calculation method. The calculation process first requires obtaining the longitude and latitude coordinates of the user's current location and the longitude and latitude coordinates of the candidate device location. Then, the Haversine formula is used to calculate the spherical distance between the two points. This method takes into account the curvature of the earth and converts the longitude and latitude coordinates into a straight-line distance in kilometers or meters to obtain the distance value between the user and the candidate device. Finally, the hard constraint requires that this distance must be less than or equal to the maximum detour distance threshold that the user can accept to ensure that the user will not be inconvenienced due to excessive path deviation when going for delivery.

[0064] S3.2, Device remaining capacity optimization:

[0065] The optimization of the remaining capacity of the device aims to select devices with a larger remaining capacity to ensure that there is still a relatively large amount of available space on the device after delivery. The calculation process first determines the full-load capacity of each device, that is, the maximum load that the device can accommodate. Then, the current load of each device is obtained, and the full-load capacity is subtracted from the current load to get the remaining capacity of the device. The optimization goal is to preferentially select devices with a larger remaining capacity, so that the available space on the device will not be depleted too quickly after the user's delivery, thereby improving the utilization efficiency of the device.

[0066] S3.3, Regional Load Difference Optimization:

[0067] The regional load difference optimization is achieved by measuring the distribution uniformity of the loads of all devices in the region, and the goal is to minimize the dispersion degree of the load distribution after the user's delivery. The calculation process first calculates the total load of all devices in the region, divides it by the number of devices to get the average load of the devices in the region. Then, for each device, the ratio of its load to the average load is calculated, and then 1 is subtracted from this ratio, and the absolute value is taken to get the deviation degree of the device. To amplify the influence of the deviation effect, the absolute value is raised to a power greater than 1, such as the 2nd power, to generate the deviation index of the device. Subsequently, the deviation indices of all devices are added up to get the load dispersion of the region. The optimization goal is to make the change of the regional load dispersion as small as possible by selecting appropriate devices after the user's delivery.

[0068] S3.4, Construction of the Initial Solution Set of the Multi-Objective Decision Algorithm:

[0069] The multi-objective decision algorithm is based on an improved non-dominated sorting method and first constructs an initial solution set. The initial solution set is composed of a set of candidate devices, and each candidate device corresponds to a candidate path from the user's current location to the location of the device. The set of candidate devices is part of the input parameters and includes all devices that meet the hard constraint of the user's detour distance, serving as the basis for the subsequent evaluation of the optimization goal.

[0070] S3.5, Target Evaluation Process of the Multi-Objective Decision Algorithm:

[0071] The target evaluation calculates the values of two optimization targets for each candidate device separately. First, for the remaining capacity of the device, the result is directly obtained by subtracting the current load from the full-load capacity of the device. Second, for the regional load dispersion, after simulating that the user adds the delivery volume to this candidate device, the load of the device is updated, and then the average load of all devices in the region is recalculated. For each device, calculate the ratio of its updated load to the new average load, subtract 1 and take the absolute value, and then take a power greater than 1, such as the second power, to obtain the deviation index. Add up the deviation indexes of all devices to get the regional load dispersion of this candidate device after the user's delivery. Through this process, two target values, namely the remaining capacity of the device and the regional load dispersion after delivery, are generated for each candidate device.

[0072] S3.6, the non-dominated sorting process of the multi-objective decision-making algorithm:

[0073] Non-dominated sorting is used to screen out the optimal solution set from candidate devices to form the Pareto front. The sorting process compares all candidate devices and evaluates them based on two objectives: the remaining capacity of the device and the regional load dispersion after delivery. For any two devices, if the remaining capacity of one device is greater than that of the other device, and its regional load dispersion after delivery is less than or equal to that of the other device, then the former is determined to dominate the latter. By comparing one by one, all candidate devices are divided into multiple levels, and the non-dominated devices form the Pareto front solution set, which represents the solutions that cannot be surpassed by other devices simultaneously in the two objectives.

[0074] S3.7, the calculation of the solution set and the comprehensive score of the multi-objective decision-making algorithm:

[0075] To select the unique optimal device from the Pareto front solution set, a target preference factor is introduced and the comprehensive score is calculated. The target preference factor is a value between 0 and 1, which is used to balance the importance of the remaining capacity of the device and the regional load dispersion. When calculating the comprehensive score, first normalize the remaining capacity of the device. Divide the remaining capacity of each device by the maximum full-load capacity of all devices to get the normalized value, and then multiply it by the target preference factor. Second, normalize the regional load dispersion after delivery. First calculate the minimum and maximum values of the regional load dispersion after delivery of all candidate devices. Subtract the difference between the current device's load dispersion and the minimum value from the maximum value, divide by the difference between the maximum value and the minimum value, and then subtract the result from 1 to get the normalized value, and then multiply it by 1 minus the target preference factor. Finally, add the normalized result of the remaining capacity of the device to the normalized result of the regional load dispersion after delivery to get the comprehensive score of this device. Select the device with the highest comprehensive score as the optimal device.

[0076] S3.8, the generation of the optimal path of the multi-objective decision-making algorithm:

[0077] After determining the optimal device, calculate the shortest path from the user's current location to the location of the device as the optimal delivery path. The Dijkstra algorithm is used in the calculation process. First, with the user's current location as the starting point and the location of the optimal device as the end point, a path network is constructed based on the geographical distance. The algorithm starts from the starting point, gradually explores all possible paths, records the cumulative distance to each node, and always selects the node with the smallest current cumulative distance for expansion until the end point is reached. Finally, the shortest path from the user's current location to the location of the optimal device is generated as the recommended route for user delivery.

[0078] The set of candidate devices is first screened by the hard constraint of the user's detour distance to ensure that the geographical distances of all devices meet the requirements. Subsequently, based on these parameters, the remaining capacity of the devices and the load dispersion in the area after delivery are calculated, and a multi-objective decision-making algorithm is executed. The output result is the optimal device and its corresponding shortest path, which are used for the update of prediction parameters and the adjustment of thresholds in subsequent steps.

[0079] The processing technical logic of the path optimization module uses the user's detour distance as a hard constraint. By optimizing two objectives, namely the remaining capacity of the device and the load dispersion in the area, an improved non-dominated sorting method is used to determine the optimal delivery path. The calculation process includes the screening of candidate devices, the evaluation of objective values, non-dominated sorting, the refinement of comprehensive scores, and the generation of the shortest path, ensuring the efficiency of user delivery and the balance of device loads.

[0080] The feedback adjustment module includes the following:

[0081] S4.1, Collect and process user behavior feedback data:

[0082] In the process of collecting and processing user behavior feedback data, first, relevant information about the user's actual delivery needs to be recorded, including the device selected by the user, the time of delivery, the number of items delivered, and the path actually selected by the user. These actual data are then compared with the recommended device, recommended delivery time, and recommended delivery volume in the path optimization module to calculate three types of deviation data. Specifically, the device selection deviation is determined by judging whether the device actually delivered by the user is the same as the recommended device. If they are the same, the deviation is zero; if not, the deviation is a non-zero value. The time deviation is obtained by calculating the difference between the actual delivery time of the user and the recommended delivery time, indicating the time difference between the two. The delivery volume deviation is determined by calculating the difference between the actual delivery volume of the user and the recommended delivery volume, reflecting the gap between the actual delivery volume and the predicted value. These deviation data will be used as the basis for subsequent parameter adjustment.

[0083] S4.2, Adjust the device load prediction parameters:

[0084] When adjusting the prediction parameters of the device load, first calculate the actual capacity consumption rate. The method is to divide the actual delivery volume of the user by the delivery time interval to obtain the capacity consumption of the device per unit time. Then, compare this actual capacity consumption rate with the predicted capacity consumption rate in the trajectory clustering module and update the predicted value using a non-linear fusion method. The specific update process is as follows: Take the predicted capacity consumption rate as the base value and then add an adjustment term; this adjustment term is determined by calculating the ratio of the difference between the actual capacity consumption rate and the predicted capacity consumption rate, and a regulation factor is introduced to control the magnitude of the adjustment. The updated capacity consumption rate is then used to recalculate the full-load time of the device. The calculation method is: starting from the current time, add the quotient obtained by dividing the remaining capacity of the device by the updated capacity consumption rate, and the result is the time point when the device is expected to be fully loaded.

[0085] S4.3, Adjust the maximum detour distance threshold:

[0086] When adjusting the maximum detour distance threshold in the path screening threshold, first check whether the device actually delivered by the user is the same as the recommended device and whether the actual detour distance exceeds the current maximum detour distance threshold. If both conditions are met, the maximum detour distance threshold needs to be adjusted. The specific adjustment method is: Take the current maximum detour distance threshold as the base value and add an adjustment term; this adjustment term is determined by calculating the ratio of the difference between the actual detour distance and the current maximum detour distance threshold, and a regulation factor is introduced to control the magnitude of the adjustment. Through this process, the maximum detour distance threshold is increased to adapt to the possible greater detour requirements in the actual behavior of the user. The path screening threshold includes the maximum detour distance threshold, the remaining capacity threshold, and the coupling coefficient threshold.

[0087] S4.4, Adjust the remaining capacity threshold:

[0088] For the adjustment of the remaining capacity threshold, first check whether the remaining capacity of the device after the user's delivery is lower than the current remaining capacity threshold. If it is lower than the current threshold, the remaining capacity threshold needs to be adjusted. The specific adjustment method is: Take the current remaining capacity threshold as the base value and add an adjustment term; this adjustment term is determined by calculating the ratio of the difference between the actual remaining capacity of the device after delivery and the current remaining capacity threshold, and a regulation factor is introduced to control the magnitude of the adjustment. Through this process, the remaining capacity threshold is increased to ensure that the recommended device in the future still has sufficient remaining capacity to meet the requirements after delivery.

[0089] S4.5, Adjust the coupling coefficient threshold:

[0090] For the adjustment of the coupling coefficient threshold, first check whether the device actually delivered by the user is different from the recommended device, and whether the coupling coefficient of the actual path is lower than the current coupling coefficient threshold. If these two conditions are met, the coupling coefficient threshold needs to be adjusted. The specific adjustment method is as follows: taking the current coupling coefficient threshold as the base value, subtract an adjustment term; this adjustment term is determined by calculating the ratio of the difference between the actual coupling coefficient and the current coupling coefficient threshold, and a regulation factor is introduced to control the magnitude of the adjustment. Through this process, the coupling coefficient threshold is reduced, thereby relaxing the path screening conditions and enabling more potential paths to be included in the recommended range.

[0091] S4.6, Closed-loop optimization mechanism:

[0092] In the closed-loop optimization mechanism, the adjusted device load prediction parameter, that is, the updated capacity consumption rate, is fed back to the trajectory clustering module for the prediction calculation of the next device full-load time. At the same time, the adjusted path screening thresholds, including the maximum detour distance threshold, the remaining capacity threshold, and the coupling coefficient threshold, are fed back to the load prediction module for the subsequent screening process of the candidate device set. By continuously collecting user behavior feedback data and repeatedly performing the above parameter and threshold adjustment processes, the prediction parameters and screening thresholds gradually approach the true distribution of user behavior characteristics and the actual state of the device, thereby achieving the closed-loop optimization effect of prediction and scheduling.

[0093] The feedback adjustment module analyzes the user behavior feedback data, calculates the device selection deviation, time deviation, and delivery volume deviation, and then adjusts the device load prediction parameters and path screening thresholds, including the maximum detour distance threshold, the remaining capacity threshold, and the coupling coefficient threshold, and feeds the adjustment results back to the previous steps to form a closed-loop optimization mechanism, ultimately improving the accuracy of device full-load time prediction and the adaptability of path screening.

[0094] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0095] It should be noted that the system of the present invention can be deployed on the device itself to achieve embedded applications, or can also run on a PC or other terminals with a user interface, so as to meet various hardware environments and usage requirements.

[0096] Only some exemplary embodiments of the present invention are described above by way of illustration. Undoubtedly, for those of ordinary skill in the art, without departing from the spirit and scope of the present invention, the described embodiments can be modified in various different ways. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

[0097] It should be noted that in this text, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0098] As described above, the above is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A multi-mode interactive garbage classification guidance and recycling system, characterized in that, Including: A trajectory clustering module, a load prediction module, a path optimization module, and a feedback adjustment module; The trajectory clustering module identifies high-frequency delivery paths through spatio-temporal clustering of user movement trajectories, and extracts dynamic prediction parameters of the device full-load time by modeling the spatio-temporal correlation of device load data; The load prediction module quantifies the maximum detour range based on the user's current location and path preference, utilizes the spatio-temporal prediction results of the device capacity consumption rate, and generates an index reflecting the matching degree between the load and the path, that is, the coupling coefficient, through the dynamic calculation of the device load urgency and path selection resistance, so as to dynamically screen out a set of candidate devices that meet the remaining capacity requirements; The path optimization module takes the user's detour distance as a hard constraint condition, takes the device remaining capacity and regional load difference as optimization objectives, and solves the optimal delivery path through a multi-objective decision-making algorithm; The feedback adjustment module adjusts the prediction parameters of the device load in the spatio-temporal correlation model based on the user behavior feedback data, synchronously optimizes the path screening threshold, and realizes the closed-loop optimization of prediction and scheduling; Among them, the path screening threshold includes a maximum detour distance threshold, a remaining capacity threshold, and a coupling coefficient threshold.

2. The multi-mode interactive garbage classification guidance and recycling system according to claim 1, wherein The trajectory clustering module includes the following: Identifying high-frequency delivery paths through spatio-temporal clustering of legal movement trajectories actively provided by users, specifically including collecting a set of trajectory points uploaded by users, performing clustering processing on the set of trajectory points in the spatio-temporal dimension using a density-based spatio-temporal clustering algorithm to generate multiple spatio-temporal clusters, and screening out high-frequency delivery paths by statistically counting the occurrence frequency of trajectory points within each spatio-temporal cluster within a specific time period; at the same time, modeling the spatio-temporal correlation of device load data, specifically including constructing a vector autoregressive model to analyze the dynamic changes of device load in time and space, calculating the device load change rate, and then extracting dynamic prediction parameters of the device full-load time.

3. The multi-mode interactive garbage classification guidance and recycling system according to claim 2, characterized in that: Among them, the dynamic prediction parameters include the capacity consumption rate and the predicted full-load time, and the predicted full-load time is calculated by the ratio of the remaining capacity to the capacity consumption rate.

4. The multi-mode interactive garbage classification guidance and recycling system according to claim 2, wherein The load prediction module includes the following: Obtaining the user's current position in real time and extracting the preference feature vector by analyzing the user's historical delivery behavior data, calculating the maximum detour distance threshold acceptable to the user to determine the detour range; then obtaining the dynamic prediction parameters of the device from the trajectory clustering module, and calculating the coupling coefficient using the device capacity consumption rate; finally, through multiple screenings of the user's maximum detour range, the device remaining capacity being greater than the remaining capacity threshold, and the coupling coefficient being greater than or equal to the preset coupling coefficient threshold, a set of candidate devices is dynamically generated.

5. The multi-mode interactive garbage classification guidance and recycling system according to claim 4, characterized in that: Among them, the coupling coefficient is obtained through the ratio of the spatio-temporal inertia decay index to the path topology resistance index and the adjustment of the regional device load difference degree; The spatio-temporal inertia decay index is calculated by weighting the instantaneous change gradient of the device capacity consumption rate and the information entropy of the user's historical delivery time distribution by the reciprocal of the regional pedestrian flow density; The path topological resistance index is obtained by multiplying the difference in the topological depth of the path network from the user's current location to the device location by the decay memory factor of the user's historical detour behavior.

6. The multi-mode interactive garbage classification guidance and recycling system according to claim 4, characterized in that, The path optimization module includes the following: Taking the user's detour distance as a hard constraint and the remaining capacity of the device and the regional load difference as optimization objectives, first calculate the geographical distance from the user's current location to the candidate device location to verify whether it meets the hard constraint of the user's detour distance; then define the remaining capacity of the device as the full-load capacity of the device minus the current load, and at the same time define the regional load difference as the sum of the deviation indices of all device loads from the average load; then use an improved non-dominated sorting method to construct the initial solution set, evaluate each candidate device for its remaining capacity and the regional load difference after delivery one by one, and generate the Pareto front solution set through non-dominated sorting; Subsequently, introduce the objective preference factor to calculate the comprehensive score for the solutions in the Pareto front solution set, and select the device with the highest comprehensive score as the optimal device; finally, use the Dijkstra algorithm to generate the shortest path from the user's current location to the optimal device location, so as to solve the optimal delivery path.

7. The multi-mode interactive garbage classification guidance and recycling system according to claim 6, characterized in that, The feedback adjustment module includes the following: First, collect multiple pieces of feedback data on the user's actual delivery and compare them with the recommended results generated by the path optimization module to analyze and obtain the device selection deviation, delivery time deviation, and delivery volume deviation; then, by comparing the difference between the actual capacity consumption rate and the predicted capacity consumption rate, use a non-linear fusion method to update the capacity consumption rate of the device, and recalculate the full-load time of the device according to the updated capacity consumption rate; at the same time, adjust the maximum detour distance threshold, remaining capacity threshold, and coupling coefficient threshold respectively according to the user's actual detour distance, the remaining capacity of the device after delivery, and the coupling coefficient of the actual path; finally, feedback the updated prediction parameters and path screening thresholds to the trajectory clustering module and the load prediction module.

Images