Urban operation fleet collaborative operation method and system based on cloud platform

Abstract

The invention relates to the technical field of cloud platforms, and discloses a cloud platform-based urban operation fleet collaborative operation method and system, and the method comprises the steps: transmitting data collected by a garbage collection point intelligent monitoring device and a garbage vehicle-mounted terminal to a cloud platform in real time, and obtaining garbage management comprehensive data; performing dynamic time-varying analysis on the garbage accumulation condition of each garbage collection point to obtain garbage collection point overflow prediction time and garbage type distribution proportion data; performing digital mapping on execution resources of the urban garbage fleet to obtain collaborative task partition data; according to the method, garbage disposal recursive optimization and scheduling analysis are carried out based on cooperative task partition data, and a target scheduling instruction set is obtained, garbage can capacity and vehicle load constraints can be met, effective control over the garbage overflow risk is achieved, multi-target Pareto optimal scheduling is further achieved, and the method is suitable for large-scale popularization and application. And meanwhile, the clearing efficiency, energy conservation and service quality are met.

Description

Technical Field

[0001] The present invention relates to the field of cloud platform technology, and in particular to a method and system for collaborative operation of urban operation fleets based on a cloud platform. Background Art

[0002] Traditional waste disposal systems typically rely on fixed routes and frequencies, lacking the ability to respond to dynamic changes in waste generation. This leads to overflowing trash cans and wasted resources. Furthermore, information silos exist between different types of vehicles, such as collection and transfer vehicles. Resource scheduling at key nodes like transfer stations and treatment plants is disconnected from the front-end collection process, resulting in overall system inefficiency. This is particularly true during peak waste generation periods, where the system's response capacity is insufficient, making it difficult to meet the demands of modern urban smart sanitation initiatives.

[0003] Existing waste transfer and removal systems lack effective state models, making it difficult to establish accurate waste generation prediction mechanisms. This is particularly true when it comes to optimizing the categorized transfer of different types of waste, such as recyclables, food waste, and other types of garbage. Furthermore, under conditions of traffic uncertainty, existing systems struggle to achieve robust coordination among multiple types of waste vehicles, and are unable to effectively address the uncertainties in waste generation and initial collection times, resulting in inefficient waste treatment. Summary of the Invention

[0004] The present invention provides a cloud-based urban fleet collaborative operation method and system. The present invention can effectively control the risk of garbage overflow while meeting the constraints of garbage bin capacity and vehicle load, thereby achieving multi-objective Pareto optimal scheduling while meeting the requirements of transportation efficiency, energy conservation and service quality.

[0005] In a first aspect, the present invention provides a method for collaborative operation of an urban operation fleet based on a cloud platform, the method comprising: Transmit data collected by intelligent monitoring devices at garbage collection points and terminals on garbage trucks to the cloud platform in real time to obtain comprehensive garbage management data; Performing a dynamic time-varying analysis of garbage accumulation at each garbage collection point based on the comprehensive garbage management data to obtain garbage collection point overflow prediction time and garbage type distribution ratio data; Based on the predicted overflow time of the garbage collection point, the garbage type distribution ratio data, and the real-time road traffic status data, the execution resources of the urban garbage fleet are digitally mapped to obtain collaborative task partition data; Based on the collaborative task partition data, garbage disposal recursive optimization and scheduling analysis are performed to obtain a target scheduling instruction set.

[0006] In a second aspect, the present invention provides a cloud-based urban operation fleet collaborative operation system, the cloud-based urban operation fleet collaborative operation system comprising: Real-time transmission module, used to transmit data collected by intelligent monitoring devices at garbage collection points and garbage truck-mounted terminals to the cloud platform in real time to obtain comprehensive garbage management data; A dynamic time-varying analysis module is used to perform a dynamic time-varying analysis on the garbage accumulation situation at each garbage collection point based on the comprehensive garbage management data, and obtain the overflow prediction time of the garbage collection point and the garbage type distribution ratio data; A digital mapping module is used to digitally map the execution resources of the urban garbage fleet based on the predicted overflow time of the garbage collection point, the garbage type distribution ratio data, and real-time road traffic status data to obtain collaborative task partition data; The scheduling analysis module is used to perform garbage disposal recursive optimization and scheduling analysis based on the collaborative task partition data to obtain a target scheduling instruction set.

[0007] The technical solution provided by this invention achieves real-time collection and transmission of multi-dimensional data, including garbage weight, type, and capacity, as well as vehicle location, load, and driving status, by installing intelligent monitoring devices at garbage collection points and onboard terminals on vehicles. This eliminates the data silos in traditional systems. Dynamic, time-varying analysis of garbage status based on a Markov state model enables the system to accurately predict the overflow time and garbage type distribution at each garbage collection point, thereby shifting from passive response to proactive prevention and effectively avoiding garbage bin overflows. By digitally mapping the execution resources of the urban garbage fleet and combining garbage overflow prediction times with real-time road traffic conditions, the system generates more rational collaborative task partitions and achieves precise allocation of garbage vehicle resources. Regional decomposition and task splitting calculations of task partition data significantly improve system processing speed. The introduction of a garbage overflow risk function for risk assessment and the use of an adaptive cooperative game control mechanism to address system constraints enable the system to effectively control garbage overflow risks while meeting garbage bin capacity and vehicle load constraints. By constructing a comprehensive evaluation function that takes into account the balance of total transportation time, total energy consumption, garbage overflow risk and vehicle utilization, the system can achieve multi-objective Pareto optimal scheduling while meeting multiple requirements such as transportation efficiency, energy conservation and service quality.

[0008] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the present invention. The purposes and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0009] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A schematic diagram of an embodiment of a method for collaborative operation of an urban operation fleet based on a cloud platform in an embodiment of the present invention; Figure 2 This is a schematic diagram of an embodiment of a cloud platform-based urban operation fleet collaborative operation system in an embodiment of the present invention. DETAILED DESCRIPTION

[0011] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0012] The terms "including," "having," and any variations thereof, as used in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to the process, method, product, or device.

[0013] To facilitate understanding of this embodiment, a cloud platform-based urban operation fleet collaborative operation method disclosed in an embodiment of the present invention is first introduced in detail. Figure 1 As shown, this method includes the following steps: 101. Transmit data collected by intelligent monitoring devices at garbage collection points and terminals on garbage trucks to the cloud platform in real time to obtain comprehensive garbage management data; It is understandable that the execution subject of the present invention can be a cloud-based urban operation fleet collaborative operation system, or a terminal or server, which is not limited here. The embodiment of the present invention is described by taking the server as the execution subject as an example.

[0014] Specifically, the raw garbage data collected by intelligent monitoring devices at garbage collection points undergoes outlier filtering and edge computing preprocessing. Using multiple detection algorithms, basic sensor-generated data, such as garbage weight, garbage type, and capacity utilization, is verified item by item, eliminating obvious outliers caused by equipment failure, environmental interference, or communication anomalies. The edge computing module also performs preliminary data compression and format standardization locally to filter out redundant and noisy data, effectively reducing subsequent transmission pressure and improving data structuring. Furthermore, a fusion algorithm is used to comprehensively process and transform the GPS and Beidou dual-system positioning signals collected by the garbage truck terminal, eliminating drift and blind spots associated with a single positioning signal and improving the real-time and accuracy of vehicle location data. Regarding the vehicle status monitoring module, the raw vehicle data collected by the garbage truck terminal is processed and formatted in real time to obtain load and driving status data, including load weight, speed, fuel consumption, and driving trajectory. The pre-processed garbage weight data, garbage type data, and capacity occupancy data, as well as the vehicle location data obtained after fusion positioning, formatted load data, and driving status data, are encapsulated according to unified standard fields to generate the original data packet. In terms of data security, to prevent data theft, tampering, or leakage during transmission, a layered encryption mechanism is used to perform multi-level security encryption on the original data packet. Advanced symmetric encryption algorithms such as AES-256 are used at the bottom layer to encrypt the data content. The transport layer enhances overall security through secure hash verification and dynamic key management. Integrity verification and retransmission mechanisms are introduced at the network protocol layer to ensure the confidentiality and integrity of the data throughout the transmission process. The data stream after layered encryption is transmitted in real time to the cloud platform via high-speed networks such as 5G. Upon receiving the layered encrypted data stream, the cloud platform automatically performs data decryption, integrity verification, and format verification, integrating waste management data from different sources and types into unified comprehensive waste management data.

[0015] 102. Conduct dynamic time-varying analysis of garbage accumulation at each garbage collection point based on comprehensive garbage management data to obtain overflow prediction time and garbage type distribution ratio data for garbage collection points; Specifically, a dynamic, time-varying analysis of garbage accumulation at each collection point is conducted using high-frequency, comprehensive garbage management data. This process aims to enhance the foresight and intelligence of garbage collection and transportation. Time series data on weight, type, and capacity occupancy at each collection point are extracted. By converting the continuously collected data stream into ordered time series state data, the dynamic process of garbage generation, accumulation, and classification is reflected. Based on this time series data, a garbage collection point state matrix is established, containing multi-dimensional parameters such as current garbage weight, current capacity occupancy percentage, garbage growth rate, and garbage type distribution ratio. This matrix describes the real-time operating status and evolution trends of each collection point at different time points. The garbage collection point state matrix is segmented into patterns, identifying typical variation patterns such as weekdays, weekends, seasonal fluctuations, and morning and evening rush hours. Based on this, a Markov process prediction algorithm is used to probabilistically model the state changes at each collection point, resulting in a prediction of the initial garbage state, reflecting the trend and rate of change of garbage accumulation over time. Initial predictions are revised and calculated based on environmental parameters, real-time traffic conditions, and historical anomaly data. By introducing external interference factors and adaptive parameter tuning mechanisms, the robustness and practicality of the prediction model are enhanced, generating a target waste status prediction. Based on this target waste status prediction, the system calculates when each waste collection point will reach its capacity threshold—the point at which it will overflow—and analyzes the generation and distribution of each type of waste.

[0016] In this embodiment, a hierarchical cluster analysis of the historical and real-time state matrices of garbage collection points is performed along the time dimension. Using clustering algorithms such as hierarchical clustering and K-means, time periods such as those within a 24-hour day, day of the week, and season are categorized according to the changing characteristics of garbage generation and accumulation. This analysis identifies various typical garbage generation patterns, such as those during peak hours in the morning and evening, weekdays and holidays, and summer and winter. Conditional probabilities are then calculated for garbage state transitions within each pattern. All state pairs that occur when state vectors (e.g., weight, capacity occupancy, and type distribution) transition from one time point to the next within a specific pattern are counted. These state pairs are aggregated into a pattern-specific state transition frequency matrix, capturing the evolution of garbage states under different scenarios. To mitigate transition probability distortion caused by data sparsity, the transition frequency matrix is normalized and a Laplace smoothing factor is introduced to correct the transition frequencies of all state pairs, resulting in a garbage state transition probability matrix. Based on the current date and specific time, the most suitable garbage generation pattern is automatically retrieved and matched, and the corresponding state transition probability matrix is selected. Combined with the state vector of the garbage collection point at the current moment, a first-order Markov prediction value is calculated through matrix multiplication. This prediction value reflects the most likely direction of change in the garbage state at the next time step. Considering the continuity of the garbage generation process and the need for long-term prediction, the state transition equation is used to perform an n-step iterative prediction of the above first-order Markov prediction value, and a state sequence containing multi-step prediction uncertainty is obtained through recursion. To improve prediction accuracy, the above state sequence is calibrated in combination with the garbage volume estimation parameter matrix. These parameter matrices are continuously adaptively updated using historical real data based on optimization methods such as least squares, thereby ensuring that the initial garbage state prediction results output by the model are as close to the actual situation as possible.

[0017] 103. Based on the overflow prediction time of garbage collection points, garbage type distribution ratio data, and real-time road traffic data, the execution resources of the urban garbage fleet are digitally mapped to obtain collaborative task partition data; Specifically, a digital twin modeling of the road network is constructed across the entire city. Using GIS technology and high-precision map data, a basic road network model is established. This model structures all road nodes, connecting edges, and their spatial topological relationships, providing quantifiable geometric attributes and connectivity descriptions for each road. As the city's operations progress, the system receives real-time traffic data, including traffic flow, road congestion, and construction closures. Based on this data, the system dynamically adjusts the attribute values of each road edge set in the basic road network model, including key parameters such as current average speed, congestion coefficient, and capacity. This creates a dynamic road network state model that reflects real-time changes in road conditions. Furthermore, attribute data such as the predicted overflow time and garbage type distribution ratio of each garbage collection point, obtained through intelligent analysis, is spatially aligned with the dynamically updated road network state model according to its geographic coordinates. Each garbage collection point is treated as a special attribute node on the network, and its overflow time and garbage type structure are incorporated into the network structure in the form of weights and labels, resulting in a network distribution map with garbage spatiotemporal attributes. A geospatial cluster analysis is performed on the garbage collection network distribution map. Using methods such as DBSCAN, Mean-Shift, or graph clustering, multidimensional clustering is performed based on factors such as the spatial distribution of garbage collection points, garbage generation intensity, and overflow urgency. Collection points with dense distribution and similar spatiotemporal attributes are then divided into preliminary task partitions, achieving a preliminary balance between workload and spatial distance. The preliminary task partitioning results are adjusted for resource balancing by comprehensively considering the garbage fleet's existing execution resources, including vehicle number, vehicle type, operational capacity, and current vehicle location. The number and type of vehicles assigned to each partition are dynamically optimized to best meet the garbage collection needs of their respective areas, maximize fleet resource utilization, and minimize the probability of empty runs and congestion. A vehicle resource allocation plan is then generated. A collaborative topological relationship diagram is constructed for each vehicle in the vehicle resource allocation plan, clarifying inter-vehicle information transmission paths, collaborative priorities, and task succession mechanisms. By constructing topological structures such as directed spanning trees, this ensures efficient distribution of dispatch information and emergency events within the fleet, enabling coordinated responses. All partitions, vehicles, tasks, and collaborative relationships are uniformly encoded and abstracted, forming structured collaborative task partition data.

[0018] 104. Perform garbage disposal recursive optimization and scheduling analysis based on collaborative task partition data to obtain the target scheduling instruction set.

[0019] Specifically, the city's entire operating area is divided into K sub-areas based on multi-dimensional factors such as geographic distribution, transportation connectivity, and waste disposal needs. Each sub-area exhibits spatial aggregation and reflects diverse characteristics such as the density of waste generation, the urgency of overflow, and the type structure within the area. Within each sub-area, a sub-objective function is set that comprehensively considers key performance indicators such as timely waste removal rate, fleet energy consumption, removal time, overflow risk, and vehicle utilization. By quantifying these objectives into mathematical models, a dynamic programming algorithm can efficiently retrieve optimal or near-optimal local solutions within the state space, ensuring that each sub-area achieves the optimal balance between its own operating efficiency and service quality within existing resources and constraints. Once the local optimal solution sets for all sub-areas are generated, the regional coordination variable matrix is constructed, taking into account the dynamic flow of resources and coordination needs between sub-areas. Each element in the matrix represents the dynamic adjustment relationship between vehicles, time, or tasks between areas. Based on this collaborative variable matrix, recursive optimization tools such as the alternating direction multiplier method are employed to decompose the global optimization objective into regional sub-objectives and regional collaborative objectives. Through the interaction of variables within and outside the region and iterative corrections using Lagrange multipliers, cross-regional information transfer and global convergence are achieved, efficiently converging to a global collaborative optimal solution under multi-objective constraints. Based on the global collaborative optimal solution, each garbage truck is precisely assigned a service area, collection route, and task sequence. A preliminary task allocation plan is generated and distributed to each vehicle's terminal system in real time. To ensure the safety and adaptability of the scheduling plan, after the allocation is completed, the overflow risk assessment mechanism is combined to analyze the overflow prediction time and tolerance range of each collection point to identify high-risk nodes. Furthermore, combined with vehicle handling capacity matching analysis, task instructions are adjusted or fine-tuned based on parameters such as vehicle load capacity, garbage type adaptability, energy consumption, and historical scheduling performance. After multiple rounds of optimization and matching analysis, a target scheduling instruction set is formed, defining each garbage truck's partition, priority, route, operating time, and contingency plan.

[0020] In this embodiment, a garbage overflow risk function is established for each garbage collection point. This function comprehensively considers key indicators such as the current capacity occupancy percentage, predicted overflow time, garbage growth rate, and the time interval since the last collection. Through weighted superposition and penalty coefficient design, it forms a quantitative expression of the probability and urgency of future garbage overflows at each collection point. Based on this, a garbage collection point risk level matrix is generated, which structured and graded the risk status of all collection points and clearly defines which points require priority. Simultaneously, the processing capabilities of all vehicles in the fleet are refined and differentiated. For the three major types of recyclables, kitchen waste, and other waste, a garbage type processing capacity matrix is constructed, combining factors such as the vehicle's functional modules, load configuration, and delivery compatibility. This matrix reflects each vehicle's actual operating capacity for different garbage categories. The garbage type processing capacity matrix is matched with the garbage type distribution ratio data for each garbage collection point to comprehensively evaluate the compatibility between each vehicle and each collection point, namely the vehicle-collection point compatibility data. The risk level matrix of waste collection points and vehicle-collection point compatibility data are simultaneously integrated into a cooperative game framework to construct a differential game model for multi-vehicle collaboration. This model treats each vehicle as a player and treats the action combinations and decisions of all vehicles as a set of differential control variables. The model aims to improve global collection efficiency, reduce overflow risk, and increase the sorting and processing rate. By balancing each player's immediate payoff function with the overall system goal, information sharing and balanced interests among fleet members are achieved. In the actual solution process, to handle complex constraints such as vehicle capacity and operation time windows, a barrier function is introduced into the differential game model. By functionally encoding soft and hard constraints such as each vehicle's capacity limit and the latest acceptable time for waste collection, over-limit and overtime behaviors are dynamically penalized in the optimization objective, forming a modified value function equation. Utilizing the revised value function equation, alternating updates are performed within the multi-vehicle differential game strategy space. Methods such as strategy iteration and gradient ascent are employed to continuously approximate the optimal control solution set both globally and locally. This results in a Pareto-optimal scheduling strategy set under multiple objectives and constraints, achieving an optimal balance between total waste collection and transportation costs, overflow risk, and fleet utilization. To adapt to the dynamic changes in urban waste generation and road network fluctuations, the Pareto-optimal scheduling strategy set is optimized over a rolling time domain. The scheduling plan is dynamically updated based on the previous cycle's execution results and the latest monitoring data. Through real-time command conversion, each vehicle's service area, operation sequence, route planning, and emergency response plan are converted into an executable target scheduling command set, which is then distributed to the vehicle terminal system.

[0021] In this embodiment of the present invention, intelligent monitoring devices installed at garbage collection points and onboard terminals on vehicles enable real-time collection and transmission of multi-dimensional data, including garbage weight, type, and volume, as well as vehicle location, load, and driving status. This eliminates the data silos in traditional systems. Dynamic, time-varying analysis of garbage status based on a Markov state model enables the system to accurately predict overflow times and garbage type distribution at each collection point, thereby shifting from passive response to proactive prevention and effectively preventing overflowing garbage bins. By digitally mapping the execution resources of the city's garbage fleet and combining overflow prediction times with real-time road traffic conditions, the system generates more rational collaborative task partitions, enabling precise allocation of garbage truck resources. Regional decomposition and task splitting calculations of task partition data significantly improve system processing speed. The introduction of a garbage overflow risk function for risk assessment and the use of an adaptive cooperative game control mechanism to address system constraints enable the system to effectively control overflow risks while meeting garbage bin capacity and vehicle load constraints. By constructing a comprehensive evaluation function that takes into account the balance of total transportation time, total energy consumption, garbage overflow risk and vehicle utilization, the system can achieve multi-objective Pareto optimal scheduling while meeting multiple requirements such as transportation efficiency, energy conservation and service quality.

[0022] In a specific embodiment, the process of executing step 101 may specifically include the following steps: The raw garbage data collected by the intelligent monitoring device at the garbage collection point is filtered for outliers and pre-processed by edge computing to obtain garbage weight data, garbage type data, and capacity occupancy data; The GPS and Beidou dual-system positioning signals collected by the garbage truck terminal are integrated and converted into coordinates to obtain vehicle location data. The raw vehicle data collected by the vehicle status monitoring module of the garbage truck terminal is processed and formatted in real time to obtain load data and driving status data. Encapsulate garbage weight data, garbage type data, capacity occupancy data, vehicle location data, load data, and driving status data to obtain an original data packet; The original data packet is securely encrypted to obtain a layered encrypted data stream, which is then transmitted to the cloud platform in real time to obtain comprehensive garbage management data.

[0023] Specifically, the raw garbage data collected by intelligent monitoring devices at garbage collection points undergoes outlier filtering and edge computing preprocessing. These devices continuously record data such as the current weight of the garbage bin, garbage type identification results, and capacity occupancy percentage at a high-frequency sampling cycle. Data validation and cleaning algorithms are run in real time within the edge computing module. Statistical methods such as sliding window averaging, median filtering, and IQR outlier detection are used to identify anomalous data points that deviate significantly from the normal distribution. Thresholds are adaptively adjusted based on the device's health status and reporting frequency to dynamically remove abnormal records such as "jumps," "drifts," "duplications," and "missing data." The filtered data undergoes a lightweight local preprocessing process, including secondary calibration of the raw weight signal against the sensor calibration curve, feature extraction and fuzzy logic analysis of the garbage type identification results to eliminate occasional misjudgments, and smoothing and interpolation of capacity occupancy percentages based on time series trends, generating structured data on garbage weight, garbage type, and capacity occupancy. Furthermore, the garbage truck terminal, a key node in the intelligent development of the urban fleet, integrates and standardizes the collected data. Garbage truck terminals are equipped with dual GPS and Beidou positioning modules, leveraging the advantages of dual-mode positioning to enhance the continuity and accuracy of positioning signals. They utilize a multi-source signal fusion algorithm based on Kalman filtering or particle filtering to integrate raw GPS and Beidou position coordinates, velocity, and timestamp data in real time. A prediction-correction mechanism smooths out occasional signal jumps, dynamically weights the optimal positioning result, and combines the vehicle's own inertial navigation data (such as acceleration and gyroscope information) for short-term compensation, resulting in high-resolution, low-error vehicle position data. Before cloud-based analysis, vehicle position data undergoes coordinate system conversion. For example, the original WGS-84, GCJ-02, or local coordinate system is mapped to the reference system used for city maps, ensuring seamless overlay and analysis of all spatial data within the same framework. The garbage truck terminal also collects real-time vehicle operating status data. Its vehicle status monitoring module includes a load sensor, fuel sensor, speed / accelerometer, and operating condition acquisition unit. Raw vehicle status data is also processed and formatted locally in real time at the terminal. Load data undergoes multi-point sampling, temperature compensation, and zero-drift calibration to eliminate reading errors caused by mechanical wear and ambient temperature changes. It is then normalized to standard units to produce comparable vehicle load data. Data such as speed and driving trajectory exclude periods of stillness, abnormal acceleration, or data packet loss. Interpolation, filtering, and other algorithms are used to obtain continuous, smooth driving status data, and all data is structured into a unified format. After completing the local cleaning and structuring of the aforementioned garbage collection point and on-board terminal data, the garbage weight data, garbage type data, capacity occupancy data, vehicle location data, load data, and driving status data are packaged and integrated according to unified data fields based on the set data encapsulation protocol to generate a standardized raw data packet.The original data packets are encrypted in layers. Symmetric encryption algorithms such as AES-256 are used for underlying data encryption to ensure that the content cannot be decrypted during wireless transmission. Uploaded instructions and authentication information utilize asymmetric encryption techniques such as RSA to prevent identity forgery. Furthermore, all data streams are appended with integrity check codes generated using hash algorithms such as SHA-256 before transmission to effectively prevent tampering. Considering occasional communication interruptions in 5G or 4G network environments, data packets are designed with local caching and retransmission mechanisms, breakpoint resumption, and traffic tiering to prioritize urgent data uploads, while regular data is uploaded in batches. All encrypted and integrity-checked data streams are ultimately uploaded to the cloud platform in real time via a high-bandwidth wireless network. The cloud platform decrypts and verifies the data integrity at the receiving end, enabling seamless data storage through multi-layer authentication and protocol parsing. The cloud platform then uniformly formats and archives all uploaded data, aggregating, sorting, deduplicating, and aligning time series with historical data to form a comprehensive database for urban waste management.

[0024] In a specific embodiment, the process of executing step 102 may specifically include the following steps: Perform time series extraction on garbage weight data, garbage type data, and capacity occupancy data in the comprehensive garbage management data to obtain time series status data; Establish a garbage collection point status matrix based on time series status data, including current garbage weight, current capacity occupancy percentage, garbage growth rate and garbage type distribution ratio; Perform pattern division and Markov prediction on the garbage collection point state matrix to obtain the initial garbage state prediction result; The initial garbage status prediction result is corrected and calculated based on the external factor influencing data to obtain the target garbage status prediction result; According to the target garbage status prediction results, the time point when each garbage collection point reaches the capacity threshold and the generation ratio of each type of garbage are calculated to obtain the overflow prediction time of the garbage collection point and the garbage type distribution ratio data.

[0025] Specifically, time series data on garbage weight, garbage type, and capacity utilization are extracted from the comprehensive garbage management data. This data is continuously collected at a set sampling frequency, and the data from each garbage collection point at different time points is arranged chronologically to construct a time series status dataset. During this process, the system automatically corrects for discontinuities caused by packet loss, sampling anomalies, or timestamp desynchronization, completing and smoothing the time series through methods such as linear interpolation and sliding average. Based on the time series status data, a multidimensional state vector matrix is dynamically constructed to describe the operational characteristics of the garbage collection point. A two-dimensional garbage collection point status matrix is formed, with each sampling time step as the horizontal axis and parameters such as the current garbage weight, current capacity utilization percentage, garbage growth rate, and garbage type distribution ratio as the vertical axis. Garbage weight data reflects the current total amount of garbage, capacity utilization percentage reflects the actual filling level of the garbage bin, and the garbage growth rate is calculated based on the weight or capacity changes between two or more consecutive sampling points through differential or sliding window calculations, reflecting the trend of garbage accumulation. The system uses data from intelligent recognition sensors to accurately quantify the proportion of recyclables, food waste, and other waste at each time step, thereby constructing a waste classification structure. This state matrix captures the temporal evolution of waste collection points and changes in their classification structure. Pattern segmentation and Markov prediction are then performed on the state matrix of the waste collection points, yielding preliminary predictions of their future states. Pattern segmentation involves the system automatically identifying the cyclical and phased characteristics of waste generation based on analysis of historical and real-time time series. For example, it distinguishes between various operating modes, such as peak and off-peak hours, holidays, and special events. Based on this, the time series data is segmented into subsets with similar statistical properties. Within each mode, a Markov model is used to model the state transition patterns of waste. The current state vector of the waste collection point is used as the state node of the Markov chain. The state transition probability matrix is obtained by counting and normalizing the state transition frequency within each mode. A Laplace smoothing factor is then introduced to address low-frequency state pairs, ensuring good generalization of the model. Based on the garbage generation pattern corresponding to the current time, the most appropriate state transition probability matrix is selected. Starting from the current state vector, matrix multiplication is used to calculate the state probability distribution for the next step, or even multiple steps, to obtain first-order and multi-order Markov prediction values, generating a preliminary garbage state prediction sequence. Based on the preliminary Markov prediction, external factors influencing data are introduced to correct the prediction results. This correction process uses methods such as multivariate regression, weighting factors, time-varying bias compensation, or neural network external disturbance compensation. External factor parameters are used as auxiliary inputs to dynamically adjust the confidence interval and center value of the prediction results to obtain the target garbage state prediction results. Based on the target garbage state prediction results, the time when each garbage collection point reaches the capacity threshold is estimated.Specifically, the system analyzes the evolution of capacity utilization percentages within the forecast sequence, identifies the time step when the capacity threshold is first exceeded, and uses this time step as the predicted overflow point. Furthermore, the system calculates the cumulative proportion of each type of waste generated within the forecast interval, outputting the changing trends of the waste classification structure over different time periods.

[0026] In a specific embodiment, the process of performing pattern division and Markov prediction on the garbage collection point state matrix to obtain the initial garbage state prediction result may specifically include the following steps: Perform hierarchical cluster analysis on the time dimension of the garbage collection point status matrix to obtain multi-modal time classification data; The conditional probability of garbage state transition relationship under each mode is calculated based on multi-mode time classification data to obtain the mode-related state transition frequency statistical matrix; The mode-related state transition frequency statistics matrix is normalized, and the Laplace smoothing factor is introduced to deal with the data sparsity problem to obtain the garbage state transition probability matrix; Determine the corresponding garbage generation mode according to the current date and time, select the corresponding garbage state transition probability matrix, and perform matrix multiplication operation on the current garbage collection point state vector and the selected garbage state transition probability matrix to obtain the first-order Markov prediction value; The state transition equation is used to perform n-step iterative prediction on the first-order Markov prediction value to obtain a state sequence containing prediction uncertainty; The garbage quantity estimation parameter matrix is applied to the state sequence containing prediction uncertainty to calibrate and obtain the initial garbage state prediction result.

[0027] Specifically, a hierarchical cluster analysis of the garbage collection point state matrix along the time dimension is performed to structurally identify and summarize the diversity and complexity of garbage generation and accumulation patterns. By combining collection frequency and time span, the time period is divided into multiple basic units. Then, using machine learning algorithms such as K-means, hierarchical clustering, spectral clustering, or self-organizing maps, cluster analysis is performed on the state vectors of garbage collection points for each time period (including current weight, capacity percentage, growth rate, type distribution, etc.). Automatic learning from historical data identifies typical garbage generation patterns, such as those during weekdays and holidays, peak and off-peak hours in the morning and evening, seasonal variations, and special events. Each clustering result represents a statistically significant garbage generation pattern. These patterns are associated with corresponding time windows using pattern labels to form multimodal temporal classification data. Based on this multimodal temporal classification data, conditional probabilities are calculated for garbage state transitions within each pattern. The time series state data within each pattern are paired in chronological order, and the number of transitions from state Si to state Sj is counted to form a state transition frequency matrix associated with the pattern. Each element of this matrix records the frequency of transitions from one state vector to another under a specific pattern. The system traverses the entire dataset and constructs a transition frequency matrix for each pattern, thereby characterizing the probabilistic characteristics of garbage state changes under different scenarios. The state transition frequency matrix is normalized, normalizing all elements in each row by the sum of that row. This yields the conditional probability distribution of each state transitioning to another state under a specific pattern. Furthermore, to prevent some state transition probabilities from being zero due to short sampling periods, rare extreme cases, or uneven distribution of pattern samples, which could affect the generalization and robustness of the prediction, a Laplace smoothing factor is introduced. A very small positive number (usually 1) is added to all transition frequencies to eliminate zeroing caused by sparse data. After normalization and smoothing, garbage state transition probability matrices are obtained for various typical time patterns. The corresponding garbage generation pattern is determined based on the current date and time, and the corresponding garbage state transition probability matrix is selected. Based on the results of previous clustering and label binding, the corresponding transition probability matrix is retrieved. The state vector of the current garbage collection point is used as the initial state. By performing matrix multiplication with the state transition probability matrix for the selected mode, a first-order Markov prediction value is directly calculated. This prediction value describes the most likely garbage state distribution at the next time step. To meet the requirements of continuous multi-period prediction and uncertainty assessment, this first-order Markov prediction value is iterated n times using the state transition equation. This means that the prediction result of one time is used as the starting point for the next step, and forward recursion is continuously performed through the state transition matrix to form a state sequence covering multiple future time points. This state sequence, which includes prediction uncertainty, is calibrated using the garbage volume estimation parameter matrix.The waste volume estimation parameter matrix is a dynamic adjustment factor derived through long-term historical observations and statistical analysis. It is dynamically updated based on factors such as different time patterns, solar term changes, external influencing factors (such as temperature, rainfall, and major events), and the residual distribution between historical actual and predicted values. During the calibration process, the initial predicted state sequence is compared with the actual collected data for the current period and external constraint variables. Parameters such as weight, volume, and type distribution in the predicted sequence are adjusted based on historical model errors, external event characteristics, and adaptive feedback within the time window. After this calibration process, the initial waste state prediction result for the current collection point is obtained.

[0028] In a specific embodiment, the process of executing step 103 may specifically include the following steps: Conduct digital twin modeling of urban road networks and establish a basic road network model; Update the road edge set attributes in the basic road network model according to the real-time road traffic status data to obtain a dynamically updated road network status model; The overflow prediction time of garbage collection points and the distribution ratio of garbage types are spatially associated with the dynamically updated road network status model to obtain a network distribution map with garbage attributes. Conducting geospatial cluster analysis on the network distribution map with garbage attributes to obtain preliminary task partitioning results. This is then adjusted for resource balancing based on the execution resources of the city's garbage fleet to obtain a vehicle resource allocation plan. A collaborative topology diagram is established for each vehicle in the vehicle resource allocation plan, the information transmission path and collaboration priority between vehicles are determined, and the collaborative task partition data is obtained.

[0029] Specifically, a digital twin modeling of the entire city's road network is constructed to establish a basic road network model that closely replicates real-world traffic operation characteristics. This modeling process utilizes high-precision maps, road centerline data, intersection and node information, road grade attributes, and related spatial data, integrating GIS (Geographic Information System) technology with graph theory methods to abstract the urban road space into a directed weighted graph. Each node in this graph represents a key geographic coordinate point, including garbage collection points, major intersections, transportation hubs, transfer stations, and waste treatment plants. Each edge represents a traversable road segment, storing static attributes such as road length, width, and direction of travel. It also includes various elements such as historical average speed, speed limit, road grade (e.g., main road, secondary road, branch road), traffic light information, and driving rules. These are uniformly coded and managed in a structured database. Furthermore, the road edge attributes in the basic road network model are dynamically updated by integrating real-time traffic data obtained from urban traffic perception systems (e.g., traffic flow monitoring, road surface sensors, traffic video, traffic police event data, and third-party navigation real-time APIs). The weight of each road edge is based not only on the length of the road segment but also on factors such as the current actual speed, congestion coefficient, traffic volume, temporary construction or regulation, and accident alerts. The system automatically adjusts the cost and status attributes of the road edge set based on collected traffic data, transforming the previously static road network diagram into a dynamic network state model that truly reflects the real-time accessibility of urban roads. This model supports attribute refreshes at a frequency of minutes or even higher, instantly reflecting any traffic jams, road closures, or abnormal congestion. The system also utilizes short-term traffic forecasting models to estimate road traffic conditions for several future time windows, giving the network state model a spatiotemporal linkage and forward-looking nature. As the dynamic attributes of the road network model continue to improve, the system spatially links data such as predicted overflow times for garbage collection points and the distribution of garbage types, previously generated through IoT intelligent monitoring and cloud-based intelligent predictive analysis, with the dynamically updated road network state model. The geographic coordinates of each garbage collection point are precisely mapped to nodes or adjacent road nodes in the network diagram. Attributes such as the corresponding overflow prediction time (i.e., the earliest predicted time of collection), garbage classification structure (ratio of recyclables, kitchen waste, and other waste) are embedded in the network model as labels, weights, or node attributes, resulting in a network distribution map with garbage attributes. Based on this network distribution map with garbage attributes, geospatial clustering analysis is used to spatially partition all garbage nodes based on multi-dimensional characteristics such as overflow urgency, spatial distribution density, type homogeneity, and road accessibility. Clustering algorithms such as DBSCAN, spectral clustering, or network-based community discovery algorithms are used to identify clusters of nodes that are close in spatial proximity, have similar predicted overflow times, and have similar type structures, thereby dividing the city into a number of preliminary task zones.This zoning scheme takes into account both physical proximity and waste disposal needs and road timeliness. The goal is to minimize cross-region dispatching and empty trips, thereby improving single-vehicle removal efficiency and task consistency. The initial zoning scheme is adjusted for resource balancing based on the number, location, vehicle type, load capacity, waste disposal suitability, and historical task execution of each vehicle in the fleet. Through integer programming or heuristic allocation algorithms, an appropriate number and type of vehicles are allocated to each zone to achieve an optimal balance between task load, vehicle capacity, and waste disposal needs, thus forming a vehicle resource allocation plan. A collaborative topological relationship graph is established for each vehicle in the vehicle resource allocation scheme. In this relationship graph, each vehicle is abstracted as a node, and the edges between nodes represent physical reachability and reflect the priorities and transfer paths between vehicles for task handover, information sharing, and scheduling collaboration. For example, vehicles in adjacent zones or the same region communicate via point-to-point wireless communications or cloud platforms. The system sets dynamic collaboration priorities based on task priority, geographic proximity, and vehicle status (such as fuel level, load, and task completion rate). When necessary, it dynamically generates mechanisms such as task transfers and support requests. This ensures that in extreme cases, when a vehicle fails in a particular zone or tasks surge, adjacent vehicles can be quickly coordinated for support or tasks can be reallocated. The collaborative topology diagram is optimized using directed spanning trees, hierarchical networks, or small-world structures to shorten task response times and information synchronization delays. All zones, vehicles, nodes, tasks, information transfer, and collaboration mechanisms are uniformly encapsulated to form structured collaborative task partition data.

[0030] In a specific embodiment, the process of executing step 104 may specifically include the following steps: The urban area in the collaborative task partition data is divided into K sub-areas according to geographical location and garbage disposal needs, the sub-objective function of each sub-area is defined, and the local optimal solution set of the sub-area is solved by the dynamic programming algorithm; Based on the sub-region local optimal solution set, a regional collaborative variable matrix is defined, and recursive optimization using the alternating direction multiplier method is performed based on the regional collaborative variable matrix and the sub-region local optimal solution set to obtain a global collaborative optimization result. According to the global collaborative optimization results, each garbage truck is assigned a service area and task sequence to obtain the initial task allocation plan; The initial task allocation plan is evaluated for garbage overflow risk and vehicle handling capacity matching analysis to obtain the target scheduling instruction set.

[0031] Specifically, based on collaborative task partitioning data, the entire urban area is divided according to geographic location and waste disposal needs. Using a network distribution map with waste attributes, combined with spatial clustering algorithms such as DBSCAN, K-means, and spectral clustering, the urban space is divided into K sub-areas based on multiple dimensions, including geographic proximity, road accessibility, predicted overflow time, and waste type distribution. Each sub-area corresponds to a set of waste collection points that are spatially close, have similar temporal overflow trends, and exhibit high homogeneity in waste type structure. A sub-objective function is defined for each sub-area. The design of the sub-objective function comprehensively considers various operational performance indicators, such as total waste collection time within the area, vehicle energy consumption, waste overflow risk, vehicle utilization balance, and waste sorting and processing rate. The objective function uses a weighted multi-objective expression, with weight coefficients flexibly set based on actual needs. To determine the optimal operational strategy for each sub-area within local resource constraints, a dynamic programming algorithm is used to recursively search its task space and construct the optimal solution in stages. The advantage of dynamic programming lies in its ability to effectively address state transitions and stage coupling. It decomposes complex decisions such as subregional route selection, vehicle allocation, and clearing sequence into a series of iteratively solvable subproblems, yielding a local optimal solution set for each subregion. Based on the subregional local optimal solution set, a regional coordination variable matrix is defined. Each element of this matrix reflects the dynamic coordination relationships between adjacent subregions, such as resource exchange, task handover, vehicle support, and cross-regional operations. By combining the local optimal solutions with the regional coordination variables, scheduling variables such as vehicles, operation times, and routes are dynamically adjusted globally without violating the service constraints of each region. To effectively solve this global multi-objective, multi-constrained optimization problem, an alternating direction multiplier method is used for recursive optimization. The alternating direction multiplier algorithm decomposes the global objective into several subproblems with local constraints. Through repeated iterations, multiplier corrections, and information exchange in the variable space, it achieves a dynamic balance between the optimal variables within and outside the region. While each subregion optimizes its local operation plan under dynamic programming, the regional coordination variables continuously adjust the local solution set using the multiplier method to approach the global optimum. In each round of recursion, local variables and global variables converge synchronously, and eventually the city's job scheduling system reaches the Pareto optimal collaborative solution in terms of vehicle resource utilization, timely garbage removal, overflow risk control, and energy consumption constraints. Based on the global collaborative results obtained by ADMM optimization, the most suitable service area and task sequence are assigned to each garbage truck to generate a preliminary task allocation plan. The service area of each vehicle not only takes into account its current geographical location, available load capacity, and garbage disposal adaptability, but also combines the globally optimal path design and time sequence arrangement to ensure that all job tasks can be completed efficiently in the shortest time, with the lowest energy consumption, and with the lowest risk. The system presets backup tasks and flexible scheduling windows in task allocation to deal with special scenarios such as road emergencies, vehicle failures, and task surges.To ensure the practicality and robustness of the initial task allocation plan, a garbage overflow risk assessment and vehicle handling capacity matching analysis are conducted. In terms of risk assessment, the collection time calculated according to the target scheduling plan for each garbage collection point is compared with its overflow prediction time. An overflow risk function is used to quantify the risk level of each node under the current allocation. If a high-risk point is found, its task priority is dynamically increased or reinforcement vehicles are temporarily assigned. In terms of vehicle capacity matching, based on parameters such as each vehicle's processing efficiency, load capacity, and delivery adaptability for different types of garbage, combined with the distribution of garbage types in each region, an adaptability analysis of vehicle task allocation is conducted. If necessary, the partition affiliation or collection order of some vehicles is adjusted to ensure that each vehicle can perform at its best. Integrating the overflow risk level and adaptability indicators, multi-vehicle collaborative game or mixed integer programming is used to further fine-tune task allocation to form a target scheduling instruction set.

[0032] In a specific embodiment, the execution step performs garbage overflow risk assessment and vehicle handling capacity matching analysis on the initial task allocation plan to obtain a target scheduling instruction set, which may specifically include the following steps: Calculate the garbage overflow risk function for each garbage collection point in the initial task allocation plan to obtain the garbage collection point risk level matrix; Based on the processing capacity coefficients of recyclables, kitchen waste and other waste of each vehicle in the garbage fleet, a garbage type processing capacity matrix is constructed. The garbage type processing capacity matrix is then matched with the garbage type distribution ratio data of the garbage collection points to obtain the vehicle-collection point compatibility data. The garbage collection point risk level matrix and vehicle-collection point compatibility data are integrated into the cooperative game framework to construct a differential game model for multi-vehicle collaboration. A barrier function is introduced into the differential game model to handle vehicle capacity and time window constraints, and a revised value function equation is obtained. Based on the revised value function equation, the optimal control strategy is alternately updated to obtain the Pareto optimal scheduling strategy set. The Pareto optimal scheduling strategy set is optimized in rolling time domain and the vehicle task instruction is converted to obtain the target scheduling instruction set.

[0033] Specifically, a waste overflow risk function is calculated for each collection point in the initial task allocation plan. A comprehensive risk function is established for each collection point, incorporating parameters such as the current capacity occupancy percentage, waste growth rate, time since last collection, predicted overflow time, and category specificity. Risk values are calculated sequentially for all collection points, and the results are discretized into several levels (such as high, medium, and low) to form a waste collection point risk matrix. The actual handling capacity of each vehicle in the garbage fleet for different waste types is considered. The handling capacity coefficients of all vehicles for recyclables, food waste, and other waste are calculated and summarized to form a waste type handling capacity matrix. Each element in this matrix represents a vehicle's handling efficiency, load adaptability, or sorting capacity for a particular type of waste. This metric is derived from multiple sources, including vehicle configuration parameters, historical dispatch records, and equipment compatibility. Furthermore, standardized data on the distribution of waste types at each collection point is extracted to clearly define the proportion of each type of waste at each point in the recent or forecast period. Using methods such as matrix multiplication or cosine similarity, each vehicle's processing capacity vector is matched with the waste type distribution vector at each collection point. This yields vehicle-collection point compatibility data, reflecting the suitability of each vehicle for sorting operations at a particular collection point. After determining the risk level and compatibility, the multi-vehicle collaborative differential game modeling phase begins. All vehicles and tasks are considered players and targets in the game, and the waste collection point risk level matrix and vehicle-collection point compatibility data are integrated into the payoff function and task allocation strategy. The model assigns an immediate payoff function to each vehicle, incorporating both priority for clearing high-risk points and bonus points for highly adapted sorting operations. It also incorporates collaborative, supportive, and resource-switching strategies across multiple vehicles to optimize overall payoffs globally. The goal of the entire game system is not only to maximize the efficiency of individual vehicles but also to emphasize cooperation, complementarity, and dynamic adjustment among multiple vehicles. The differential game model employs recursive optimization and strategy evolution mechanisms, continuously revising each vehicle's route and task sequence through state-space differential equations until a global optimum or Nash equilibrium is reached. Actual transportation scheduling is inevitably influenced by physical constraints such as vehicle capacity and task time windows. To explicitly address these constraints within the differential game model, a barrier function is introduced into the optimization process. This barrier function uses a negative logarithm to impose extremely high penalties on non-compliant behaviors such as vehicle overloading and task timeouts, ensuring that all optimization searches are conducted within the constrained feasible domain. The correction value function equation with the barrier function combines immediate benefits, risk indicators, capacity adaptation, and constraint penalties to become the objective function for updating vehicle dynamic policies. Based on this equation, the system uses alternating updates (such as hybrid methods such as policy iteration, gradient ascent and descent, and Lagrange multipliers) to repeatedly adjust the task sequences and coordination modes of all vehicles, ensuring that each round of decision-making converges towards minimizing risk, optimizing capacity matching, and maximizing global resource utilization.The optimization output is a Pareto-optimal set of scheduling strategies. Each solution in this set cannot improve a key performance indicator without degrading other objectives, reflecting a multi-objective balance between efficiency, risk, constraints, and cooperation. To address the dynamic changes and emergencies in urban waste collection and transportation, a rolling time-domain optimization of the Pareto-optimal scheduling strategy set is performed. This optimization strategy uses a dynamic scheduling cycle and uses the actual execution results of the previous cycle, the latest waste monitoring, and traffic conditions as input to incrementally adjust and fine-tune the existing scheduling plan. Through rolling optimization, the system promptly detects and responds to anomalies such as road congestion, vehicle breakdowns, and sudden increases in waste volume, achieving real-time closed-loop iteration of the scheduling plan. After each time-domain optimization, the system outputs the scheduling decision as a standardized vehicle task instruction set, including detailed information such as each vehicle's operation route, collection sequence, task priority, waste type handling instructions, and emergency dispatch recommendations. This information is then distributed to each vehicle terminal via wireless communication networks such as 5G / 4G.

[0034] The above describes the urban operation fleet collaborative operation method based on the cloud platform in the embodiment of the present invention. The following describes the urban operation fleet collaborative operation system based on the cloud platform in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an urban operation fleet collaborative operation system based on a cloud platform includes: The real-time transmission module 201 is used to transmit the data collected by the intelligent monitoring device at the garbage collection point and the terminal on the garbage truck to the cloud platform in real time to obtain comprehensive garbage management data; Dynamic time-varying analysis module 202 is used to perform dynamic time-varying analysis on the garbage accumulation situation of each garbage collection point based on the comprehensive garbage management data, and obtain the overflow prediction time of the garbage collection point and the distribution ratio data of garbage types; The digital mapping module 203 is used to digitally map the execution resources of the urban garbage fleet based on the predicted overflow time of the garbage collection point, the garbage type distribution ratio data, and the real-time road traffic status data to obtain the collaborative task partition data; The scheduling analysis module 204 is used to perform garbage processing recursive optimization and scheduling analysis based on the collaborative task partition data to obtain a target scheduling instruction set.

[0035] Through the collaborative efforts of these components, intelligent monitoring devices installed at waste collection points and onboard terminals installed on vehicles enable real-time collection and transmission of multi-dimensional data, including waste weight, type, and volume, as well as vehicle location, load, and driving status. This breaks the data silos of traditional systems. Dynamic, time-varying analysis of waste status based on a Markov state model enables the system to accurately predict overflow times and waste type distribution at each collection point, effectively shifting from passive response to proactive prevention and effectively avoiding overflowing bins. By digitally mapping the execution resources of the city's waste fleet, combined with predicted overflow times and real-time road traffic conditions, the system generates more rational collaborative task partitions, enabling precise allocation of waste truck resources. Regional decomposition and task splitting of task partition data significantly improves system processing speed. The introduction of a waste overflow risk function for risk assessment and the use of an adaptive cooperative game control mechanism to address system constraints enable the system to effectively control overflow risks while meeting waste bin capacity and vehicle load constraints. By constructing a comprehensive evaluation function that takes into account the balance of total transportation time, total energy consumption, garbage overflow risk and vehicle utilization, the system can achieve multi-objective Pareto optimal scheduling while meeting multiple requirements such as transportation efficiency, energy conservation and service quality.

[0036] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0037] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0038] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A collaborative operation method for urban operation fleet based on cloud platform, characterized in that: include: Transmit data collected by intelligent monitoring devices at garbage collection points and terminals on garbage trucks to the cloud platform in real time to obtain comprehensive garbage management data; Performing a dynamic time-varying analysis of garbage accumulation at each garbage collection point based on the comprehensive garbage management data to obtain garbage collection point overflow prediction time and garbage type distribution ratio data; Based on the predicted overflow time of the garbage collection point, the garbage type distribution ratio data, and the real-time road traffic status data, the execution resources of the urban garbage fleet are digitally mapped to obtain collaborative task partition data; Based on the collaborative task partition data, garbage disposal recursive optimization and scheduling analysis are performed to obtain a target scheduling instruction set.

2. The cloud platform-based urban operation fleet collaborative operation method according to claim 1 is characterized in that: The data collected by the intelligent monitoring device at the garbage collection point and the garbage truck terminal are transmitted to the cloud platform in real time to obtain comprehensive garbage management data, including: The raw garbage data collected by the intelligent monitoring device at the garbage collection point is filtered for outliers and pre-processed by edge computing to obtain garbage weight data, garbage type data, and capacity occupancy data; The GPS and Beidou dual-system positioning signals collected by the garbage truck terminal are integrated and converted into coordinates to obtain vehicle location data. The raw vehicle data collected by the vehicle status monitoring module of the garbage truck terminal is processed and formatted in real time to obtain load data and driving status data. Encapsulating the garbage weight data, the garbage type data, the capacity occupancy data, the vehicle position data, the load data, and the driving status data to obtain an original data packet; The original data packet is securely encrypted to obtain a layered encrypted data stream, and the layered encrypted data stream is transmitted to the cloud platform in real time to obtain comprehensive garbage management data.

3. The cloud platform-based urban operation fleet collaborative operation method according to claim 1 is characterized in that: The dynamic time-varying analysis of the garbage accumulation situation at each garbage collection point is performed based on the comprehensive garbage management data to obtain the overflow prediction time of the garbage collection point and the garbage type distribution ratio data, including: Performing time series extraction on the garbage weight data, garbage type data, and capacity occupancy data in the garbage management comprehensive data to obtain time series status data; Establishing a garbage collection point status matrix including current garbage weight, current capacity occupancy percentage, garbage growth rate, and garbage type distribution ratio based on the time series status data; Performing pattern division and Markov prediction on the garbage collection point state matrix to obtain an initial garbage state prediction result; Correcting the initial garbage status prediction result based on the external factor impact data to obtain a target garbage status prediction result; The time point at which each garbage collection point reaches the capacity threshold and the generation ratio of each type of garbage are calculated based on the target garbage status prediction result, and the overflow prediction time of the garbage collection point and the garbage type distribution ratio data are obtained.

4. The cloud platform-based urban operation fleet collaborative operation method according to claim 3 is characterized in that: The performing pattern division and Markov prediction on the garbage collection point state matrix to obtain an initial garbage state prediction result includes: Performing a time dimension hierarchical cluster analysis on the garbage collection point status matrix to obtain multi-modal time classification data; Performing conditional probability calculation on the garbage state transition relationship under each mode based on the multi-mode time classification data to obtain a mode-related state transition frequency statistical matrix; The state transition frequency statistics matrix related to the pattern is normalized, and a Laplace smoothing factor is introduced to deal with the data sparsity problem to obtain a garbage state transition probability matrix; Determine the corresponding garbage generation mode according to the current date and time, select the corresponding garbage state transition probability matrix, and perform matrix multiplication operation on the current garbage collection point state vector and the selected garbage state transition probability matrix to obtain the first-order Markov prediction value; Using a state transfer equation to perform n-step iterative prediction on the first-order Markov prediction value to obtain a state sequence containing prediction uncertainty; The state sequence containing prediction uncertainty is calibrated by applying a garbage quantity estimation parameter matrix to obtain an initial garbage state prediction result.

5. The cloud platform-based urban operation fleet collaborative operation method according to claim 1 is characterized in that: The method of digitally mapping the execution resources of the urban garbage fleet based on the predicted overflow time of the garbage collection point, the garbage type distribution ratio data, and the real-time road traffic status data to obtain collaborative task partition data includes: Conduct digital twin modeling of urban road networks and establish a basic road network model; updating the road edge set attributes in the basic road network model according to the real-time road traffic status data to obtain a dynamically updated road network status model; Spatially associating the predicted overflow time of the garbage collection point and the garbage type distribution ratio data with the dynamically updated road network state model to obtain a network distribution map with garbage attributes; Performing a geospatial cluster analysis on the network distribution map with garbage attributes to obtain a preliminary task partitioning result, and performing a resource balancing adjustment on the preliminary task partitioning result based on the execution resources of the urban garbage fleet to obtain a vehicle resource allocation plan; A collaborative topology diagram is established for each vehicle in the vehicle resource allocation scheme, the information transmission path and collaboration priority between vehicles are determined, and collaborative task partition data is obtained.

6. The cloud platform-based urban operation fleet collaborative operation method according to claim 1 is characterized in that: The garbage disposal recursive optimization and scheduling analysis based on the collaborative task partition data is performed to obtain a target scheduling instruction set, including: Divide the urban area in the collaborative task partition data into K sub-areas according to geographical location and garbage disposal needs, define a sub-objective function for each sub-area, and solve the local optimal solution set of the sub-area through a dynamic programming algorithm; defining a regional collaborative variable matrix based on the sub-region local optimal solution set, and performing recursive optimization using an alternating direction multiplier method based on the regional collaborative variable matrix and the sub-region local optimal solution set to obtain a global collaborative optimization result; Allocating a service area and a task sequence for each garbage truck according to the global collaborative optimization result to obtain an initial task allocation plan; The initial task allocation plan is subjected to garbage overflow risk assessment and vehicle handling capacity matching analysis to obtain a target scheduling instruction set.

7. The cloud platform-based urban operation fleet collaborative operation method according to claim 6 is characterized in that: The initial task allocation plan is subjected to garbage overflow risk assessment and vehicle handling capacity matching analysis to obtain a target scheduling instruction set, including: Calculating a garbage overflow risk function for each garbage collection point in the initial task allocation plan to obtain a garbage collection point risk level matrix; Based on the processing capacity coefficients of recyclables, kitchen waste, and other waste for each vehicle in the garbage fleet, a garbage type processing capacity matrix is constructed. The garbage type processing capacity matrix is then matched with the garbage type distribution ratio data of the garbage collection points to obtain vehicle-collection point compatibility data; Integrating the garbage collection point risk level matrix and the vehicle-collection point compatibility data into a cooperative game framework to construct a multi-vehicle collaborative differential game model; Introducing a barrier function into the differential game model to process vehicle capacity and time window constraints, obtaining a revised value function equation, and alternately updating and searching for an optimal control strategy based on the revised value function equation to obtain a Pareto optimal scheduling strategy set; The Pareto optimal scheduling strategy set is subjected to rolling time domain optimization and vehicle task instruction conversion to obtain a target scheduling instruction set.

8. A cloud-based urban fleet collaborative operation system, characterized by: For executing the cloud platform-based urban operation fleet collaborative operation method according to any one of claims 1 to 7, the cloud platform-based urban operation fleet collaborative operation system comprises: Real-time transmission module, used to transmit data collected by intelligent monitoring devices at garbage collection points and garbage truck-mounted terminals to the cloud platform in real time to obtain comprehensive garbage management data; A dynamic time-varying analysis module is used to perform a dynamic time-varying analysis on the garbage accumulation situation at each garbage collection point based on the comprehensive garbage management data, and obtain the overflow prediction time of the garbage collection point and the garbage type distribution ratio data; A digital mapping module is used to digitally map the execution resources of the urban garbage fleet based on the predicted overflow time of the garbage collection point, the garbage type distribution ratio data, and real-time road traffic status data to obtain collaborative task partition data; The scheduling analysis module is used to perform garbage disposal recursive optimization and scheduling analysis based on the collaborative task partition data to obtain a target scheduling instruction set.

Images