Intelligent scheduling system for electric sanitation vehicle

Abstract

The invention relates to the technical field of intelligent scheduling of sanitation vehicles, and discloses an intelligent scheduling system for an electric sanitation vehicle, and the system comprises an information collection module which is used for obtaining the operation state information of the electric sanitation vehicle; the state modeling module is used for constructing a multi-dimensional data model for describing a vehicle scheduling state based on the operation state information; the task relation modeling module is used for constructing an association structure between the vehicle and the task; the scheduling optimization module is used for executing scheduling strategy optimization operation and generating a vehicle task allocation result; the instruction control module issues the scheduling instruction to a target vehicle to drive the target vehicle to execute a task; and the scheduling monitoring module is used for receiving manual intervention operation, and the intervention operation is fed back to the state modeling module and the task relation modeling module. According to the method, a tensor expression structure based on the vehicle, electric quantity, task and time four-dimensional state is constructed, and a sparse tensor compression mechanism is introduced, so that the storage and calculation cost can be remarkably reduced while the task performability is reserved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent scheduling of sanitation vehicles, in particular to an intelligent scheduling system for electric sanitation vehicles. BACKGROUND

[0002] With the continuous improvement of the automation level of urban cleaning operations, electric sanitation vehicles, as green, low-noise, and low-emission operation equipment, have gradually replaced traditional fuel vehicles and are widely deployed in main roads, branch roads, parks, campuses, and other public areas. Their low-energy-consumption operation characteristics meet the requirements of urban energy-saving and emission-reducing policies and have strong potential for promotion. Due to their limited range and high charging frequency, the scheduling system needs to not only meet the basic requirements of task completion but also comprehensively consider multiple factors such as power distribution, path planning, and task timeliness to achieve efficient, safe, and stable fleet operation scheduling strategies.

[0003] Some existing sanitation management systems begin to integrate vehicle real-time monitoring, task allocation, and scheduling execution functions. A common implementation method is to match vehicle positioning information with a preset task list and assign tasks based on the current location or static scheduling rules. Some systems issue tasks centrally through a scheduling software platform, and vehicles execute them in turn.

[0004] However, the existing part of the sanitation management system has a single expression of the scheduling state, mostly represented by a two-dimensional table or a task queue to represent the relationship between vehicles and tasks, which is difficult to capture the dynamic relationship between multiple factors such as power, time, and task accessibility. Secondly, the scheduling strategy mostly optimizes the single target of the shortest distance and the earliest time, ignoring key factors such as power remaining, task fairness, and execution continuity, which leads to frequent interruptions or unbalanced loads in vehicle scheduling results during execution. Therefore, the present application provides an intelligent scheduling system for electric sanitation vehicles to solve the problems existing in the prior art. SUMMARY

[0005] In view of the deficiencies of the prior art, the purpose of the present application is to provide an intelligent scheduling system for electric sanitation vehicles, which solves the problems of single dimension in state modeling, lack of expression in scheduling graph structure, narrow optimization target, lack of energy constraints, and ineffective closed-loop feedback of artificial intervention in the existing scheduling method.

[0006] To achieve the above purpose, the present application is implemented by the following technical scheme: an intelligent scheduling system for electric sanitation vehicles, comprising the following modules: An information acquisition module is used to acquire the running state information of the electric sanitation vehicle and upload the running state information to the system platform. The state modeling module is used to construct a multi-dimensional data model describing the vehicle scheduling state based on the running state information; the task relationship modeling module is used to construct the association structure between vehicles and tasks based on the scheduling state data in the multi-dimensional data model. The scheduling optimization module is used to perform scheduling strategy optimization calculations based on the multidimensional data model and the associated structure, and generate vehicle task allocation results. The instruction control module is used to generate vehicle dispatch instructions based on the vehicle task allocation results, and to send the dispatch instructions to the target vehicles to drive them to perform tasks; The scheduling and monitoring module is used to display vehicle status and task execution status, and to receive manual intervention operations. The intervention operations are fed back to the status modeling module and the task relationship modeling module to update the scheduling input data.

[0007] Preferably, the information acquisition module includes: The location acquisition unit is used to obtain the vehicle's location coordinates in real time via a GPS device; The power monitoring unit is used to acquire the current battery power of the vehicle at a frequency of 1Hz and map it to an energy level. The status upload unit is used to encrypt and upload the location coordinates, power level and operation identifier to the dispatch system platform via 4G communication.

[0008] Preferably, the state modeling module includes: Multidimensional tensor building blocks are used to construct tensors in four dimensions: number of vehicles, number of tasks, number of time slices, and number of battery levels. Where N represents the total number of currently active vehicles in the system, M represents the total number of tasks to be assigned, K represents the number of future time slices, and L represents the number of battery level ranges; A sparse compression unit is used to compress the tensor into a weighted sum of several rank-one tensors using the CANDECOMP / PARAFAC decomposition method. The tensor state update unit is used to update the tensor element values ​​based on real-time vehicle feedback data to reflect the current scheduling feasibility.

[0009] Preferably, the task relationship modeling module includes: Graph structure building unit, used to build task-vehicle relationship graph based on the executability of vehicle tasks in tensors; The edge weight calculation unit is used to assign a weight to each edge in the graph; The graph reconstruction unit is used to perform subgraph pruning and edge weight reconstruction to locally update the graph structure when a vehicle fails or a task is interrupted.

[0010] Preferably, the scheduling optimization module includes: A multi-objective modeling unit is used to define task completion time, vehicle idle distance, power utilization, and task allocation fairness as scheduling optimization objectives; An optimized execution unit is used to generate Pareto optimal scheduling solutions through evolutionary optimization algorithms; The weighted adaptive unit is used to dynamically adjust the weights of each objective function based on the tensor feedback state, forming an adaptive fitness function, which is: Where s represents the current scheduling solution; This represents the normalized evaluation value of the j-th objective function under the scheduling solution s; j is the index number of the objective function; w j The dynamic weight coefficient corresponding to the j-th objective function; F(s) represents the final comprehensive fitness function value.

[0011] Preferably, the instruction control module includes: The decoding unit is used to parse the scheduling scheme generated by the optimization module into a set of vehicle-task mappings. The route generation unit is used to generate the optimal driving route for vehicles based on the current traffic conditions. The instruction generation and distribution unit is used to package task, route and power limit data into scheduling instructions and send them to the target vehicle.

[0012] Preferably, the scheduling and monitoring module includes: The status display unit is used to display the current task status, battery status and location information of each vehicle on the interface; An anomaly intervention unit is used to allow manual reassignment of tasks when scheduling anomalies or unexpected tasks occur; The intervention feedback unit is used to synchronously feed back the results of manual operations to the tensor modeling module and the task graph module to update the scheduling input status.

[0013] Preferably, the construction task-vehicle relationship diagram is calculated according to the following formula: Among them, w m,n This represents the association weight between vehicle n and task m in the task-vehicle relationship graph; α, β, and γ are system-preset weighting factors; d m,n Indicates the Euclidean distance between the vehicle and the mission; e n The remaining battery level of the vehicle; ρ m This represents the task priority level coefficient.

[0014] Preferably, the fitness function is based on the normalization function of each optimization objective. With adaptive weights w j The construction, in the form of: Where s represents the current scheduling solution; This is the normalized value of the total task completion time; This is the normalized value of the empty driving distance of the vehicle during the mission. This is the normalized value for the vehicle's energy utilization efficiency. Normalized values ​​for the fairness index of task allocation among vehicles; w j The dynamic weight coefficients corresponding to the j-th objective function.

[0015] A method for intelligent dispatching of electric sanitation vehicles is also provided, including the following steps: Obtain the location, battery level, and operational status information of each electric sanitation vehicle; Based on the information, a multidimensional data model is constructed to describe the feasibility of scheduling. Based on the data model, construct the association structure between vehicles and tasks; Perform scheduling strategy optimization calculations to generate vehicle task allocation schemes; The scheduling instructions are generated according to the proposed scheme and sent to the target vehicle; The scheduling model and associated structure are updated based on vehicle feedback and human intervention results during the scheduling process for the next round of scheduling optimization.

[0016] In summary, this application includes at least one of the following beneficial technical effects: 1. This invention constructs a tensor representation structure based on a four-dimensional state of vehicle, battery level, task, and time, and introduces a sparse tensor compression mechanism, which can significantly reduce storage and computation costs while preserving task executability. Compared with existing technologies that simply use two-dimensional matrices or linear queues to describe vehicle scheduling states, this scheme solves the problem of insufficient data dimensionality and weak expressive power leading to ambiguous scheduling boundaries.

[0017] 2. This invention introduces a triple-weighted mechanism of distance, power consumption, and task urgency between tasks and vehicle nodes, supporting a quantitative assessment of task execution feasibility. Compared to traditional scheduling systems that only rely on coarse-grained location distance to determine scheduling relationships, this improvement achieves a fine-grained, multi-factor integrated controllable scheduling structure, overcoming the bottleneck of execution failure caused by ignoring power constraints in scheduling decisions.

[0018] 3. This invention introduces a multi-objective evolutionary optimization algorithm and constructs an adaptive weighted scheduling fitness function, which can dynamically adjust the optimization focus according to the real-time state, balancing efficiency and fairness. Unlike previous scheduling logics that use fixed weights for single-objective or dual-objective optimization, this method effectively avoids the problem of system imbalance caused by excessive scheduling bias towards individual objectives, thus improving the overall sustainable scheduling capability.

[0019] 4. This invention designs a closed-loop mechanism of task insertion, state tracking, and feedback in the rescheduling and monitoring stage, enabling human intervention to participate in scheduling correction in real time and automatically feed back into the modeling structure. In traditional systems, human intervention is usually limited to surface display and static correction. This solution breaks the human-machine separation and realizes the immediate impact of human intentions on system modeling, thereby making up for the weakness of intelligent scheduling algorithms in adapting to sudden scenarios. Attached Figure Description

[0020] Figure 1 This is the system architecture diagram of this application; Figure 2 This is a flowchart of the method in this application. Detailed Implementation

[0021] The following is in conjunction with the appendix Figure 1 -Appendix Figure 2 This application will be described in further detail below.

[0022] Please see the appendix Figure 1 This invention provides an intelligent dispatch system for electric sanitation vehicles, including the following modules: an information acquisition module, used to acquire the operating status information of electric sanitation vehicles and upload the operating status information to the system platform; The state modeling module is used to construct a multi-dimensional data model describing the vehicle scheduling state based on the running state information; the task relationship modeling module is used to construct the association structure between vehicles and tasks based on the scheduling state data in the multi-dimensional data model. The scheduling optimization module is used to perform scheduling strategy optimization calculations based on the multidimensional data model and the associated structure, and generate vehicle task allocation results. The instruction control module is used to generate vehicle dispatch instructions based on the vehicle task allocation results, and to send the dispatch instructions to the target vehicles to drive them to perform tasks; The scheduling and monitoring module is used to display vehicle status and task execution status, and to receive manual intervention operations. The intervention operations are fed back to the status modeling module and the task relationship modeling module to update the scheduling input data.

[0023] In this embodiment, the information acquisition module acquires data through a sensing unit deployed on the vehicle terminal. The collected data includes, but is not limited to, key operating parameters such as the vehicle's geographical location, current battery level, and operational status. This data is encoded and uploaded to the system's cloud platform, providing structured input for the state modeling module. The information acquisition module specifically includes: A location information acquisition unit, deployed on the vehicle terminal, uses a GNSS module to acquire the vehicle's two-dimensional or three-dimensional spatial coordinate information in real time. In one possible implementation, the sampling frequency is 1Hz, and the data error is controlled within ±3 meters. The acquisition results are expressed as follows: Where s represents the current scheduling solution; This represents the normalized evaluation value of the j-th objective function under the scheduling solution s; j is the index number of the objective function; w j The dynamic weight coefficient corresponding to the j-th objective function; F(s) represents the final comprehensive fitness function value.

[0024] The battery monitoring unit, connected to the vehicle's battery management system, is used to monitor the remaining battery power of the electric sanitation vehicle in real time. Battery power data is represented in SOC (State of Charge) format and mapped to discrete energy levels for use in the construction of the tensor discrete layer in subsequent modeling. Specifically, the SOC value e... n (t)∈[0,1], an energy level function can be defined: Among them, l n (t)∈{0,1,…,10} represents the power level, where level 0 indicates that the power is depleted and level 10 indicates that the power is fully charged.

[0025] The operation status detection unit, configured in the vehicle operation control system, is used to identify whether the vehicle is currently performing a task. Specifically, the status is detected by sensors on the operating equipment. For example, in a water spraying task, the water pump current is detected, and in a garbage collection task, the opening and closing action of the rear bin is used to determine the status. In some embodiments, this unit combines the vehicle speed to determine the operation status; if the speed remains at zero for an extended period and the operating equipment does not move, it is determined to be in a "waiting state".

[0026] Alternatively, the information acquisition module may also include a data packaging and encryption unit. This unit packages the acquired data into structured data frames and encodes them using AES symmetric encryption to improve data security during public network transmission. In one implementation, the uploaded data format is as follows: D n (t)=[P n (t),e n (t),l n (t),s n (t),ID n ]; Among them, D n (t) represents the overall data structure packet uploaded by vehicle n at time t; P n (t) represents the location information of vehicle n at time t; e n(t) represents the remaining battery level of vehicle n at the current moment; s n (t) represents the vehicle operation status code, l n (t) represents the current load status of vehicle n or load indicators such as cumulative distance and time of completed tasks; ID n It is a unique vehicle identification number.

[0027] Data uploads utilize a 4G wireless link. If the 4G signal is weak or unavailable, it can automatically switch to low-power wide-area communication methods such as NB-IoT or LTE Cat-M1. In some embodiments, the system supports a resume mechanism to ensure that previously unuploaded data is automatically retransmitted after a temporary connection loss is restored.

[0028] Specifically, before the scheduling cycle begins, the system cloud will call the data receiving thread at fixed intervals (default 30 seconds) to batch access the current state snapshots from each vehicle information collection module, which are used to initialize or update the scheduling state tensor model.

[0029] In one possible implementation, the system supports a signal receipt mechanism to confirm that the collected data has been successfully uploaded. If a vehicle loses packets continuously for more than a set threshold, it enters a "data loss warning mode" and the relevant vehicle will be temporarily removed from the scheduling list until communication is restored and status synchronization is completed.

[0030] In this embodiment, the state modeling module is mainly used to construct a four-dimensional scheduling tensor. This tensor uses vehicles, tasks, time slices, and power levels as dimensions to represent the Boolean relationships of task executability in the scheduling state space.

[0031] Generally, the real-time status information uploaded by a vehicle includes the vehicle number (v). n The system processes the following information in batches: current location coordinates, current battery level, and operational status indicator. This information is used to construct a four-dimensional tensor of dimensions N×M×K×L. Where N represents the total number of currently active vehicles in the system; M represents the total number of tasks to be assigned; K represents the number of future time slices, the length of each time slice is configurable, and the default value is 15 minutes; L represents the number of battery level intervals, which are positive integers and are usually 11 levels (0 to 10). Specifically, the Boolean values ​​in the tensor are jointly determined by a set of scheduling constraints. For example, the vehicle starts from its current position P. n (t) Departure to the mission starting point Q m Whether the required energy consumption is less than the maximum operating distance supported by the current power level; whether the task time window covers the scheduled time slice; whether the job status allows task switching, etc.

[0032] In one possible implementation, the system calculates the vehicle's completed task t. mRequired energy E n,m As follows: E n,m =η·D n,m ; Where η represents energy consumption per unit distance (unit: kWh / km), this value is related to vehicle type and workload; D n,m Indicates starting from the current position P n (t) to the task starting point Q m Path distance; E n,m The unit is kWh, and it is related to the vehicle's remaining energy. Compare. If Furthermore, the task allows the time range to cover the current time slice θ. k Then

[0033] As an alternative, to reduce storage and computational burden, tensors can be sparsely represented using CANDECOMP / PARAFAC(CP) decomposition after construction. Specifically, tensor x is decomposed into a weighted sum of tensors of a finite set of rank, i.e.: Where, λ r The weight of the r-th rank tensor; Each dimension of the vector represents a different vector. This indicates the outer product operation.

[0034] This decomposition method preserves the original feasibility distribution structure in the tensor while significantly reducing system memory usage and subsequent optimization computational overhead. The tensor construction and decomposition process employs an incremental update strategy, performing partial reconstruction only on vehicle or task indices with changing states, avoiding full recalculation and improving the system's real-time responsiveness.

[0035] In some embodiments, the state modeling module also dynamically adjusts the time slice dimension in the tensor dimension. For example, when scheduling demands change abruptly (such as a sudden large increase in tasks), the system will automatically extend the time slice prediction window K to enhance the foresight of task scheduling; conversely, it will compress the window to reduce computational latency.

[0036] In the state modeling module, task heat information can be attached as an annotation tensor. This structure can be used to reflect secondary information such as traffic density, obstacle factors, or historical completion rate within the task area, providing auxiliary priority indicators for the subsequent task matching stage.

[0037] For the task relationship modeling module, in this embodiment, the task relationship modeling module is used to construct a heterogeneous bilateral graph G = (U, E) based on the output of the tensor model, where: U = V∪T is the set of nodes, and V = {v1, v2, ..., v...}N Let T = {t1, t2, ..., t} represent the set of vehicle nodes. M} represents a set of task nodes; Let e ​​represent the set of edges, indicating the probability that a vehicle in a certain state has the ability to complete a certain task; each edge e n,m Each ∈E is accompanied by a scalar weight w m,n , used to represent vehicle v n Execute task t m The cost assessment value.

[0038] Generally, the system obtains tensors from the state modeling module. All of the above satisfy The index combination is used as the initial task feasibility pair. Subsequently, it is represented as a set of edges (v) in the graph structure using graph structure building units. n ,t m ).

[0039] Specifically, the edge weights are calculated using a weighted evaluation model, and the calculation formula is as follows: Among them, w m,n This represents the association weight between vehicle n and task m in the task-vehicle relationship graph; α, β, and γ are system-preset weighting factors; e n The remaining battery level of the vehicle; d m,n This represents the Euclidean distance between the vehicle's current position and the mission starting point, in meters, and is calculated as follows: Among them, (x n ,y n (x) represents the vehicle's location. m ,y m ) represents the coordinates of the task's starting point; ρ m The task priority coefficient is defined as follows: Where, Δt m The time difference is the current time and the task deadline, where ∈ is a small positive quantity to prevent the denominator from being zero.

[0040] In some embodiments, to adapt to situations where the urgency of tasks rapidly increases in a sudden scheduling environment, the system can adjust ρ m Introducing an exponential scaling function, such as ρ ′ m =exp(ρ m This is to enhance the expression of differences in task priorities.

[0041] As an alternative, the task relationship modeling module also supports dynamic reconstruction of the graph structure. When a vehicle's status is abnormal, a task is temporarily canceled, or its priority is adjusted, the system will perform a graph update operation, reconstructing only the affected subgraph portion, thus improving response speed and local scheduling agility.

[0042] In one possible implementation, the graph structure is encoded and stored using a sparse adjacency list to reduce space complexity in high-density task environments. The system also standardizes the edge weight matrix to ensure consistent scaling of all metrics, facilitating cost evaluation and ranking calculations by subsequent optimization modules.

[0043] Furthermore, node attribute extension fields can be introduced into the graph structure. For example, vehicle node v n It can carry attributes such as its maximum load capacity, vehicle size class, and operating zone; task node t m It can carry additional information such as task type identifier and estimated task time. This type of attribute field can be bound to the node index through hash mapping to support task adaptation and rule scheduling under subsequent policies.

[0044] For the scheduling optimization module, in this embodiment, the main function of the scheduling optimization module is: based on the constructed vehicle-task graph structure G=(U,E) and the scheduling state tensor A multi-objective scheduling problem is constructed, and the vehicle task matching result is solved by an evolutionary optimization algorithm.

[0045] Generally, the scheduling optimization module first defines a target vector: f(s)=[f1(s),f2(s),f3(s),f4(s)]; Where f1(s) represents the task completion time; f2(s) represents the vehicle's idle driving distance; f3(s) represents the efficiency of completing the task per unit of electricity; and f4(s) represents the fairness of task allocation.

[0046] To transform this multi-objective problem into a fitness form suitable for evolutionary optimization, the system constructs a weighted fitness function as follows: Where s represents the current scheduling solution; This is the normalized value of the total task completion time; This is the normalized value of the empty driving distance of the vehicle during the mission. This is the normalized value for the vehicle's energy utilization efficiency. Normalized values ​​for the fairness index of task allocation among vehicles; w j The dynamic weight coefficients corresponding to the j-th objective function.

[0047] Specifically, the empty driving distance f2(s) is usually based on the edge weights d in the graph structure. m,n The summation yields the total distance the vehicle travels from its current position to the mission starting point, while the unit energy efficiency f3(s) is the reciprocal weighted average of the energy consumed by the vehicle to complete the mission.

[0048] The task fairness index f4(s) can be measured by the variance of the number of tasks assigned to each vehicle. The smaller the variance, the more balanced the scheduling.

[0049] In one possible implementation, the scheduling optimization module employs the Differential Evolution (DE) algorithm to perform multiple rounds of population search on the aforementioned objective function. The initial population is a set of different vehicle task allocation schemes encoded, and the system performs mutation, crossover, and selection operations in each generation to find the optimal solution.

[0050] As an alternative, the system can also introduce a local perturbation mechanism to directionally perturb parts of the population when evolution stalls, in order to escape local optima. The perturbation method can be directionally adjusted based on the gradient of the current fitness function, improving the ability to explore the solution space.

[0051] After the optimization algorithm has iterated, the system selects the individual solution with the optimal fitness value as the scheduling result output. Each output solution clearly records the task-vehicle matching relationship, the estimated start and finish times, and the input parameters for the subsequent path planning module.

[0052] Furthermore, to support real-time scheduling capabilities, the system can set a maximum iteration time limit (e.g., 5 seconds). If the algorithm does not converge, it will automatically return the current optimal solution and stop the calculation, so as to ensure that the overall scheduling process can be completed within minutes.

[0053] In this embodiment, the instruction control module takes the task-vehicle matching result output by the optimization module as input, performs a series of parsing, path calculation and instruction encapsulation operations, and finally forms a task scheduling instruction package with the ability to be issued.

[0054] Generally, the scheduling result is represented as a matching set. Each pair (v) n ,t m ) indicates vehicle v n Task t assigned to perform m To further refine the execution details, the system needs to convert task matching pairs into structured control instructions, a process completed by the instruction decoding unit.

[0055] Specifically, the decoding unit converts each item in the scheduling result into a structured record containing the following fields: I n,m =[IDv ID t ,T s ,T e E min ,R n,m ]; Among them, ID v A unique vehicle identification number; ID t A unique identifier for the task; T s Suggest a start time for the task; T e E represents the estimated completion time of the task. min R represents the minimum remaining power threshold required for the task (in kWh); n,m Recommended route information for the vehicle to perform this task.

[0056] The path generation unit uses the current vehicle position P n (t) and the starting point Q of the task m Using the city traffic map as the starting and ending point, the shortest energy-consuming path is constructed. Path planning comprehensively considers various constraints such as city traffic maps, real-time congestion information, road closures, and regional traffic restrictions. In one possible implementation, path search uses an improved A / B algorithm. * The algorithm, the objective function can be expressed as: f(i) = g(i) + h(i); Where g(i) is the cumulative energy consumption from the starting point to the current node i; h(i) is the estimated minimum energy consumption from the current node to the target task point.

[0057] As an option, a backup path field can be included in the path; if the primary path becomes unavailable due to an unforeseen event, the vehicle can switch to the suboptimal path to continue execution. The generated path R n,m Represented as a sequence of nodes: R n,m ={p0,p1,…,p k}; Among them, each node p i It includes coordinate information, suggested road speeds, and predicted power consumption.

[0058] The instruction generation and distribution unit receives the above-mentioned structure record and encapsulates it into a control instruction packet. Generally, the instruction packet uses binary encoding for easy transmission over low-bandwidth links. The data structure also includes redundant fields for verification, priority identification, and task type marking. For example, cleaning tasks and watering tasks have different path width requirements during execution, which need to be explicitly marked using the task type field.

[0059] In one possible implementation, the system also signs and encrypts each issued instruction using the AES encryption algorithm to ensure that the dispatch instructions are not tampered with during transmission. After receiving the instruction, the vehicle terminal executes a confirmation receipt, with the following format: ACK n,m =[ID v ID t ,STATUS,T recv ], where ID v A unique vehicle identification number; ID t The task is uniquely identified by its ID; STATUS indicates the confirmation status; T recv This is the timestamp for receiving the instruction.

[0060] If the system does not receive a confirmation reply within the set time window, it will automatically resend the instruction or temporarily mark the vehicle as "unschedulable".

[0061] Furthermore, in some embodiments, the instruction control module also records all historically issued instructions and execution feedback data, forming a traceable scheduling log. This log data can be used for system performance evaluation and scheduling behavior analysis.

[0062] In this embodiment, the scheduling and monitoring module mainly includes a status display unit, an anomaly intervention unit, and an intervention feedback unit.

[0063] Typically, the status display unit visualizes data based on data transmitted from the vehicle terminal. The vehicle terminal periodically uploads operational parameters such as task execution status, current location, and battery level changes, and the system matches and updates these parameters in real time according to each vehicle's ID.

[0064] In one possible implementation, the interface supports a nested map display, with each vehicle marked with a number and colors used to distinguish different operational statuses. The task execution path, current location information, and remaining energy can be displayed floating on the interface.

[0065] As an option, the system interface also supports historical task trajectory review, statistics of completed tasks, and reminders for incomplete tasks. Some interface controls support batch operation functions, suitable for multi-vehicle linkage intervention in central control scenarios.

[0066] The anomaly intervention unit is used by dispatchers to manually reassign tasks in abnormal scenarios such as task conflicts, vehicle malfunctions, and route interruptions. The system provides a task reassignment operation interface, where dispatchers can select the vehicle to be replaced and the new target vehicle.

[0067] Specifically, this operation is triggered based on the following basic conditions: The current vehicle has timed out due to consecutive upload failures; Battery level is below the set threshold; The current task execution time has exceeded the expected range; Sudden road closures or changes to area restrictions.

[0068] In some embodiments, manual intervention can also perform task interruption, such as when a sudden cleaning task is issued, the scheduler can specify a priority to forcibly interrupt the original planned task. This mechanism usually comes with a forced switching flag so that subsequent modeling modules can identify the data source.

[0069] The intervention feedback unit is responsible for packaging all manual operation instructions and sending them back to the state modeling module and the task relationship modeling module. The returned data structure is as follows: F op =[ID v ID t ,T op ,op type ,op result ]; Among them, ID v A unique vehicle identification number; ID t A unique identifier for the task; T op For operation timestamp; op type For operation type identification (redispatch, insert, cancel, etc.); op result This indicates the status of the operation execution result.

[0070] This feedback information is used for local tensor updates in the state modeling module and for subgraph structure pruning and edge reconstruction in the task relationship modeling module, thus forming a complete scheduling logic closed loop.

[0071] In some embodiments, the system may set a limit on the number of interventions or configure operation channels based on permission levels to prevent unauthorized interventions from affecting normal scheduling.

[0072] The intelligent dispatching method for electric sanitation vehicles described below can be referred to in correspondence with the intelligent dispatching system for electric sanitation vehicles described above.

[0073] Please see the appendix Figure 2 The present invention also provides an intelligent dispatching method for electric sanitation vehicles, comprising the following steps: Obtain the location, battery level, and operational status information of each electric sanitation vehicle; Based on the information, a multidimensional data model is constructed to describe the feasibility of scheduling. Based on the data model, construct the association structure between vehicles and tasks; Perform scheduling strategy optimization calculations to generate vehicle task allocation schemes; The scheduling instructions are generated according to the proposed scheme and sent to the target vehicle; The scheduling model and associated structure are updated based on vehicle feedback and human intervention results during the scheduling process for the next round of scheduling optimization.

[0074] The method in this embodiment can be used to execute the above system embodiment, and its principle and technical effect are similar, so it will not be described again here.

[0075] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Identical components are represented by the same reference numerals. Therefore, all equivalent changes made to the structure, shape, and principle of this application should be included within the scope of protection of this application.

Claims

1. An intelligent dispatching system for electric sanitation vehicles, characterized in that, Includes the following modules: The information acquisition module is used to acquire the operating status information of the electric sanitation vehicle and upload the operating status information to the system platform; The state modeling module is used to construct a multi-dimensional data model to describe the vehicle scheduling state based on the operational state information. The task relationship modeling module is used to construct the association structure between vehicles and tasks based on the scheduling status data in the multidimensional data model. The scheduling optimization module is used to perform scheduling strategy optimization calculations based on the multidimensional data model and the associated structure, and generate vehicle task allocation results. The instruction control module is used to generate vehicle dispatch instructions based on the vehicle task allocation results, and to send the dispatch instructions to the target vehicles to drive them to perform tasks; The scheduling and monitoring module is used to display vehicle status and task execution status, and to receive manual intervention operations. The intervention operations are fed back to the status modeling module and the task relationship modeling module to update the scheduling input data.

2. The intelligent dispatching system for electric sanitation vehicles according to claim 1, characterized in that, The information collection module includes: The location acquisition unit is used to obtain the vehicle's location coordinates in real time via a GPS device; The power monitoring unit is used to acquire the current battery power of the vehicle at a frequency of 1Hz and map it to an energy level. The status upload unit is used to encrypt and upload the location coordinates, power level and operation identifier to the dispatch system platform via 4G communication.

3. The intelligent dispatching system for electric sanitation vehicles according to claim 1, characterized in that, The state modeling module includes: Multidimensional tensor building blocks are used to construct tensors in four dimensions: number of vehicles, number of tasks, number of time slices, and number of battery levels. Where N represents the total number of currently active vehicles in the system, M represents the total number of tasks to be assigned, K represents the number of future time slices, and L represents the number of battery level ranges; A sparse compression unit is used to compress the tensor into a weighted sum of several rank-one tensors using the CANDECOMP / PARAFAC decomposition method. The tensor state update unit is used to update the tensor element values ​​based on real-time vehicle feedback data to reflect the current scheduling feasibility.

4. The intelligent dispatching system for electric sanitation vehicles according to claim 1, characterized in that, The task relationship modeling module includes: Graph structure building unit, used to build task-vehicle relationship graph based on the executability of vehicle tasks in tensors; The edge weight calculation unit is used to assign a weight to each edge in the graph; The graph reconstruction unit is used to perform subgraph pruning and edge weight reconstruction to locally update the graph structure when a vehicle fails or a task is interrupted.

5. The intelligent dispatching system for electric sanitation vehicles according to claim 1, characterized in that, The scheduling optimization module includes: A multi-objective modeling unit is used to define task completion time, vehicle idle distance, power utilization, and task allocation fairness as scheduling optimization objectives; An optimized execution unit is used to generate Pareto optimal scheduling solutions through evolutionary optimization algorithms; The weighted adaptive unit is used to dynamically adjust the weights of each objective function based on the tensor feedback state, forming an adaptive fitness function, which is: Where s represents the current scheduling solution; This represents the normalized evaluation value of the j-th objective function under the scheduling solution s; j is the index number of the objective function; w j The dynamic weight coefficient corresponding to the j-th objective function; F(s) represents the final comprehensive fitness function value.

6. The intelligent dispatching system for electric sanitation vehicles according to claim 1, characterized in that, The instruction control module includes: The decoding unit is used to parse the scheduling scheme generated by the optimization module into a set of vehicle-task mappings. The route generation unit is used to generate the optimal driving route for vehicles based on the current traffic conditions. The instruction generation and distribution unit is used to package task, route and power limit data into scheduling instructions and send them to the target vehicle.

7. The intelligent dispatching system for electric sanitation vehicles according to claim 1, characterized in that, The scheduling and monitoring module includes: The status display unit is used to display the current task status, battery status and location information of each vehicle on the interface; An anomaly intervention unit is used to allow manual reassignment of tasks when scheduling anomalies or unexpected tasks occur; The intervention feedback unit is used to synchronously feed back the results of manual operations to the tensor modeling module and the task graph module to update the scheduling input status.

8. The intelligent dispatching system for electric sanitation vehicles according to claim 4, characterized in that, The construction task-vehicle relationship diagram is calculated according to the following formula: Among them, w m,n This represents the association weight between vehicle n and task m in the task-vehicle relationship graph; α, β, and γ are system-preset weighting factors; d m,n Indicates the Euclidean distance between the vehicle and the mission; e n The remaining battery level of the vehicle; ρ m This represents the task priority level coefficient.

9. The intelligent dispatching system for electric sanitation vehicles according to claim 5, characterized in that, The fitness function is based on the normalization function of each optimization objective. With adaptive weights w j The construction, in the form of: Where s represents the current scheduling solution; This is the normalized value of the total task completion time; This is the normalized value of the empty driving distance of the vehicle during the mission. This is the normalized value for the vehicle's energy utilization efficiency. Normalized values ​​for the fairness index of task allocation among vehicles; w j The dynamic weight coefficients corresponding to the j-th objective function.

10. A method for intelligent dispatching of electric sanitation vehicles, applied to an intelligent dispatching system for electric sanitation vehicles as described in any one of claims 1-9, characterized in that, Includes the following steps: Obtain the location, battery level, and operational status information of each electric sanitation vehicle; Based on the information, a multidimensional data model is constructed to describe the feasibility of scheduling. Based on the data model, construct the association structure between vehicles and tasks; Perform scheduling strategy optimization calculations to generate vehicle task allocation schemes; The scheduling instructions are generated according to the proposed scheme and sent to the target vehicle; The scheduling model and associated structure are updated based on vehicle feedback and human intervention results during the scheduling process for the next round of scheduling optimization.

Images