Unmanned vehicle remote non-inductive scheduling and multi-vehicle cluster charging collaborative management system

Abstract

The invention relates to the technical field of unmanned vehicle scheduling and charging management, and discloses an unmanned vehicle remote non-inductive scheduling and multi-vehicle cluster charging cooperative management system, which comprises a cluster state sensing module, a multi-vehicle cluster charging cooperative management module, an unmanned vehicle remote non-inductive scheduling module and a multi-vehicle cluster charging cooperative management module, the scheduling feature analysis module performs modeling analysis on the two types of flows to generate feature sets; the collaborative decision-making module is used for matching scheduling strategy levels based on strategy matrix fusion features; the charging strategy generation module calls a charging strategy according to the grade; and the central processing unit verifies and generates a final instruction set. The system is further provided with a communication relay interface, a charging pile dispatching module and the like, and data transmission, charging optimization, instruction execution and data archiving are achieved. The system improves unmanned vehicle cluster scheduling and charging cooperation efficiency, is suitable for large-scale management, and has the advantages of being high in real-time performance, accurate in decision making and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned vehicle scheduling and charging management, in particular to an unmanned vehicle remote non-inductive scheduling and multi-vehicle cluster charging collaborative management system. BACKGROUND

[0002] With the rapid development of intelligent transportation technology, unmanned vehicle clusters are increasingly widely used in fields such as logistics distribution and intelligent travel. However, current unmanned vehicle cluster management faces core problems such as low scheduling efficiency and insufficient energy supply collaboration, which seriously restricts the feasibility and economy of large-scale application of unmanned vehicles. From the perspective of scheduling, traditional scheduling systems are mostly based on single vehicle state or static path planning, and lack the ability to dynamically model the overall space-time relationship of the cluster. For example, in the multi-vehicle collaborative operation scenario, the path topological relationship between vehicles (such as real-time position, speed vector, heading angle, etc.) presents complex space-time evolution characteristics, and traditional methods are difficult to capture path conflicts or inefficient routing problems in real time during cluster operation, resulting in phenomena such as vehicle congestion and repeated routes during task execution, which significantly reduces the efficiency of cluster operation. At the same time, the existing scheduling system manages the vehicle energy consumption in a rough way, and can only realize simple threshold judgment based on the remaining power, and cannot deeply analyze the dynamic decay characteristics in the energy consumption flow, such as the capacity decay caused by the state of health of the battery and the influence of instantaneous load fluctuation on the endurance, which makes the scheduling decision and the actual energy demand of the vehicle disconnected, and easily causes problems such as vehicle stop due to insufficient power and unreasonable allocation of charging resources.

[0003] In terms of charging management, multi-vehicle cluster charging faces the dual challenges of low utilization rate of charging facilities and poor coordination of charging task scheduling. Traditional charging strategies are often independent of the scheduling system and only trigger charging based on the current power of the vehicle, without effective linkage with the scheduling strategy. For example, when multiple unmanned vehicles need to be charged at the same time, the lack of comprehensive consideration of factors such as charging facility capacity and vehicle scheduling priority easily leads to problems such as charging pile resource occupation and long charging queue, which prolongs the non-operation time of the vehicle and reduces the overall operational efficiency of the cluster. In addition, the existing charging management system lacks dynamic sensing capability for the topological structure and real-time state of the charging facility (such as available pile position, charging power, etc.), making it difficult to achieve optimal matching of charging tasks and charging pile resources, further exacerbating the inefficiency of the energy supply process.

[0004] At the system integration and data processing level, the unmanned vehicle cluster involves the collection, transmission and processing of multi-source heterogeneous data, including vehicle state data from different manufacturers and communication protocol differences of the Internet of Vehicles platform. Traditional systems lack efficient data classification and analysis mechanisms, resulting in high data processing delay and poor compatibility, making it difficult to meet the needs of real-time scheduling and charging coordination of the unmanned vehicle cluster. For example, the basic running data (such as position and speed) and auxiliary running data (such as battery temperature and motor load) in the heterogeneous data frame cannot be effectively separated and classified, making the subsequent feature analysis and decision-making process inaccurate due to data mixing, affecting the overall performance of the system. SUMMARY

[0005] The purpose of the present application is to provide an unmanned vehicle remote non-inductive scheduling and multi-vehicle cluster charging coordination management system to solve the problems raised in the background art.

[0006] To achieve the above purpose, the present application provides the following technical solution: an unmanned vehicle remote non-inductive scheduling and multi-vehicle cluster charging coordination management system, the system comprising: A cluster state perception module is used to collect real-time running state data of the unmanned vehicle cluster, and the running state data is divided into path topology flow and energy consumption flow based on a preset data classification rule; A scheduling feature analysis module is used to model the path topology flow in terms of space-time relationship to generate a first scheduling feature set, and to analyze the energy consumption flow in terms of dynamic attenuation to generate a second scheduling feature set; A collaborative decision-making module is used to match the first and second scheduling feature sets to the corresponding scheduling strategy level after feature fusion according to a preset scheduling strategy matrix, and the scheduling strategy level is used as the scheduling basis for the current cluster; A charging strategy generation module is used to call a target charging strategy from a preset charging scheme library based on the scheduling strategy level, and the target charging strategy is used as the charging control parameter for the current cluster; A central processing unit is used to send the running state data to the scheduling feature analysis module, and the first and second scheduling feature sets to the collaborative decision-making module, and is also used to perform logical verification on the scheduling basis and the charging control parameter to generate a final scheduling instruction set.

[0007] Preferably, the dynamic attenuation analysis of the energy consumption flow by the scheduling feature analysis module comprises: The continuous time series data in the energy consumption flow is divided into an instantaneous load group and a steady-state load group, and a residual endurance prediction is performed on the instantaneous load group based on a preset energy attenuation model to generate an energy feature set; performing electrochemical state segmentation processing on the battery health data in the energy consumption flow, extracting capacity attenuation characteristics of each state interval, and constructing an attenuation atlas; correlating the energy feature set with the attenuation atlas in multiple dimensions to generate the second scheduling feature set.

[0008] Preferably, the preset data classification rule includes a basic operation set and an auxiliary operation set; the basic operation set includes position coordinate identification, speed vector identification, and heading angle identification; the auxiliary operation set includes battery temperature identification and motor load identification, and each identification corresponds to an independent data processing channel.

[0009] Preferably, the system further includes a communication relay interface for realizing protocol adaptation of the cluster state perception module, the scheduling feature analysis module, the collaborative decision module, and the charging strategy generation module with a vehicle networking platform, respectively; The cluster state perception module divides operation state data based on the preset data classification rule, including: receiving a heterogeneous data frame from the vehicle networking platform through the communication relay interface, and matching a header field of the heterogeneous data frame according to the identification in the basic operation set to separate a basic state section; traversing an extension field of the heterogeneous data frame according to the identification in the auxiliary operation set to extract an auxiliary state section; aligning the basic state section and the auxiliary state section according to timestamps, and respectively writing them into a path topology storage area and an energy consumption buffer area.

[0010] Preferably, when the preset scheduling strategy matrix adopts a static partition model, the scheduling strategy level is a discrete mapping result of a weighted fusion value of the first scheduling feature set and the second scheduling feature set; When the preset scheduling strategy matrix adopts a dynamic partition model, the scheduling strategy level is a continuous decision vector obtained by dynamically dividing a collaborative analysis result of the first scheduling feature set and the second scheduling feature set through a reinforcement learning algorithm.

[0011] Preferably, the system further includes a charging pile scheduling module connected with the central processing unit, and the charging pile scheduling module is connected with a charging facility database through the communication relay interface; The charging pile scheduling module is configured to filter an available pile list from the charging facility database according to a charging pile scheduling requirement in the final scheduling instruction set, and generate a charging task sequence to optimize an energy supply process.

[0012] Preferably, the charging pile scheduling module generating the charging task sequence includes: Loading a regional charging station topology map, and locating the physical coordinate node of each charging station in the available charging station list in the topology map; Calculating the optimal service path from the deployment location of each charging pile to the target dispatch vehicle based on a path planning algorithm, and prioritizing the list of available charging piles according to charging efficiency; The optimal service path and the priority ranking are integrated into the topology map to generate a visual charging task sequence.

[0013] Preferably, when the central processing unit performs logical verification on the scheduling basis and charging control parameters, a two-layer verification mode of conflict detection mechanism and resource constraint mechanism is adopted. The conflict detection mechanism is used to identify task conflicts, and the resource constraint mechanism is used to solve the capacity limitations of charging facilities.

[0014] Preferably, the system also includes an instruction execution module connected to the central processing unit, and the instruction execution module is used to convert the final scheduling instruction set into vehicle control instructions, and send the vehicle control instructions to the target unmanned vehicle through the communication relay interface to perform the scheduling task.

[0015] Preferably, the system also includes a state storage module connected to the central processing unit, and the state storage module is used to archive the operating status data, the first scheduling feature set, the second scheduling feature set, the scheduling strategy level and the final scheduling instruction set, and generate a cluster scheduling record chain according to the timeline.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The swarm state perception module uses pre-set data classification rules to precisely divide operational status data into path topology flows and energy consumption flows, enabling multi-dimensional, real-time monitoring of the operational status of swarms of unmanned vehicles. The independent data processing channels for the basic and auxiliary operational sets ensure the efficient separation and simultaneous processing of fundamental state data (such as position coordinates and velocity vectors) and auxiliary state data (such as battery temperature and motor load). This provides structured, low-latency data support for subsequent feature analysis, addressing the processing inefficiency associated with data clutter in traditional systems.

[0017] The scheduling feature analysis module models the spatiotemporal relationship of the path topology flow to generate the first scheduling feature set, which can dynamically capture path conflicts and collaboration opportunities between vehicles in the cluster. For example, by analyzing the spatiotemporal evolution of the speed vectors and heading angles of multiple vehicles, it can predict the potential congestion risks at path intersections in advance, providing an accurate basis for optimizing scheduling strategies. At the same time, the dynamic attenuation analysis of the energy consumption flow, by dividing the instantaneous load group and the steady-state load group and combining it with the energy attenuation model, achieves an accurate prediction of the vehicle's remaining range; the electrochemical state segmentation and attenuation map construction of the battery health data deeply reveals the inherent laws of battery capacity attenuation. The multi-dimensional correlation between the two generates the second scheduling feature set, which enables the system to fully grasp the vehicle's energy status and avoid stagnation or waste of charging resources due to extensive power management.

[0018] The collaborative decision-making module, based on a preset scheduling strategy matrix, achieves intelligent matching of feature fusion and scheduling strategy levels. The discrete mapping of weighted fusion values ​​in the static partitioning model and the dynamic partitioning through reinforcement learning in the dynamic partitioning model enable the system to flexibly respond to diverse operational scenarios. For example, in static environments with urgent tasks and stable scenarios, the static partitioning model can quickly generate deterministic scheduling strategies. In complex, dynamically changing scenarios, the reinforcement learning algorithm can optimize scheduling decisions through continuous learning, significantly improving the cluster's adaptability and scheduling accuracy in dynamic environments.

[0019] The charging strategy generation module calls the target charging strategy from the preset charging solution library based on the scheduling strategy level, achieving deep coordination between scheduling and charging. By combining scheduling priorities with vehicle energy requirements, it ensures that vehicles with high-priority tasks have priority access to charging resources, avoiding task delays caused by long charging queues. At the same time, the charging pile scheduling module achieves the optimal match between charging tasks and charging pile resources by loading the regional charging station topology, calculating the optimal service path, and prioritizing it, effectively improving the utilization rate of charging facilities and shortening the vehicle's non-operating time. For example, by using a path planning algorithm to match each vehicle to be charged with the nearest available charging pile and generate a visual task sequence, it can reduce the vehicle's idle driving distance and reduce the overall energy consumption of the cluster.

[0020] The central processing unit's dual-layer logic verification mechanism (conflict detection and resource constraints) ensures the reliability of dispatch instructions and charging control parameters. The conflict detection mechanism promptly identifies task conflicts, preventing multiple vehicles from simultaneously occupying the same resource or path. The resource constraint mechanism optimizes the capacity allocation of charging facilities, addressing the common charging pile resource overload problem in traditional systems and ensuring the stability and security of cluster operations. The design of the instruction execution module and state storage module further enhances the system's closed-loop control and data traceability capabilities. By converting dispatch instructions into vehicle control commands and archiving operating data in real time, they provide strong support for continuous system optimization and troubleshooting. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a working principle diagram of the unmanned vehicle remote sensorless dispatch and multi-vehicle cluster charging collaborative management system according to the present invention; Figure 2 This is the working principle diagram of the scheduling feature analysis module; Figure 3 This is the working principle diagram of the cluster status perception module; Figure 4 This is the working principle diagram of the charging pile scheduling module. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] See also Figures 1-4 The present invention relates to a collaborative management system for remote, seamless dispatching of unmanned vehicles and charging of multiple vehicles in a cluster. The system comprises a cluster status perception module, a dispatching feature analysis module, a collaborative decision-making module, a charging strategy generation module, and a central processing unit. The following details the overall implementation scheme and specific embodiments of the invention, combining the functions and relationships of each module.

[0024] During system operation, the cluster state perception module first collects real-time operational status data from the swarm of autonomous vehicles. Based on pre-set data classification rules, this data is divided into a path topology flow and an energy consumption flow. The path topology flow primarily reflects the spatial position and motion status of the autonomous vehicles, while the energy consumption flow focuses on the vehicles' energy usage and battery status.

[0025] After the collected data is transmitted to the central processing unit, the central processing unit sends the operating status data to the scheduling feature analysis module. The scheduling feature analysis module models the spatiotemporal relationships of the path topology flow to generate a first scheduling feature set containing features such as spatial location associations and time series trajectories. Simultaneously, it analyzes the dynamic attenuation of the energy consumption flow to generate a second scheduling feature set reflecting energy consumption trends and battery health status.

[0026] The central processing unit transmits the first and second scheduling feature sets to the collaborative decision-making module, which fuses the two feature sets based on a preset scheduling strategy matrix. The fused features are matched to the corresponding scheduling strategy level, which serves as the basis for the current cluster's scheduling. Subsequently, the charging strategy generation module, based on the scheduling strategy level, retrieves the target charging strategy from a preset charging solution library and uses it as the cluster's charging control parameters.

[0027] The central processing unit needs to perform logical verification on the scheduling basis and charging control parameters. After the verification is passed, the final scheduling instruction set is generated to guide the coordinated management of scheduling and charging of the unmanned vehicle cluster.

[0028] The present invention will be further described below in conjunction with Examples 1 to 5: Example 1: The system's scheduling feature analysis module analyzes the dynamic attenuation of energy consumption flows as follows: First, the continuous time-series data in the energy consumption flow is structured and divided into transient and steady-state load groups. The transient load group primarily includes load data that exhibits dramatic fluctuations over short periods of time. For example, during dynamic operating scenarios like starting and accelerating, emergency braking, and climbing and overcoming obstacles, parameters like battery output power and motor current can experience significant instantaneous fluctuations. This type of data is categorized as transient. The steady-state load group, on the other hand, covers load data from relatively stable operating conditions like constant speed driving and straight-line cruising. During these conditions, parameters like battery output power and motor speed remain relatively stable, with minimal fluctuations.

[0029] After the data is grouped, the remaining battery life of the instantaneous load group is predicted based on a preset energy decay model. This preset energy decay model is an analytical model constructed based on the battery's physical characteristics and historical operating data. The model construction process comprehensively considers multiple factors, including the magnitude and duration of the instantaneous load fluctuation, the current remaining battery charge, and ambient temperature. Specifically, for each data set in the instantaneous load group, the model first identifies the peak and valley values ​​of the load fluctuation, as well as its duration. For example, during an acceleration, if the motor current suddenly rises from a baseline value to a peak value and then falls back for 5 seconds, the model records the peak current value, the rate of increase, and the duration. Then, based on the battery's energy capacity characteristic curve, the model calculates the energy consumption for this instantaneous load scenario. The remaining battery life prediction is then subtracted from the initial remaining battery life value to obtain the remaining battery life prediction. The model also converts the remaining battery life prediction value into a remaining battery life estimate based on the autonomous vehicle's current average speed, thereby generating an energy signature set that includes the remaining battery life prediction value, the remaining battery life estimate, and the instantaneous load characteristic parameters.

[0030] Perform electrochemical state segmentation on the battery health data in the energy consumption flow. Battery health data includes key parameters such as battery voltage, internal resistance, number of cycles, and self-discharge rate. Changes in these parameters can directly reflect the health status and decay trend of the battery. During the electrochemical state segmentation process, the battery state is first divided into multiple continuous state intervals based on the historical changes in battery voltage and internal resistance, combined with the technical standards and industry experience provided by the battery manufacturer. For example, it can be divided into a healthy state interval (battery voltage is higher than the standard threshold and internal resistance is lower than the critical value, and the number of cycles is small), a sub-healthy state interval (battery voltage is close to the lower limit of the standard threshold, internal resistance is slightly increased, and the number of cycles reaches a certain range), a decay state interval (battery voltage is lower than the standard threshold, internal resistance is significantly increased, and the number of cycles exceeds the service life threshold), etc.

[0031] For each state interval, the corresponding capacity decay characteristics are extracted. In the healthy state interval, the main extractions include the capacity decline rate per unit time (such as a monthly capacity decay rate of less than 0.5%), the stability index of the decay rate, etc.; in the sub-healthy state interval, the focus is on the changing trend of the capacity decay rate (such as whether the decay rate shows a gradually accelerating trend), the correlation between internal resistance and capacity decay, and other characteristics; in the decay state interval, the focus is on analyzing the mutation point of capacity decay (such as a sudden and significant drop in capacity after a certain charge and discharge), the critical value of the remaining available capacity, and other characteristics. By quantitatively analyzing and statistically analyzing the capacity decay characteristics of each state interval, a decay map is constructed. The decay map is presented in the form of a chart, with the horizontal axis as the time axis and the vertical axis as the degree of capacity decay. Different color areas represent different battery state intervals, and the curve represents the trend of capacity decay over time, which intuitively shows the decay characteristics and laws of the battery at different stages.

[0032] The energy feature set is correlated with the decay profile in multiple dimensions. In the time dimension, the relationship between the predicted remaining battery life and the battery capacity decay trend is analyzed. For example, if the predicted remaining battery life decreases continuously over a certain period of time, the decay profile can be used to determine whether the battery has entered a sub-healthy state and whether the capacity decay rate has accelerated during that period. In the load type dimension, the correlation between the energy feature set and the decay profile is examined under different transient load and steady-state load scenarios. For example, for unmanned vehicles that frequently experience high-intensity transient loads, the predicted remaining battery life in the energy feature set may decrease more rapidly, and the corresponding capacity decay curve in the decay profile has a steeper slope, indicating that high-intensity transient loads accelerate battery degradation. Through multi-dimensional correlation analysis, a second scheduling feature set is generated that couples energy consumption patterns with battery health status. This feature set not only covers single energy consumption data or battery health data but also reveals the inherent connection between the two. This provides the collaborative decision-making module with more comprehensive and in-depth energy status information, enabling the system to formulate more targeted scheduling strategies and charging plans based on the actual energy consumption and battery health of the unmanned vehicle.

[0033] Example 2: The system's preset data classification rules include a basic operation set and an auxiliary operation set. The basic operation set covers position coordinate identifiers, velocity vector identifiers, and heading angle identifiers, which correspond to the core operating status parameters such as the unmanned vehicle's spatial position, movement speed, and driving direction; the auxiliary operation set includes battery temperature identifiers and motor load identifiers, which are used to supplement the status information reflecting the vehicle's energy system and power system, and each identifier corresponds to an independent data processing channel to ensure the independence and accuracy of different types of data during collection, transmission, and processing, and to avoid data confusion or interference.

[0034] The system features a communication relay interface, which facilitates protocol adaptation between the cluster status perception module, scheduling feature analysis module, collaborative decision-making module, and charging strategy generation module, and the connected vehicle platform. Specifically, the communication relay interface performs protocol conversion and format parsing on heterogeneous data transmitted by the connected vehicle platform, enabling each module to identify and process the diverse data from the connected vehicle platform.

[0035] When the cluster status perception module classifies operational status data based on preset data classification rules, it first receives heterogeneous data frames from the connected vehicle platform via the communication relay interface. These heterogeneous data frames may originate from different brands and models of unmanned vehicles and various sensor devices, and their data formats, encoding methods, and protocol standards may vary. For example, some data frames are encapsulated in JSON format, while others use binary format. Furthermore, different manufacturers may have different field names and storage locations for the same type of data.

[0036] The cluster state perception module matches the header fields of heterogeneous data frames based on the identifiers in the basic operation set to separate the basic state segments. The header field typically contains metadata such as the data frame type identifier, version information, and sending device identifier. By matching the field names or field positions corresponding to the position coordinate identifier, velocity vector identifier, and heading angle identifier, the basic state segment containing the unmanned vehicle's current longitude, latitude, and altitude, as well as basic operating state data such as speed magnitude, velocity direction vector, and heading angle value, can be extracted from the data frame header. For example, if the "Pos" field in a data frame header corresponds to the position coordinate, the "Spd" field corresponds to the velocity vector, and the "Hdg" field corresponds to the heading angle, then the basic state segment data can be obtained by parsing these fields.

[0037] After completing the separation of the basic state segments, the cluster state perception module traverses the extended fields of the heterogeneous data frame according to the identifiers in the auxiliary operation set to extract the auxiliary state segments. The extended field is located after the data frame header and is used to store non-core data that has auxiliary analysis value for system operation. The fields corresponding to the battery temperature identifier and the motor load identifier may be distributed in different positions in the extended field. By traversing all fields in the extended field and matching the fields related to battery temperature (such as "BatTemp") and motor load (such as "MotLoad"), the system extracts auxiliary state data such as the current battery temperature value, temperature change rate, motor load percentage, load fluctuation amplitude, etc., to form an auxiliary state segment.

[0038] Since there may be a time difference between the data collection moments of the basic state segment and the auxiliary state segment (for example, the location coordinate data is collected by the GPS module once per second, while the battery temperature data is collected by the temperature sensor every 0.5 seconds), in order to ensure the time consistency of the data, the cluster state perception module needs to align the basic state segment and the auxiliary state segment by timestamp. The timestamp is the precise time mark of the data collection moment, usually in milliseconds. During the alignment process, the system first sorts the basic state segment and the auxiliary state segment by timestamp from small to large, and then for each basic state segment data point, finds the data point with the closest timestamp in the auxiliary state segment. Through methods such as linear interpolation or nearest neighbor interpolation, the data with different sampling frequencies are unified onto the same time axis, so that the same time point has complete data of both the basic operating state and the auxiliary operating state.

[0039] After completing the time alignment, the basic state segment data is written into the path topology storage area, which is used to store data related to the spatial position and motion trajectory of the unmanned vehicle. It supports chronological query of the position coordinates, speed and heading angle of the unmanned vehicle at different times, and provides a data basis for the subsequent scheduling feature analysis module to perform spatiotemporal relationship modeling; the auxiliary state segment data is written into the energy consumption buffer, which is used to temporarily store data related to energy consumption and battery status, so as to facilitate the scheduling feature analysis module to perform dynamic attenuation analysis on the energy consumption flow and generate energy feature sets and attenuation maps.

[0040] Through the above process, the cluster state perception module realizes the classification processing and structured storage of heterogeneous data, ensuring the accuracy, completeness and time consistency of path topology flow and energy consumption flow data, and providing reliable data support for subsequent links such as scheduling feature analysis, collaborative decision-making and charging strategy generation of the entire system.

[0041] Example 3: The preset scheduling strategy matrix serves as the core analysis tool for the collaborative decision-making module. Its operational mechanism differs depending on the model type (static partitioning model or dynamic partitioning model). The following details the specific processing logic, feature fusion process, and scheduling strategy level generation mechanism for both models.

[0042] When the preset scheduling strategy matrix uses a static partitioning model, a weighted fusion process is first performed on the first and second scheduling feature sets. The first scheduling feature set is generated by the scheduling feature analysis module through spatiotemporal modeling of the path topology flow. It includes spatiotemporal features such as the spatial location distribution of the AV cluster, trajectory correlation, driving speed coordination, and heading angle consistency. The second scheduling feature set is derived through dynamic decay analysis of energy consumption flows. It includes energy-related features such as remaining battery life prediction, battery capacity decay rate, and energy consumption patterns under different load scenarios. The system assigns different weight coefficients to the two feature sets based on the scheduling objective priority (e.g., prioritizing transportation efficiency or optimizing energy utilization). For example, if the current scheduling task focuses on improving transportation efficiency, the first scheduling feature set is assigned a higher weight (e.g., 0.6) and the second scheduling feature set is assigned a lower weight (e.g., 0.4). If the system prioritizes cluster charging needs, the weight ratio is adjusted to 0.4 for the first scheduling feature set and 0.6 for the second scheduling feature set.

[0043] The weighted fusion value is calculated as follows: Each characteristic parameter in the first scheduling feature set is normalized (e.g., to the range [0, 1]), multiplied by the corresponding weight coefficient, and summed to obtain a spatiotemporal feature composite value. Similarly, the second scheduling feature set is normalized and weighted to obtain an energy feature composite value. Finally, the spatiotemporal and energy feature composite values ​​are added together to form a weighted fusion value. This fusion value is a continuous value and needs to be converted to a scheduling policy level through discrete mapping. The static partitioning model predefines several discrete level intervals, such as "low," "medium," and "high," each corresponding to a fixed numerical range. Assuming the weighted fusion value range is [0, 2], the "low" level can be set to [0, 0.6), the "medium" level to [0.6, 1.4), and the "high" level to [1.4, 2]. The collaborative decision-making module maps the weighted fusion value to the corresponding scheduling policy level based on the range in which it falls. Each level is associated with a preset scheduling policy. For example, a "low" level strategy may only perform simple path optimization to avoid collisions between vehicles in the cluster; a "medium" level strategy needs to consider both path optimization and energy balance to ensure that some low-battery vehicles are given priority to drive to the charging area; a "high" level strategy initiates multi-vehicle collaborative scheduling, combines the distribution of charging facilities with traffic conditions, and generates a globally optimal scheduling plan, including vehicle platooning and pre-allocation of charging tasks.

[0044] When the preset scheduling strategy matrix adopts a dynamic partitioning model, the system uses a reinforcement learning algorithm to collaboratively analyze the first and second scheduling feature sets to achieve dynamic classification of scheduling strategy levels. The reinforcement learning algorithm optimizes cluster scheduling efficiency, energy utilization, and task completion time. By constructing a state space, an action space, and a reward function, the model learns the optimal decision-making strategy through continuous interaction with the environment. The state space consists of all feature parameters of the first and second scheduling feature sets, for example, N spatiotemporal feature parameters and M energy feature parameters, forming an N+M-dimensional state vector. The action space is a set of available scheduling strategies, such as route replanning, charging priority adjustment, and vehicle group scheduling. The reward function is designed based on the alignment of the current scheduling results with the goals. For example, successfully completing an emergency task is rewarded with +10 points, while interrupting the task due to insufficient energy is penalized with -20 points.

[0045] During the dynamic analysis process, the reinforcement learning algorithm first converts the first and second scheduling feature sets acquired in real time into a state vector. Then, based on the current policy (e.g., the ε-greedy policy), it selects an action (i.e., a scheduling policy) to execute. It then observes environmental feedback after execution (e.g., whether the vehicle's driving path is optimized or whether the charging task completion rate has improved). It then calculates a reward value based on a reward function and updates policy parameters (e.g., a Q-value table or neural network weights). Through continuous iterative training, the algorithm gradually learns the optimal actions corresponding to different state vectors, forming a dynamic classification rule for scheduling policies. At this point, the scheduling policy level is no longer a discrete interval, but a continuous decision vector, with each dimension corresponding to the execution probability or parameter value of a scheduling policy. For example, the decision vector may include a path optimization coefficient, a charging priority coefficient, a vehicle coordination coefficient, and other factors. The value range of each coefficient is dynamically adjusted based on the state vector, enabling refined control of the scheduling policy.

[0046] The advantage of the dynamic partitioning model is its ability to adapt to changes in the cluster's operating state in real time. For example, when a sudden traffic jam blocks some vehicles' routes, the trajectory correlation and speed coordination features in the first scheduling feature set will change significantly. The reinforcement learning algorithm, sensing these changes, automatically adjusts the path optimization coefficients in the decision vector, generates a new scheduling strategy level, and triggers path replanning. For another example, when the battery levels of multiple vehicles are detected to be below a critical value simultaneously, the remaining range prediction and capacity decay rate features in the second scheduling feature set will highlight energy shortages. The algorithm then increases the weight of the charging priority coefficient in the decision vector, prioritizing charging resources for vehicles with low battery levels, thus preventing the overall cluster scheduling from failing due to insufficient energy.

[0047] The choice between the two models depends on the actual application scenario. The static partitioning model is suitable for scenarios with relatively stable operating environments and fixed scheduling objectives, and features low computational complexity and fast decision-making. The dynamic partitioning model is more suitable for complex and changing environments, and can improve the adaptability and optimization of scheduling strategies through continuous learning, but it requires high computing resources and a certain amount of training time.

[0048] Example 4: The system includes a charging pile scheduling module connected to the central processing unit (CPU). This module establishes a connection to the charging facility database via a communication relay interface. The charging facility database stores detailed information about all charging stations in the area, including their geographic location, the number of charging piles, the charging power type of each pile (e.g., DC fast charging, AC slow charging), current usage status (idle, charging, faulty), and reservation status. The core function of the charging pile scheduling module is to filter available resources from the charging facility database and optimize the energy replenishment process based on the charging pile scheduling requirements in the final scheduling instruction set generated by the CPU. The specific implementation is as follows: After the central processing unit generates the final dispatch instruction set containing the charging station scheduling requirements, the charging station scheduling module first parses the requirements to determine information such as the number of unmanned vehicles requiring charging, their current locations, their estimated arrival times at the charging station, and the required charging power level. The charging station scheduling module then uses this information to filter eligible available charging stations from the charging facility database, creating a list of available charging stations. These criteria include: the station is currently idle or can complete the current charging task before the vehicle's estimated arrival time; the charging power level at the station meets the vehicle's requirements; and the distance between the charging station where the station is located and the vehicle's current location is within a reasonable range.

[0049] After generating a list of available charging stations, the charging station scheduling module loads a regional charging station topology map. This topology map, built using a geographic information system (GIS), visually displays the spatial distribution of charging stations within the region, the physical layout of charging stations within the station (such as station numbering, spacing between adjacent stations, and charging space dimensions), and the surrounding road network (including entrance and exit locations and connecting road grades). In the topology map, each charging station corresponds to a physical coordinate node. Node information includes the station's latitude and longitude coordinates, the unique identifier of the charging station where it is located, and the distance from the station's entrance and exit.

[0050] Based on the path planning algorithm, the optimal service path from the deployment location of each charging pile to the target dispatch vehicle is calculated. The path planning algorithm uses the improved Dijkstra algorithm or The algorithm aims to minimize driving distance or travel time, while also taking into account factors such as real-time traffic data (such as road congestion and speed limit information) and road type (such as highways and urban roads). For each available pile position, the algorithm uses the physical coordinate node of the pile position as the starting point and the current position or expected stop position of the target vehicle as the end point. It searches for one or more feasible paths and calculates the comprehensive cost value of each path based on preset weight factors (such as 70% for distance and 30% for time). The path with the smallest comprehensive cost value is selected as the optimal service path. Assume that the calculation formula for the comprehensive cost value of the path is:

[0051] in, Represents the comprehensive cost value of the path, Indicates the path distance (unit: kilometers), Indicates the estimated travel time of the route (unit: minutes). and is the weight coefficient and satisfies (For example, Indicates that the distance weight accounts for 70%, Indicates that the time weight accounts for 30%).

[0052] After completing the calculation of the optimal service path, the charging pile scheduling module needs to prioritize the list of available pile positions according to charging efficiency. The evaluation indicators of charging efficiency include the charging power of the pile position (the higher the power, the higher the efficiency), the historical average service time of the pile position (including vehicle entry and exit time, charging connection time, etc.), the utilization rate of the charging station where the pile position is located (excessive utilization may cause queuing), etc. For example, the charging power of a DC fast charging pile is usually above 60kW, and that of an AC slow charging pile is about 7kW. Therefore, under the same other conditions, a DC fast charging pile has a higher priority than an AC slow charging pile. During the sorting process, the system calculates a charging efficiency index for each pile position. The higher the index, the higher the priority. The calculation of the charging efficiency index can take into account factors such as charging power and historical service time. For example:

[0053] in, represents the charging efficiency index, is the maximum charging power of the charging station (unit: kW), is the average charging power of charging piles in the area, is the standard service time (preset value, such as 30 minutes), = is the historical average service time of the charging station (unit: minutes). According to this formula, the charging station with higher charging power and shorter historical service time has a higher charging efficiency index and a higher priority ranking.

[0054] Finally, the charging station scheduling module integrates the optimal service path and priority ranking results into the regional charging station topology map, generating a visual sequence of charging tasks. During this integration process, each available charging station node on the topology map is labeled with its priority number (such as "1," "2," or "3"), and the corresponding optimal service path is drawn with a line of different colors. The thickness of the line indicates the overall cost of the path (thick lines indicate higher costs, thin lines indicate lower costs). The topology map also displays information such as the target vehicle number, estimated charging start time, and estimated charging completion time corresponding to each charging station, creating an intuitive visual interface. This interface allows system administrators or dispatchers to view the allocation of charging tasks, path planning results, and schedules in real time, facilitating manual intervention or adjustments. It also provides the automated scheduling system with a clear execution instruction sequence to ensure that charging tasks are efficiently executed according to priority and path planning.

[0055] Through the above process, the charging pile scheduling module realizes the full process management from available pile position screening, route planning, priority sorting to visual task sequence generation, effectively optimizing the energy supply process of the unmanned vehicle cluster and improving the utilization rate and scheduling efficiency of charging facilities.

[0056] Example 5: The system is equipped with an instruction execution module and a state storage module connected to the central processing unit. The two modules respectively undertake the core functions of scheduling instruction execution and running data archiving. The following is a detailed description of the module architecture, data interaction process and specific implementation logic.

[0057] The command execution module serves as the control bridge between the system and the unmanned vehicles. Its core function is to convert the final dispatch instruction set generated by the central processing unit into vehicle control commands that the unmanned vehicles can recognize and execute, and to ensure reliable transmission of these commands via the communication relay interface. The final dispatch instruction set, based on the dispatch strategy level output by the collaborative decision-making module and the charging control parameters determined by the charging strategy generation module, forms a complete control instruction sequence after logical verification by the central processing unit. These instructions cover specific commands for each unmanned vehicle in the swarm, such as route adjustment, speed control, charging start / stop operations, and task priority switching. The command execution module first performs protocol parsing on the final dispatch instruction set. Based on the communication protocol standards for different brands or models of unmanned vehicles, the abstract dispatch instructions are converted into specific control signal formats. For example, for unmanned vehicles using the CAN bus protocol, path planning instructions must be converted into CAN data frames containing parameters such as target coordinates and speed thresholds. For intelligent connected vehicles supporting Ethernet communication, these instructions must be encapsulated into TCP / IP protocol packets.

[0058] After completing the protocol conversion, the command execution module sends vehicle control commands to the target unmanned vehicle via the communication relay interface. The communication relay interface plays a dual role in this process: on the one hand, it performs power amplification and anti-interference encoding on the control signals output by the command execution module to ensure signal stability during wireless transmission; on the other hand, it establishes a two-way communication link with the Internet of Vehicles platform to monitor the command transmission status in real time. If there is a signal interruption or command loss, the retransmission mechanism is automatically triggered. For example, if an unmanned vehicle does not provide a command reception confirmation within a preset time, the communication relay interface will resend the command until confirmation is received, or mark it as an abnormal command and feed it back to the central processing unit for troubleshooting.

[0059] The command execution module also has a command execution status tracking function. After sending a control command, the module continuously monitors the status feedback data returned by the target unmanned vehicle via the communication relay interface, including the command reception status (success / failure), execution progress (such as the proportion of the distance traveled to the total route length), and execution results (such as whether the charging connection was successful or whether the task was completed on time). This feedback data is transmitted to the central processing unit in real time and serves as the basis for the system to dynamically adjust the scheduling strategy. For example, if a vehicle fails to follow the planned route due to an unexpected obstacle, the path deviation information feedback from the command execution module can prompt the central processing unit to trigger the collaborative decision-making module to regenerate the scheduling strategy and adjust the driving paths of surrounding vehicles to avoid congestion.

[0060] The state storage module, serving as the system's data hub, is responsible for archiving and managing critical data generated during operation, ensuring data integrity, traceability, and long-term availability. This module archives data including: raw operational status data collected by the cluster state perception module (such as position coordinates, velocity vector, and battery temperature); the first and second scheduling feature sets generated by the scheduling feature analysis module (such as spatiotemporal relationship modeling results, energy feature sets, and attenuation maps); the scheduling strategy level output by the collaborative decision-making module; the final scheduling instruction set generated by the central processing unit; and instruction execution status feedback data recorded by the instruction execution module.

[0061] The state storage module utilizes a time-series database architecture, organizing and storing data in chronological order to generate a cluster scheduling record chain. The time-series database offers efficient time series data processing capabilities, supporting fast querying, inserting, and updating data by timestamp. It also provides data compression and backup mechanisms to conserve storage resources and prevent data loss. During data archiving, each data entry includes a unique timestamp, data type identifier, source module identifier, and the data content. For example, a raw operating status data entry might include the following: timestamp "2025-06-10 14:30:05.001," data type "Basic Operating Status," source module "Cluster Status Perception Module," and content "Longitude 116.4810, Latitude 39.9219, Speed ​​20 km / h, Heading Angle 180°, Battery Temperature 25°C, Motor Load 30%."

[0062] The cluster scheduling record chain uses the timeline as an index, concatenating different types of data entries in order of occurrence to form a complete record of the scheduling process. This structured storage method facilitates subsequent system monitoring, fault tracing, and performance analysis. For example, when a scheduling task encounters an anomaly (such as a vehicle failing to arrive at the charging station on time), the state storage module can be used to query the vehicle's operating status data, corresponding scheduling feature set, scheduling strategy level, and instruction execution records in chronological order, trace the execution status of each link, and locate the root cause of the problem (such as whether the route deviates due to an error in the path planning algorithm, or whether the charging station fails to charge in time due to a failure). In addition, the state storage module also supports data visualization, which can generate trend charts (such as battery capacity decay trend, cluster driving speed distribution), heat maps (such as charging station usage frequency distribution), etc. from archived data to provide data support for system optimization.

[0063] To ensure data security, the state storage module employs access control and data encryption technologies. Access control mechanisms set different data access permissions based on user roles (e.g., system administrator, general operator) to prevent sensitive data leaks. Data encryption technology encrypts stored raw data and feature sets, ensuring data security during storage and transmission. Furthermore, the module regularly performs integrity checks on archived data, using hash checks and other methods to detect data tampering or corruption, ensuring data credibility.

[0064] The instruction execution module and the state storage module realize data interaction and cooperative work through the central processing unit. After completing instruction sending and state feedback collection, the instruction execution module transmits relevant records to the state storage module for archiving; the state storage module provides the instruction execution module with historical instruction record query function, supports analysis and reuse of instruction execution modes of similar tasks. For example, when scheduling the same type of unmanned vehicle to execute similar tasks again, the optimal instruction sequence in the historical record can be referred to, the scheduling decision time is shortened, and the system operation efficiency is improved.

[0065] Through the cooperative operation of the instruction execution module and the state storage module, the system realizes full-process closed-loop management from scheduling instruction generation to execution and then to data archiving, ensures the efficiency, reliability and traceability of the unmanned vehicle cluster scheduling and charging cooperative management, and provides solid technical support for unmanned vehicle cluster management in complex scenarios.

[0066] It should be noted that, in this article, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device.

[0067] Although the embodiments of the present application have been shown and described, it can be understood by those of ordinary skill in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. Unmanned vehicle remote sensorless dispatching and multi-vehicle cluster charging collaborative management system, characterized by: include: A cluster state perception module is used to collect the operating state data of the unmanned vehicle cluster in real time and divide the operating state data into path topology flow and energy consumption flow based on preset data classification rules; a scheduling feature analysis module, configured to perform spatiotemporal relationship modeling on the path topology flow to generate a first scheduling feature set, and perform dynamic attenuation analysis on the energy consumption flow to generate a second scheduling feature set; A collaborative decision-making module is configured to fuse the first and second scheduling feature sets according to a preset scheduling strategy matrix, match them to a corresponding scheduling strategy level, and use the scheduling strategy level as the scheduling basis for the current cluster; A charging strategy generation module is configured to call a target charging strategy in a preset charging solution library based on the scheduling strategy level, and use the target charging strategy as a charging control parameter of the current cluster; A central processing unit is used to send the operating status data to the scheduling feature analysis module, send the first scheduling feature set and the second scheduling feature set to the collaborative decision-making module, and perform logical verification on the scheduling basis and the charging control parameters to generate a final scheduling instruction set.

2. The unmanned vehicle remote sensorless dispatch and multi-vehicle cluster charging collaborative management system according to claim 1 is characterized in that: The scheduling feature analysis module performs dynamic attenuation analysis on the energy consumption flow, including: Dividing the continuous time series data in the energy consumption flow into a transient load group and a steady-state load group, and performing a remaining endurance prediction on the transient load group based on a preset energy decay model to generate an energy feature set; Performing electrochemical state segmentation processing on the battery health data in the energy consumption flow, extracting the capacity decay characteristics of each state interval and constructing a decay map; The energy feature set is multi-dimensionally associated with the attenuation map to generate the second scheduling feature set.

3. The unmanned vehicle remote sensorless dispatching and multi-vehicle cluster charging collaborative management system according to claim 1 is characterized in that: The preset data classification rules include a basic operation set and an auxiliary operation set; the basic operation set includes a position coordinate identifier, a velocity vector identifier, and a heading angle identifier; the auxiliary operation set includes a battery temperature identifier and a motor load identifier, and each identifier corresponds to an independent data processing channel.

4. The unmanned vehicle remote sensorless dispatching and multi-vehicle cluster charging collaborative management system according to claim 3 is characterized in that: The system further includes a communication relay interface, which is used to implement protocol adaptation of the cluster state perception module, the scheduling feature analysis module, the collaborative decision module, and the charging strategy generation module to the Internet of Vehicles platform; The cluster status perception module divides the operation status data based on the preset data classification rule, including: Receiving heterogeneous data frames from the Internet of Vehicles platform through the communication relay interface, and matching header fields of the heterogeneous data frames according to identifiers in the basic operation set to separate basic state segments; Traversing the extended field of the heterogeneous data frame according to the identifier in the auxiliary operation set to extract the auxiliary state segment; The basic state segment and the auxiliary state segment are aligned according to timestamps and then written into the path topology storage area and the energy consumption buffer area respectively.

5. The unmanned vehicle remote sensorless dispatch and multi-vehicle cluster charging collaborative management system according to claim 1 is characterized in that: When the preset scheduling strategy matrix adopts a static partition model, the scheduling strategy level is a discrete mapping result of a weighted fusion value of the first scheduling feature set and the second scheduling feature set; When the preset scheduling strategy matrix adopts a dynamic partition model, the scheduling strategy level is a continuous decision vector that dynamically divides the collaborative analysis results of the first scheduling feature set and the second scheduling feature set through a reinforcement learning algorithm.

6. The unmanned vehicle remote sensorless dispatch and multi-vehicle cluster charging collaborative management system according to claim 1 is characterized in that: Also included is a charging pile scheduling module connected to the central processing unit, the charging pile scheduling module being connected to the charging facility database via the communication relay interface; The charging pile scheduling module is used to filter the available pile position list from the charging facility database according to the charging pile scheduling requirements in the final scheduling instruction set, and generate a charging task sequence to optimize the energy replenishment process.

7. The unmanned vehicle remote sensorless dispatching and multi-vehicle cluster charging collaborative management system according to claim 6 is characterized in that: The charging pile scheduling module generates a charging task sequence including: Loading a regional charging station topology map, and locating the physical coordinate node of each charging station in the available charging station list in the topology map; Calculating the optimal service path from the deployment location of each charging pile to the target dispatch vehicle based on a path planning algorithm, and prioritizing the list of available charging piles according to charging efficiency; The optimal service path and the priority ranking are integrated into the topology map to generate a visual charging task sequence.

8. The unmanned vehicle remote sensorless dispatching and multi-vehicle cluster charging collaborative management system according to claim 1 is characterized in that: When the central processing unit performs logical verification on the scheduling basis and charging control parameters, a two-layer verification mode of conflict detection mechanism and resource constraint mechanism is adopted. The conflict detection mechanism is used to identify task conflicts, and the resource constraint mechanism is used to solve the capacity limitations of charging facilities.

9. The unmanned vehicle remote sensorless dispatching and multi-vehicle cluster charging collaborative management system according to claim 1 is characterized in that: It also includes an instruction execution module connected to the central processing unit, which is used to convert the final scheduling instruction set into vehicle control instructions and send the vehicle control instructions to the target unmanned vehicle through the communication relay interface to perform the scheduling task.

10. The unmanned vehicle remote sensorless dispatch and multi-vehicle cluster charging collaborative management system according to claim 1 is characterized in that: It also includes a state storage module connected to the central processing unit, which is used to archive the operating state data, the first scheduling feature set, the second scheduling feature set, the scheduling strategy level and the final scheduling instruction set, and generate a cluster scheduling record chain according to the timeline.

Images