Method and system for evaluating the operating risk of unmanned delivery vehicles

Abstract

The present invention discloses a method for evaluating the operating risk of unmanned delivery vehicles, which includes the following steps: operating area division: dividing the operating area of ​​the unmanned delivery vehicle into multiple honeycomb plots to ensure that there is no overlap between the plots and that the operating area is fully covered; key node determination; plot congestion probability calculation: for each honeycomb plot, the truncated mean of the congestion probability of all key nodes in it is calculated, and the maximum value is selected as the overall congestion probability of the plot; distribution resource allocation: according to the congestion probability of each honeycomb plot, the appropriate amount of unmanned delivery vehicles in the plot and the number of safety personnel responsible for handling potential accidents are determined. The present invention can identify the key nodes of risk in urban traffic operation, propose a regional risk assessment method for the operation of unmanned delivery vehicles from the two dimensions of congestion and accidents, and then form unmanned delivery vehicle deployment strategy constraints.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation and logistics technology, and more specifically, to a system for real-time evaluation and early warning of potential risks of unmanned delivery vehicles during operation. More specifically, the present invention relates to a method and system for evaluating the operating risks of unmanned delivery vehicles. Background Art

[0002] In the field of intelligent transportation and logistics distribution, unmanned delivery vehicles are gradually becoming an important tool for improving delivery efficiency and reducing operating costs. In complex urban traffic scenarios, in order to ensure that unmanned delivery vehicles can operate efficiently and safely, it is crucial to scientifically and reasonably plan and manage the operation plan. First of all, urban transportation is a highly open and complex system. As a new participant in urban transportation, the operation plan of unmanned delivery vehicles needs to match the time-varying characteristics of urban transportation operation; secondly, the technology of unmanned delivery vehicles is still in the process of continuous development and improvement, and its operation plan should be able to dynamically match the technical capabilities of unmanned delivery vehicles; finally, the operation plan of unmanned delivery vehicles should be spatially independent, so that it can adapt to the deployment and operation strategy planning needs of unmanned vehicles in different operating areas.

[0003] In order to solve the above problems, the present invention proposes a method for evaluating the operating risk of unmanned delivery vehicles, as a quantitative evaluation method for the feasibility of the operating plan of unmanned delivery vehicles after they are put on open roads. In the evaluation method of the operating risk of unmanned delivery vehicles of the present invention, the dynamic characteristics of urban traffic operation are integrated, the technical capability model of unmanned delivery vehicles is defined, and an innovative operating area division method is proposed based on the deployment characteristics of unmanned delivery vehicles to solve the problem of spatial independence of the evaluation method.

[0004] In actual operation, it is necessary to set up several stations to meet the needs of unmanned delivery vehicles for refueling and cargo distribution during deployment. The distribution area of ​​a single station is a circular area with the station as the center and the distribution radius as the radius. The distribution ranges of multiple stations cross-cover the entire unmanned delivery vehicle operation area. According to the above deployment characteristics of unmanned delivery vehicles, the present invention proposes a honeycomb-based operation area division method, that is, the operation areas of unmanned delivery vehicles at different stations are divided into plots in the shape of honeycombs.

[0005] The honeycomb-shaped plot division method can ensure that each plot does not interfere with each other and can also cover the entire operation area, thereby realizing refined management and optimization of the operation area. On this basis, the present invention further proposes a method for identifying and calculating traffic congestion points in each honeycomb plot.

[0006] The method first determines the attribute data of each road in the honeycomb plot based on the static road network structure, such as road width, number of lanes, traffic light settings, etc. Then, the congestion probability of each road is determined in combination with dynamic traffic operation characteristics, such as traffic volume and speed. By setting a congestion threshold, the roads with congestion data exceeding the threshold are classified and merged according to the attribute data and topological relationship to obtain a series of key nodes. These key nodes are the concentrated embodiment of traffic congestion and are of great significance for evaluating the congestion situation of the entire honeycomb plot.

[0007] In order to more accurately reflect the congestion status of the honeycomb plot, the present invention also proposes a method of calculating the truncated mean of the congestion probability of each key node and taking the maximum value as the congestion probability of the plot. This method takes advantage of the short board effect and can ensure that the plot with the least congestion impact is focused on and managed.

[0008] Finally, according to the congestion probability of the honeycomb plot, the present invention proposes a strategy to determine the number of unmanned delivery vehicles and the number of safety personnel. By dynamically adjusting the number of unmanned delivery vehicles and the number of safety personnel, congestion of the plot can be avoided to the greatest extent, the delivery efficiency can be improved, and the safety and reliability of the delivery process can be ensured.

[0009] In summary, the method proposed in the present invention of dividing plots in honeycomb shapes, identifying key nodes and calculating congestion probability, as well as the strategy of determining the number of unmanned delivery vehicles and the number of safety personnel based on the congestion probability, provides a new idea and method for the operation and management of unmanned delivery vehicles. Summary of the invention

[0010] The present invention provides a method and system for evaluating the operating risks of unmanned delivery vehicles, which can identify key risk nodes in urban traffic operation, propose regional risk assessment of unmanned delivery vehicle operation from the two dimensions of congestion and accidents, and then form strategic constraints for the deployment of unmanned delivery vehicles.

[0011] In order to achieve these purposes and other advantages according to the present invention, a method for evaluating the operating risk of an unmanned delivery vehicle is provided, the method comprising the following steps:

[0012] Operation area division: The operation area of ​​the unmanned delivery vehicle is divided into multiple honeycomb-shaped plots to ensure that there is no overlap between the plots and that the operation area is fully covered;

[0013] Key nodes determined:

[0014] Step a: According to the static road network structure, obtain and determine the attribute data of each road in each honeycomb plot;

[0015] Step b: Evaluate and determine the congestion probability of each road in each cell block based on the dynamic traffic operation characteristics;

[0016] Step c: Set a congestion threshold ρ, classify the roads whose congestion probability exceeds the threshold ρ according to their attribute data, and merge the same type of roads according to the topological relationship between the roads to form one or more key node sets;

[0017] Calculation of plot congestion probability: For each cellular plot, calculate the truncated mean of the congestion probabilities of all key nodes in it, and select the maximum value as the overall congestion probability of the plot;

[0018] Distribution resource allocation: Based on the congestion probability of each honeycomb plot, determine the appropriate number of unmanned delivery vehicles in the plot and the number of safety personnel responsible for handling potential accidents.

[0019] Preferably, the step of determining the number of unmanned delivery vehicles and the number of safety personnel according to the congestion probability of the honeycomb plot is achieved through the following micro-simulation experiment:

[0020] Honeycomb plot simulation modeling: Build a honeycomb plot simulation model that reflects the actual road network, traffic flow, and traffic rules;

[0021] Unmanned delivery vehicle technical capability simulation:

[0022] In the simulation model, the technical capabilities of the unmanned delivery vehicle are simulated, and the level of technical capabilities is adjusted by setting the state transition probability;

[0023] When an unmanned delivery vehicle encounters a risk scenario, it has a certain probability of transitioning from a risky state to a normal state or an abnormal state, depending on its technical capabilities, i.e., the state transition probability. The higher the technical capabilities, the greater the probability of transitioning to a normal state.

[0024] Simulation experiment running:

[0025] In the constructed cellular plot simulation model, traffic flow, road attributes, and traffic signal data are input;

[0026] Based on the technical capability simulation of the unmanned delivery vehicle, run simulation experiments to observe and record the key indicators of the unmanned delivery vehicle's operating efficiency and accident rate under different congestion probabilities;

[0027] Resource allocation decisions:

[0028] According to the results of simulation experiments, the impact of different congestion probabilities on the operating efficiency and safety of unmanned delivery vehicles is analyzed;

[0029] Based on the analysis results, determine the appropriate number of unmanned delivery vehicles and the number of safety officers required in honeycomb plots with different congestion probabilities.

[0030] Preferably, in the unmanned delivery vehicle technical capability simulation, the probability of state transition is determined by the unmanned delivery vehicle behavior model obtained by the following steps:

[0031] Abnormal scenario monitoring: From the perspective of the unmanned delivery vehicle, the system monitors its own operating parameters and the operating parameters of potential interactive traffic participants in the surrounding environment in real time. Based on these parameters, it detects whether it is in the set abnormal scenario. The behavioral data in the abnormal scenario includes: data on sudden deceleration, emergency stop, and collision of the unmanned delivery vehicle;

[0032] Abnormal state variable setting: In simulation modeling, the position coordinates (x, y), speed (v_x, v_y) and acceleration (a_x, a_y) of the unmanned delivery vehicle under abnormal conditions are measured as basic state variables to form a state vector S(t) = [x(t), y(t), v_x(t), v_y(t), a_x(t), a_y(t)], where t represents the time;

[0033] Expand abnormal state variables: further include the distance to surrounding vehicles or obstacles (d_{front}, d_{rear}), the vehicle in which the vehicle is located (lane(t), traffic signals and road attributes to form an expanded state vector S(t) = [x(t), y(t), v_x(t), v_y(t), a_x(t), a_y(t), lane(t), d_{front}(t), d_{rear}(t)];

[0034] Discretization of abnormal state variables: Discretize the state variables with continuous values ​​to facilitate the subsequent construction of the state transfer matrix;

[0035] Establishment of state transfer matrix: Based on the collected operating parameters of the unmanned delivery vehicle, construct the preliminary state transfer probability matrix elements, and combine the kinematic principles and traffic rules of the unmanned delivery vehicle to improve the state transfer matrix;

[0036] Iterative calculation and matrix correction: Use the established state transfer matrix to perform iterative calculations in time series, gradually derive the operating state of the unmanned delivery vehicle at subsequent moments, use the reconstructed operating state as a new state transfer matrix element, recalculate the state transfer probability, and repeat this process to correct and optimize the state transfer matrix until a stable and accurate state transfer probability is obtained;

[0037] Technical capability simulation: Based on the revised state transition matrix, simulate the state transition probability of the unmanned delivery vehicle at different technical capability levels.

[0038] Preferably, the step of calculating the traffic congestion point in each honeycomb plot and taking it as the key node is implemented according to dynamic traffic information, and specifically includes the following steps:

[0039] Road network definition: Set the road network of each honeycomb plot in the area, including the set of all nodes and the set of all links. The attribute information collected from the links includes the number of lanes and whether non-motorized vehicle lanes are set;

[0040] Dynamic traffic information update: Update the congestion information of each road section in the road network at a set frequency, including the average driving speed and congestion status of the road section;

[0041] Evaluation of road section congestion probability: For each road section in the road network, its congestion probability is calculated based on the traffic information updated during the evaluation period. Specifically, if the road section updates traffic information N times during the evaluation period, and if M of them show congestion, then the congestion probability P of the road section is M / N;

[0042] Screening of candidate road sections for key nodes: setting a congestion probability threshold ρ, and taking road sections with congestion probability exceeding the threshold ρ as candidate road sections for key nodes;

[0043] Spatial clustering operation: Perform spatial clustering on the candidate road segment set, and merge the candidate road segments with the same road characteristics into key nodes according to the topological relationship. The key nodes include key roads and key intersections.

[0044] Key node feature extraction: Extract feature information of key nodes. The node features of key roads include the number of lanes and whether vehicles and non-motorized vehicles are mixed. The node features of key intersections include the type of intersection, whether there are traffic lights, and the number of lanes of the upstream and downstream roads of the intersection and whether vehicles and non-motorized vehicles are mixed.

[0045] Definition of key nodes and calculation of congestion probability: A key node is defined as a structure containing a unique identifier, type, attribute information and road section information. For each key node, its congestion probability during the evaluation period is calculated. Specifically, suppose the key node contains M road sections, and the congestion probability of each road section is P_i. After removing extreme values, the congestion probability of the key node is the truncated mean of the congestion probabilities of each road section.

[0046] Preferably, a three-level assessment framework for the operational risk of unmanned delivery vehicles at the “key nodes of regional cellular plots” is constructed.

[0047] Regional level integration: Divide the entire service area of ​​the unmanned delivery vehicle into multiple independent areas;

[0048] Honeycomb plot definition: In each area, based on the operating characteristics and delivery needs of the unmanned delivery vehicle, an abstract definition of a hexagonal structure with the distribution center as the center and a reasonable delivery distance as the radius is used as a honeycomb plot;

[0049] Identification and evaluation of key nodes: Within each honeycomb, by calculating the congestion probability of the road section and performing spatial clustering operations, traffic congestion points or key locations are identified as key nodes. Key nodes include key roads and key intersections.

[0050] Among them, the highest traffic operation risk of the key nodes in the honeycomb plot represents the traffic operation risk level of the entire honeycomb plot; the highest traffic operation risk of the honeycomb plot represents the traffic operation risk level of the entire area.

[0051] Preferably, the method for determining the number of unmanned delivery vehicles and the number of safety personnel in a honeycomb plot comprehensively considers the risk integration of both safety and efficiency.

[0052] Comprehensive risk assessment model: A comprehensive risk assessment model is constructed, which integrates the efficiency risk and safety risk after the deployment of unmanned delivery vehicles. The calculation formula is as follows:

[0053] Comprehensive risk = α × efficiency risk + β × safety risk

[0054] Among them, α and β are the weighted coefficients of efficiency risk and safety risk respectively, and α+β=1.

[0055] Preferably, the congestion probability assessment under the maximum deployment volume: under the preset maximum deployment volume of unmanned delivery vehicles, the traffic operation conditions of key nodes are simulated and analyzed to assess the resulting congestion probability of key nodes;

[0056] Determination of the ratio of the number of safety officers to unmanned delivery vehicles: Based on the above congestion probability assessment results, a reasonable ratio of the number of safety officers to unmanned delivery vehicles is determined through quantitative analysis.

[0057] Preferably, the cellular area is set so that the travel range of the unmanned delivery vehicle is between 2 and 3.5 times the radius of the cellular with the cellular center as the center.

[0058] Preferably, the highest traffic operation risk of the key node in the honeycomb plot represents the traffic operation risk level of the entire honeycomb plot; the highest traffic operation risk of the honeycomb plot represents the traffic operation risk level of the entire area is achieved by the following steps:

[0059] Critical node assessment: used to determine the highest traffic operation risk of each critical node within a honeycomb plot, which represents the traffic operation risk level of the corresponding honeycomb plot;

[0060] Cell plot assessment: Receive data from the key node assessment module and determine the highest traffic operation risk of the cell plot based on the maximum value of the highest traffic operation risk of the key nodes in each cell plot;

[0061] Regional assessment: used to receive data from multiple cellular plot assessment modules and determine the traffic operation risk level of the entire area based on the maximum value of the highest traffic operation risk in all cellular plots.

[0062] A system for assessing the operating risks of unmanned delivery vehicles, the system comprising:

[0063] Operation area division module: used to divide the operation area of ​​the unmanned delivery vehicle into multiple honeycomb-shaped plots to ensure that there is no overlap between the plots and that the operation area is fully covered;

[0064] Key node determination module:

[0065] Attribute data acquisition submodule: According to the static road network structure, the attribute data of each road in each honeycomb plot is acquired and determined;

[0066] Congestion probability assessment submodule: evaluates and determines the congestion probability of each road in each cell plot based on dynamic traffic operation characteristics;

[0067] Key node set generation submodule: set a congestion threshold ρ, classify the roads whose congestion probability exceeds the threshold ρ according to their attribute data, and merge the same type of roads according to the topological relationship between the roads to form one or more key node sets;

[0068] Block congestion probability calculation module: For each cellular block, the truncated mean of the congestion probabilities of all key nodes in it is calculated, and the maximum value is selected as the overall congestion probability of the block;

[0069] Distribution resource allocation module: Based on the congestion probability of each honeycomb plot, determine the appropriate number of unmanned delivery vehicles in the plot and the number of safety personnel responsible for handling potential accidents.

[0070] The present invention has at least the following beneficial effects:

[0071] First, the present invention proposes a three-level assessment framework for the operation risk of unmanned delivery vehicles, namely "region-cellular structure-key nodes". By evaluating the congestion of road sections and intersections within the honeycomb structure, the key nodes of risk in urban traffic operation are automatically screened and identified. The static road network structure and dynamic traffic operation characteristics of the key nodes are referred to, and efficiency and safety are considered from the two dimensions of congestion and accidents. A regional risk assessment for the operation of unmanned delivery vehicles is proposed, thereby forming constraints on the deployment strategy of unmanned delivery vehicles.

[0072] Second, the present invention is based on the operating characteristics of unmanned delivery vehicles and independent honeycomb structures to flexibly adapt to the evaluation needs of different area sizes. The evaluation processes between different honeycombs are decoupled, thereby achieving the spatial independence of the algorithm. The comprehensive operating risk of the set area is obtained by integrating the comprehensive operating risks of the honeycomb structure, and the comprehensive operating risk of the honeycomb structure is obtained by the maximum comprehensive risk of the key nodes, ultimately realizing the formulation of rule plans after the unmanned delivery vehicles are deployed on open roads.

[0073] Third, this application can improve traffic management efficiency. Through the risk probability transfer matrix, it can more accurately predict traffic congestion and formulate effective management strategies. Ensuring traffic safety, objective evaluation of driving ability and characterization of delivery capabilities will help improve the operational safety of unmanned delivery vehicles. It will promote industry development, provide companies with standards for vehicle performance evaluation, and promote industry technological progress and industrial upgrading. It will enhance public trust and increase public trust and support for unmanned delivery vehicle technology through transparent risk assessment and management.

[0074] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 This is a schematic diagram of the three-level modeling of "region-cell-key node" of the present invention;

[0076] Figure 2 A schematic diagram of the key node risk model construction process of the present invention;

[0077] Figure 3 Modeling the operation process of the unmanned delivery vehicle of the present invention within a cell;

[0078] Figure 4 Schematic diagram of the meso-capability modeling method of the unmanned delivery vehicle of the present invention. DETAILED DESCRIPTION

[0079] The present invention is further described in detail below with reference to the description so that those skilled in the art can implement it accordingly.

[0080] It should be understood that the terms such as “having”, “including” and “comprising” used herein do not exclude the existence or addition of one or more other elements or combinations thereof.

[0081] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial sources unless otherwise specified; in the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "set" should be understood in a broad sense, for example, they can be fixedly connected, set, or detachably connected, set, or connected and set in one piece. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood in specific circumstances. The orientation or position relationship indicated by the terms "lateral", "longitudinal", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc. is based on the orientation or position relationship shown in the accompanying drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention.

[0082] like Figure 1-4 As shown, the present invention provides a method for evaluating the operating risk of an unmanned delivery vehicle, comprising:

[0083] 1. Operation area division:

[0084] Step description: First, the entire operating area of ​​the unmanned delivery vehicle needs to be carefully divided. This division is based on honeycomb plots to ensure that each plot can fully cover the operating area and there is no overlap between plots. The design of honeycomb plots is intended to simplify the complex road network structure and facilitate subsequent congestion analysis and resource allocation.

[0085] Implementation details: Use geographic information system (GIS) tools to draw a honeycomb plot map based on the actual operation area of ​​the unmanned delivery vehicle. Ensure that the size and shape of each plot are appropriately adjusted according to actual needs and road network structure to ensure the rationality and effectiveness of the division.

[0086] 2. Determination of key nodes:

[0087] Step a: Obtain road attribute data. Step description: According to the static road network structure, collect the attribute data of each road in each honeycomb plot. Implementation details: Attribute data includes but is not limited to road width, number of lanes, road type (such as main road, secondary road, branch road, etc.), speed limit, etc.

[0088] Obtain this data through field surveys, consulting transportation planning materials, or using GIS tools.

[0089] Step b: Evaluate the probability of road congestion. Step description: According to the dynamic traffic operation characteristics, evaluate the congestion probability of each road in each cell block. Implementation details: Use traffic flow monitoring data (such as vehicle passing speed, vehicle density, etc.) and traffic simulation models to calculate the congestion level of each road.

[0090] Congestion probability can be estimated based on historical data, real-time data and predictive models.

[0091] Step c: Form a set of key nodes. Step description: Set a congestion threshold ρ, classify the roads with congestion probability exceeding the threshold according to their attribute data, and merge the same type of roads according to the topological relationship between the roads to form one or more key node sets.

[0092] Implementation details: The congestion threshold ρ can be set according to the actual situation, for example, it can be set to a multiple of the historical average congestion probability. Road classification can be performed based on road attributes (such as width, number of lanes, whether there is mixed flow of people and vehicles, etc.) and congestion characteristics.

[0093] The topological relationship considers the connection relationship and relative position between roads to ensure that the merged key node set can accurately reflect the congestion situation.

[0094] 3. Calculation of land congestion probability:

[0095] Step description: For each cellular plot, calculate the truncated mean of the congestion probabilities of all key nodes within it, and select the maximum value as the overall congestion probability of the plot.

[0096] Implementation details: By calculating the congestion probability of each key node and then finding the truncated mean, the overall congestion situation in the plot can be obtained. The maximum value is selected as the overall congestion probability of the plot to ensure that the evaluation result is conservative so as to better deal with potential congestion risks.

[0097] 4. Distribution resource allocation:

[0098] Step description: Based on the congestion probability of each honeycomb plot, determine the appropriate number of unmanned delivery vehicles in the plot and the number of safety personnel responsible for handling potential accidents.

[0099] Implementation details: For plots with a higher probability of congestion, more unmanned delivery vehicles need to be deployed to meet delivery needs, and the number of safety personnel should be increased to ensure safety.

[0100] The number of delivery vehicles can be set based on the delivery demand, road capacity and congestion within the plot. The number of safety officers can be set based on factors such as traffic complexity within the plot and accident history.

[0101] Through the above specific implementation methods, the present invention provides a method for evaluating the operating risk of unmanned delivery vehicles. This method realizes a comprehensive evaluation and optimization of the operating risk of unmanned delivery vehicles through the steps of operating area division, key node determination, land congestion probability calculation and distribution resource allocation. This method not only improves the operating efficiency of unmanned delivery vehicles, but also ensures the safety and reliability of the distribution process.

[0102] With the popularity of unmanned delivery vehicles in urban traffic, their operational safety and efficiency have become the focus of attention from all walks of life. This paper aims to construct a comprehensive evaluation model, taking into account key events such as deceleration, emergency stop, and collision, and using the risk probability transfer matrix to predict the traffic congestion index under different risk scenarios and provide decision support. At the same time, the model will also objectively evaluate driving ability, characterize the delivery capabilities of vehicles on the market, and provide guidance for the healthy development of the industry.

[0103] Risk probability transfer matrix construction: event definition and classification. Deceleration: The vehicle reduces its speed due to an obstacle ahead or a change in traffic conditions. Emergency stop: The vehicle stops suddenly when encountering an emergency or serious traffic obstacle. Collision: The vehicle comes into physical contact with other traffic participants (including pedestrians, other vehicles, etc.).

[0104] Risk level classification: Based on the probability of an event occurring and its impact on traffic congestion, the risk is divided into three levels: low, medium and high.

[0105] Transition matrix construction: Based on historical data and simulation, a probability matrix of mutual transformation between different risk levels is constructed. For example, the probability of a low-risk state turning into a medium-risk state due to a deceleration event, or the probability of a high-risk state turning into a low-risk state due to effective measures.

[0106] Collision is a more serious situation among these problems and needs to be included in the accident dimension as the highest risk level, considering its direct impact on the traffic congestion index and subsequent processing costs.

[0107] Congestion index calculation and response strategy: Congestion index calculation. Combined with the risk probability transfer matrix, the congestion index in each honeycomb is calculated according to the current traffic conditions and the predicted future event development. The congestion index takes into account multiple factors such as vehicle density, driving speed, and event impact range.

[0108] Develop response strategies. Develop targeted response strategies based on the congestion index and risk level, such as adjusting the number of unmanned delivery vehicles, optimizing route planning, and strengthening traffic monitoring.

[0109] For high-risk areas or time periods, emergency measures are taken, such as temporarily closing some roads and increasing safety patrols. Objective evaluation of driving ability and characterization of delivery capabilities. Driving ability evaluation objectively evaluates driving ability by analyzing the vehicle's reaction speed, accuracy and stability when dealing with events such as deceleration, emergency stops and collisions. The evaluation results can be used as an important indicator for vehicle performance evaluation and provide a reference for vehicle selection.

[0110] The delivery capability characterization combines the driving ability evaluation results and vehicle technical parameters (such as cruising range, load capacity, etc.) to comprehensively characterize the delivery capabilities of vehicles on the market.

[0111] Provide data support for formulating relevant policies and optimizing vehicle configuration for enterprises.

[0112] The present invention also includes:

[0113] 1. Simulation modeling of honeycomb plots:

[0114] Step description: Use advanced simulation software (such as MATLAB / Simulink, SUMO, etc.) to build a cellular plot simulation model that reflects the actual road network, traffic flow, and traffic rules based on the actual operation area of ​​the unmanned delivery vehicle. The model should include detailed road structure, intersection layout, traffic signal control, vehicle driving rules, etc. to ensure the authenticity and accuracy of the simulation.

[0115] Implementation details: Collect and organize actual road network data, including road length, width, number of lanes, intersection type, signal light cycle, etc. According to traffic flow data, set parameters such as vehicle generation rate, driving speed, acceleration, etc. in the simulation model. Introduce traffic rules, such as priority rules and turning rules, to simulate the real traffic environment.

[0116] 2. Simulation of unmanned delivery vehicle technical capabilities:

[0117] Step description: In the simulation model, the technical capabilities of the unmanned delivery vehicle are simulated, and the level of technical capabilities is adjusted by setting the state transition probability. When the unmanned delivery vehicle encounters a risk scenario (such as deceleration, emergency stop, collision, etc.), according to its technical capabilities (i.e., state transition probability), there is a certain probability that it will transition from the risk state to the normal state or abnormal state. The higher the technical capability, the greater the probability of transitioning to the normal state.

[0118] Implementation details: Set the initial technical capability value of the unmanned delivery vehicle, which can be determined based on data provided by the vehicle manufacturer or actual test results.

[0119] According to the technical capability value, the state transition probability matrix is ​​set to simulate the behavioral response of the unmanned delivery vehicle in different risk scenarios. In the simulation experiment, the state transition probability matrix is ​​adjusted to observe the performance of the unmanned delivery vehicle under different technical capabilities.

[0120] 3. Simulation experiment operation:

[0121] Step description: In the constructed honeycomb plot simulation model, input data such as traffic flow, road attributes, and traffic signals. According to the technical capability simulation of the unmanned delivery vehicle, run the simulation experiment, observe and record key indicators such as the operating efficiency (such as average driving speed, driving time, etc.) and accident rate of the unmanned delivery vehicle under different congestion probabilities.

[0122] Implementation details: Set multiple simulation scenarios, each corresponding to different traffic flows and congestion probabilities. In each scenario, run multiple simulation experiments to obtain stable statistical results.

[0123] Record and organize simulation experiment data, including the driving trajectory, speed changes, accident occurrences, etc. of the unmanned delivery vehicle.

[0124] IV. Resource Allocation Decision

[0125] Step Description:

[0126] According to the results of simulation experiments, the impact of different congestion probabilities on the operating efficiency and safety of unmanned delivery vehicles is analyzed.

[0127] Based on the analysis results, determine the appropriate number of unmanned delivery vehicles and the number of safety officers required in honeycomb plots with different congestion probabilities.

[0128] Implementation details: Use data analysis tools (such as Excel, SPSS, etc.) to conduct statistical analysis on simulation experiment data, draw charts, and intuitively display the operating efficiency and accident rate of unmanned delivery vehicles under different congestion probabilities. According to the statistical analysis results, formulate a configuration plan for the number of unmanned delivery vehicles and the number of safety personnel. Consider factors such as economic costs and staffing in actual operations, and optimize and adjust the configuration plan. Apply the formulated resource allocation plan to actual operations for verification and adjustment. Collect actual operation data, compare and analyze it with the simulation experiment results, and evaluate the effectiveness and accuracy of the resource allocation plan. According to the actual operation situation, continuously optimize and improve the simulation model and resource allocation plan. Through the above specific implementation methods, the use of micro-simulation experiments to evaluate the operating risks of unmanned delivery vehicles and allocate resources can ensure the rationality and effectiveness of resource allocation, improve the operating efficiency and safety of unmanned delivery vehicles, and provide scientific decision-making support for enterprises.

[0129] This application also includes:

[0130] 1. Restore the real traffic environment:

[0131] Design and build a realistic miniature model based on the actual operating area of ​​the unmanned delivery vehicle. The model should include key features such as the actual road network, traffic flow, and traffic rules, and restore the real traffic environment as much as possible.

[0132] Model materials can be made of materials that are easy to process and observe, such as wood boards, plastic boards, paper, etc., to ensure the clarity and operability of the model.

[0133] Implementation details: Collect and organize actual road network data, including road length, width, number of lanes, intersection types, etc., and draw them on the model after scaling down.

[0134] According to the traffic flow data, set the vehicle generation point in the model to simulate the vehicle flow in different time periods and congestion conditions. Introduce traffic rules, such as priority rules, turning rules, etc., and display them through signs or indicators in the model.

[0135] 2. Simulation of the technical capabilities of unmanned delivery vehicles. Step description: Simulate the technical capabilities of unmanned delivery vehicles in a real miniature model. This can be achieved by setting different obstacles, risk scenarios and test conditions.

[0136] When an unmanned delivery vehicle encounters an obstacle or risk scenario, it has a certain probability of successfully avoiding or responding to it, depending on its technical capabilities (such as sensor accuracy, algorithm processing capabilities, etc.). The higher the technical capabilities, the greater the probability of successful response.

[0137] Implementation details: Set the initial technical capability value of the unmanned delivery vehicle, which can be determined based on data provided by the vehicle manufacturer or actual test results.

[0138] Multiple obstacles and risk scenarios are set up in the model, such as speed bumps, sharp turns, intersections, etc., to test the response capabilities of unmanned delivery vehicles.

[0139] Record the performance of unmanned delivery vehicles under different technical capabilities, such as the number of successful obstacle avoidances, response time, etc.

[0140] 3. Experimental operation and observation. Step description: In the constructed real miniature model, simulate different traffic flow and congestion probability scenarios. Run the experiment, observe and record key indicators such as the operating efficiency (such as driving speed, driving time, etc.) and accident rate of the unmanned delivery vehicle.

[0141] Implementation details: Set up multiple experimental scenarios, each corresponding to different traffic flows and congestion probabilities. In each scenario, run the experiment multiple times to obtain stable statistical results. Use tools such as timers and counters to record data such as the driving time of the unmanned delivery vehicle and the number of times it successfully avoids obstacles.

[0142] IV. Resource Allocation Decision

[0143] Step description: Based on the experimental results, analyze the impact of different congestion probabilities on the operating efficiency and safety of unmanned delivery vehicles. Based on the analysis results, determine the appropriate number of unmanned delivery vehicles and the number of safety officers required in honeycomb plots with different congestion probabilities.

[0144] Implementation details: Use data analysis tools (such as Excel, SPSS, etc.) to conduct statistical analysis on the experimental results and draw charts to intuitively display the operating efficiency and accident rate of unmanned delivery vehicles under different congestion probabilities. According to the statistical analysis results, formulate a configuration plan for the number of unmanned delivery vehicles and the number of safety personnel.

[0145] Consider factors such as economic costs and staffing in actual operations, and optimize and adjust the configuration plan.

[0146] V. Summary and Verification

[0147] Apply the resource allocation plan to actual operations for verification and adjustment. Collect actual operation data, compare and analyze with experimental results, and evaluate the effectiveness and accuracy of the resource allocation plan.

[0148] Based on actual operating conditions, the real miniature models and resource allocation plans are continuously optimized and improved.

[0149] Through the above specific implementation methods, the use of real miniature models to evaluate the operating risks of unmanned delivery vehicles and allocate resources can ensure the rationality and effectiveness of resource allocation and improve the operating efficiency and safety of unmanned delivery vehicles. At the same time, real miniature models have the advantages of being intuitive and easy to operate, which helps companies better understand the operating characteristics and risks of unmanned delivery vehicles and provide support for scientific decision-making.

[0150] In the present invention, the method for determining the state transition probability in the technical capability simulation of the unmanned delivery vehicle is as follows:

[0151] 1. Description of abnormal scene monitoring steps:

[0152] From the perspective of the individual unmanned delivery vehicle, through on-board sensors and surrounding environment monitoring equipment, its own operating parameters (such as speed, acceleration, position, etc.) and the operating parameters of potential interactive traffic participants in the surrounding environment (such as pedestrians, other vehicles, etc.) are monitored in real time.

[0153] Based on these parameters, the preset algorithms and thresholds are used to detect whether the unmanned delivery vehicle is in a set abnormal scenario. Abnormal scenarios include but are not limited to sudden deceleration, emergency stop, collision, etc.

[0154] Implementation details: Set the judgment criteria and thresholds for abnormal scenarios, such as speed change rate, acceleration magnitude, distance to surrounding objects, etc.

[0155] The operating parameters of the unmanned delivery vehicle and surrounding traffic participants are collected in real time, and pre-processed, such as filtering and denoising.

[0156] Compare the preprocessed data with the judgment criteria and thresholds to determine whether the unmanned delivery vehicle is in an abnormal scenario.

[0157] 2. Abnormal state variable setting:

[0158] Step description: In simulation modeling, the position coordinates (x, y), speed (v_x, v_y) and acceleration (a_x, a_y) of the unmanned delivery vehicle under abnormal conditions are measured as basic state variables. These basic state variables are combined into a state vector S(t)=[x(t), y(t), v_x(t), v_y(t), a_x(t), a_y(t)], where t represents the time.

[0159] Implementation details: Determine the accurate measurement method of the position, speed and acceleration of the unmanned delivery vehicle under abnormal conditions. Construct the state vector S(t) based on the measurement data.

[0160] 3. Extended exception status variables:

[0161] Step description: Based on the basic state vector, further variables such as the distance to surrounding vehicles or obstacles (d_{front}, d_{rear}), lane (lane(t), traffic signals and road attributes are included.

[0162] Combine these variables into the expanded state vector S(t)=[x(t),y(t),v_x(t),v_y(t),a_x(t),a_y(t),lane(t),d_{front}(t),d_{rear}(t)].

[0163] Implementation details: Use on-board sensors and surrounding environment monitoring equipment to measure the distance to surrounding vehicles or obstacles in real time. Determine the lane the unmanned delivery vehicle is in based on its driving trajectory and road information. Collect data on traffic signals and road attributes, such as traffic light status, road type, speed limit, etc.

[0164] 4. Discretization of abnormal state variables:

[0165] Step description: Discretize the state variables with continuous values ​​to facilitate the subsequent construction of the state transfer matrix. The discretization method can be equal interval division, cluster analysis, etc.

[0166] Implementation details: Determine the discretization granularity and method. Discretize each state variable to obtain the discretized state value.

[0167] 5. Establishment of state transfer matrix:

[0168] Step description: Based on the collected operating parameters of the unmanned delivery vehicle, construct the preliminary state transition probability matrix elements. Combine the kinematic principles and traffic rules of the unmanned delivery vehicle to improve the state transition matrix.

[0169] Implementation details: Use historical data and experimental data to count the probability of each state transition.

[0170] According to the kinematic model and traffic rules of the unmanned delivery vehicle, the state transition probability is corrected and optimized.

[0171] 6. Iterative calculation and matrix correction:

[0172] Step description: Use the established state transfer matrix to perform iterative calculations in time series, and gradually derive the operating state of the unmanned delivery vehicle at subsequent moments. Use the reconstructed operating state as a new state transfer matrix element and recalculate the state transfer probability.

[0173] This process is repeated continuously to correct and optimize the state transfer matrix until a stable and accurate state transfer probability is obtained.

[0174] Implementation details: Determine the initial conditions and step size of the iterative calculation. In each iteration, calculate the state at the subsequent time according to the state transfer matrix. Compare the calculated results with the actual situation and adjust the elements of the state transfer matrix.

[0175] 7. Technical capability simulation:

[0176] Step description: According to the modified state transition matrix, simulate the state transition probability of the unmanned delivery vehicle at different technical capability levels. The technical capability level can be achieved by adjusting the elements in the state transition matrix.

[0177] Implementation details: Set different technical capability levels, such as high, medium, and low. According to the technical capability level, adjust the elements in the state transfer matrix to reflect the behavioral characteristics of unmanned delivery vehicles under different technical capabilities.

[0178] The adjusted state transition matrix is ​​used to simulate the operating state of the unmanned delivery vehicle at different technical capability levels. Through the above specific implementation method, the state transition probability in the unmanned delivery vehicle technical capability simulation can be accurately determined, providing reliable data support for the subsequent unmanned delivery vehicle operation risk assessment and resource allocation.

[0179] The method for evaluating the operating risk of unmanned delivery vehicles of the present invention divides the entire area into three levels.

[0180] I. Regional level integration

[0181] Step description: First, divide the entire service range of the unmanned delivery vehicle into multiple independent areas according to factors such as geographical location, traffic conditions, and delivery needs. These areas should have clear boundaries and independent traffic characteristics.

[0182] Implementation details: Collect geographic information, traffic data, delivery demand and other data within the service area.

[0183] Using geographic information system (GIS) and data analysis tools, the service area is divided into multiple areas.

[0184] Ensure that each area has similar traffic conditions and delivery requirements, and that there are clear boundaries between areas.

[0185] 2. Definition of Cellular Plots

[0186] Step description: In each area, based on the operating characteristics and delivery needs of the unmanned delivery vehicle, an abstract hexagonal structure with the distribution center as the center and a reasonable delivery distance as the radius is defined as a honeycomb plot.

[0187] Implementation details: Determine the distribution center locations for each region.

[0188] Determine the reasonable delivery distance based on the operating speed and delivery time requirements of the unmanned delivery vehicle.

[0189] With the distribution center as the center and the reasonable distribution distance as the radius, a hexagonal structure is drawn as a honeycomb plot.

[0190] Ensure that each cell plot has similar delivery needs and traffic conditions.

[0191] 3. Identification and evaluation of key nodes

[0192] Road network definition: Inside each honeycomb, a road network is set, including the set of all nodes and the set of all links.

[0193] Collect road section attribute information, such as the number of lanes, whether non-motor vehicle lanes are set up, etc.

[0194] Dynamic traffic information updates:

[0195] Update the congestion information of each road section in the road network at a set frequency (such as every 5 minutes, every 10 minutes, etc.).

[0196] Congestion information includes the average driving speed of the road section, congestion status, etc.

[0197] Road congestion probability assessment:

[0198] For each road segment in the road network, its congestion probability is calculated based on its traffic information updated during the evaluation period.

[0199] Assume that the traffic information of a road section is updated N times during the evaluation period. If M of them are displayed as congested, the congestion probability P of the road section is M / N.

[0200] Screening of candidate sections for key nodes:

[0201] Set a congestion probability threshold ρ (such as 0.5, 0.6, etc.), and take the road sections with congestion probability exceeding the threshold ρ as candidate key node sections.

[0202] Spatial clustering operations:

[0203] A spatial clustering operation is performed on the candidate road segment set, and candidate road segments with the same road characteristics are merged into key nodes according to topological relationships.

[0204] Two types of key nodes are generated: key roads and key intersections.

[0205] Key node feature extraction:

[0206] Extract characteristic information of key nodes, including the number of lanes, whether vehicles and non-vehicles are mixed (for key roads), intersection type, whether there are traffic lights, the number of lanes on the upstream and downstream roads of the intersection, whether vehicles and non-vehicles are mixed (for key intersections), etc.

[0207] Definition of key nodes and calculation of congestion probability:

[0208] A key node is defined as a structure containing a unique identifier, type, attribute information, and road segment information.

[0209] For each key node, its congestion probability during the evaluation period is calculated.

[0210] Assume that the key node contains M road sections, and the congestion probability of each road section is P_i. After removing the extreme values, the congestion probability of the key node is the truncated mean of the congestion probabilities of each road section.

[0211] risk assessment:

[0212] The highest traffic operation risk of the key nodes within the cellular plot is taken as the traffic operation risk level of the entire cellular plot.

[0213] The highest traffic operation risk of the honeycomb plot is taken as the traffic operation risk level of the entire area.

[0214] Develop corresponding risk management measures and distribution strategies based on the risk level.

[0215] Through the above-mentioned specific implementation methods, a three-level assessment framework for the operation risks of unmanned delivery vehicles at the "key nodes of regional cellular plots" was constructed, enabling accurate assessment and effective management of the operation risks of unmanned delivery vehicles.

[0216] The method for evaluating the operating risk of unmanned delivery vehicles of the present invention further determines the deployment volume of unmanned delivery vehicles and the number of safety personnel in the honeycomb plot. The method comprehensively considers the risk integration of safety and efficiency, and optimizes the deployment strategy of unmanned delivery vehicles by constructing a comprehensive risk assessment model.

[0217] Comprehensive risk assessment model construction: defining risk indicators

[0218] Efficiency risk: refers to the risk of reduced delivery efficiency due to insufficient deployment of unmanned delivery vehicles, or the risk of waste of resources and increased traffic congestion due to excessive deployment.

[0219] Safety risk: refers to the safety risks caused by traffic accidents, equipment failures, etc. that may occur during the operation of unmanned delivery vehicles.

[0220] Determine the weighting coefficients:

[0221] α: The weighting coefficient of efficiency risk, reflecting the degree of emphasis on efficiency.

[0222] β: The weighted coefficient of safety risk, reflecting the degree of importance attached to safety.

[0223] The value range of α and β is set to 0 to 1, and α + β = 1. The specific value can be adjusted according to actual needs and policy orientation.

[0224] Building a comprehensive risk assessment model

[0225] The comprehensive risk assessment model formula is as follows:

[0226] Comprehensive risk = α × efficiency risk + β × safety risk

[0227] The specific calculation methods of efficiency risk and safety risk can be defined and quantified according to the actual situation. For example, efficiency risk can be measured by indicators such as delivery time and delivery cost; safety risk can be measured by indicators such as traffic accident rate and equipment failure rate.

[0228] Determine the number of unmanned delivery vehicles and the number of safety personnel: data collection and preprocessing

[0229] Collect traffic data, delivery demand data, unmanned delivery vehicle performance data, etc. within the cellular area.

[0230] Preprocess the data, including data cleaning and data standardization, to ensure the accuracy and comparability of the data.

[0231] Risk assessment and launch strategy formulation

[0232] Based on the comprehensive risk assessment model, calculate the comprehensive risk value under different release amounts.

[0233] Analyze the changing trend of the comprehensive risk value and find out the range of investment volume with the lowest risk or an acceptable risk range.

[0234] Consider the operating efficiency, cost-effectiveness and safety requirements of unmanned delivery vehicles to determine the final deployment quantity.

[0235] Assess the need for safety officers based on factors such as the number of unmanned delivery vehicles deployed, operating routes, and traffic conditions.

[0236] Set the ratio between the number of safety officers and the number of unmanned delivery vehicles deployed, or determine the number of safety officers based on specific safety risk assessment results.

[0237] Ensure that the number of safety personnel can meet the needs of safety supervision and emergency response.

[0238] Regularly evaluate the operating performance and safety status of unmanned delivery vehicles, and adjust the deployment volume and number of safety personnel based on the evaluation results.

[0239] Collect user feedback and opinions, and continuously optimize the operating strategies and service quality of unmanned delivery vehicles.

[0240] This specific implementation method optimizes the deployment strategy of unmanned delivery vehicles by building a comprehensive risk assessment model to integrate the efficiency risk and safety risk after the deployment of unmanned delivery vehicles. Through steps such as data collection, risk assessment, deployment strategy formulation, and determination of the number of safety personnel, the safe and efficient operation of unmanned delivery vehicles in the honeycomb plot is ensured. This method is flexible and scalable and can be adjusted and optimized according to actual needs.

[0241] In the present invention, the ratio of the number of safety officers to unmanned delivery vehicles is determined as follows: the probability of congestion at key nodes is used as the main risk measurement indicator, and a comprehensive risk quantification model is established in combination with the probability of accidents such as collisions and deviations from the driving path that may occur in unmanned delivery vehicles in a congested environment. Through statistical analysis of historical accident data or expert experience, corresponding weights are assigned to various types of accident risks under different congestion levels. For example, when the probability of congestion is in the range of 0-20%, the risk weight of a collision accident is 0.3, and the risk weight of deviation from the path is 0.2; when the probability of congestion is in the range of 20%-50%, the risk weight of a collision accident is increased to 0.5, and the risk weight of deviation from the path is increased to 0.3, etc.

[0242] The autonomous handling ability score of the unmanned delivery vehicle in dealing with complex traffic conditions is determined based on its technical performance and intelligent driving level. For example, the autonomous handling ability score of an unmanned delivery vehicle with L4 autonomous driving capability is 8 points (out of 10 points) under normal traffic conditions, but in high congestion scenarios with a congestion probability of more than 50%, the autonomous handling ability score drops to 4 points due to factors such as sensor occlusion and increased complexity of the decision-making algorithm.

[0243] Based on the above risk quantification model, different ratios of the number of safety officers to unmanned delivery vehicles are set for simulation analysis. The initial plan can refer to industry experience or simply set it according to a ratio of 1:10, 1:15, 1:20, etc. Under each ratio, rerun the traffic simulation, simulate the delivery process of the unmanned delivery vehicle, and monitor the changes in risk indicators in real time.

[0244] When the comprehensive risk index (such as the result of the comprehensive consideration of the risk probability weighted score and the autonomous handling ability score of the unmanned delivery vehicle) is lower than the preset safety threshold under a certain ratio scheme, the ratio scheme is considered to be initially feasible. Further fine-tuning and optimization of the feasible scheme is carried out. For example, on the basis of the 1:15 ratio scheme, more refined adjustments such as 1:14 and 1:13 are tried to observe whether the risk index is further reduced until the ratio of the number of safety officers and unmanned delivery vehicles that optimizes the comprehensive risk index is found. For example, after multiple rounds of simulation optimization, it was found that when the maximum number of unmanned delivery vehicles in a certain area is 100, 6 safety officers (i.e., a ratio of about 1:16.7) can achieve a good balance between cost and risk control, ensuring that the operating risk of unmanned delivery vehicles is within an acceptable range.

[0245] Through the above detailed implementation methods, it is possible to effectively evaluate the operating risks of unmanned delivery vehicles, determine a reasonable safety officer deployment strategy, and ensure the safe and efficient development of unmanned delivery services.

[0246] The cellular area model on which the unmanned delivery vehicle operates. In this embodiment, the cellular area is planned based on the cellular structure. Each cellular center serves as a key reference point to define the effective travel range of the unmanned delivery vehicle.

[0247] When an unmanned delivery mission is initiated, the system limits the travel of the unmanned delivery vehicle to a specific area with the center of the cell as the center according to pre-set rules. Specifically, the travel of the unmanned delivery vehicle is set to be between 2 and 3.5 times the radius of the cell with the center of the cell as the center. This range is not set arbitrarily, but takes into account many factors.

[0248] On the one hand, from the perspective of energy supply, within this range, the energy (such as battery power) carried by the unmanned delivery vehicle is sufficient to support it to complete the delivery task and safely return to the cellular center for resupply or maintenance, avoiding the risk of breaking down midway due to energy exhaustion. Taking existing battery technology as an example, after many actual tests, within this range, the energy consumption of the delivery vehicle is at a relatively reasonable level, which can ensure that there is a certain amount of power left at the end of the task to meet emergency needs.

[0249] On the other hand, considering the stability of signal transmission, maintaining a range of 2 to 3.5 times the cell radius can ensure that the unmanned delivery vehicle and the control center always maintain a high-intensity and stable communication connection. The cellular layout itself is conducive to signal coverage and relay transmission. Within this range, even if there are some situations such as building obstruction and electromagnetic interference, the coordinated switching of surrounding cellular base stations can ensure real-time data return and accurate command delivery, reducing the risk of loss of control caused by communication interruption.

[0250] Furthermore, based on the analysis of traffic flow and road condition complexity, this range is suitable for traffic planning in most cities or parks. Usually, when a delivery vehicle encounters unexpected situations such as sudden traffic congestion or road construction changes within this travel range, there are enough redundant routes to choose from, so as to flexibly adjust the delivery route, avoid high-risk sections, and reduce the possibility of delays and collisions.

[0251] In actual operation, when the unmanned delivery vehicle receives a delivery instruction, the on-board navigation system combines the cellular regional map to quickly plan the optimal route that meets the travel requirements. The route planning algorithm not only considers the principle of shortest distance, but also comprehensively weighs the above-mentioned factors such as energy, signal, and road conditions, and dynamically adjusts the driving route in real time to ensure that the entire delivery process is in a low-risk operation state. At the same time, the monitoring center always tracks the vehicle's position. Once the vehicle approaches the boundary of the journey, the system will issue a warning prompt in time, and take remote intervention measures if necessary to ensure the safe operation of the vehicle.

[0252] By setting the cellular area range so precisely, the safety and reliability of unmanned delivery vehicles during operation can be improved in an all-round and systematic manner, laying a solid foundation for the large-scale promotion of unmanned delivery services.

[0253] In a honeycomb plot, the highest traffic operation risk of a key node reflects the traffic operation risk level of the entire plot; and the highest traffic operation risk of a honeycomb plot represents the traffic operation risk level of the entire area. This process is achieved through the following steps:

[0254] First, a key node assessment is conducted to determine the highest traffic operation risk of each key node within the honeycomb plot. This risk value represents the traffic operation risk level of the corresponding honeycomb plot.

[0255] Then, a cell plot assessment is performed, which receives data from the key node assessment module and determines the highest traffic operation risk of the cell plot based on the maximum value of the highest traffic operation risk of the key nodes in each cell plot;

[0256] Finally, a regional assessment is performed, which is used to receive data from multiple cellular plot assessment modules and determine the traffic operation risk level of the entire area based on the maximum value of the highest traffic operation risk in all cellular plots.

[0257] Specifically, the three-level assessment framework is accurate and can provide decision-making references because, in order to assess and determine the traffic operation risk level of the cell plot and the entire area, we take the following specific implementation steps:

[0258] Critical node assessment: First, identify all critical nodes within the cell plot, which are usually locations with high traffic volume, prone to congestion, or of strategic importance.

[0259] Conduct detailed traffic operation risk assessments for each key node, including but not limited to traffic flow analysis, congestion frequency, accident rate and other indicators.

[0260] Based on the evaluation results, the highest traffic operation risk value of each key node is determined, which represents the traffic operation risk level of the node under extreme conditions.

[0261] Cellular plot assessment: collects data from the key node assessment module, i.e. the highest traffic operation risk value of each key node.

[0262] In each honeycomb plot, the highest traffic operation risk values ​​of all key nodes are compared, and the maximum value is selected as the highest traffic operation risk value of the honeycomb plot.

[0263] This step ensures that the traffic operation risk level of each cell plot is based on the assessment results of the highest risk node within it.

[0264] Regional assessment: Receives data from multiple cellular plot assessment modules, i.e., the highest traffic operation risk value for each cellular plot.

[0265] In the entire area, the highest traffic operation risk values ​​of all honeycomb plots are compared, and the maximum value is selected as the highest traffic operation risk value of the entire area.

[0266] This step provides a comprehensive assessment of the risk level of traffic operations in the entire region, helping decision makers to develop targeted traffic management measures and emergency plans.

[0267] Through the above steps, we can accurately and efficiently assess the traffic operation risk level of the honeycomb plots and the entire area, and provide strong support for urban traffic planning and management.

[0268] The application discloses a method and system for evaluating the operational risks of unmanned delivery vehicles. With the development of unmanned delivery vehicles, their deployment faces problems such as adaptability to urban traffic and their own technical capabilities. For this reason, a three-level evaluation framework including regions, cells, and key nodes is proposed to solve the problems of operational risk assessment and rule-making.

[0269] 1. Background and Problems Traditional manual delivery has many problems, and unmanned delivery technology has emerged. However, when it is deployed, it faces challenges such as matching with urban traffic operation characteristics, technical capability assessment, and spatial range setting. Therefore, an effective operation risk assessment algorithm with multiple characteristics is needed.

[0270] 2. Technical solution Three-level modeling solution: Based on the "short board theory of the wooden barrel", after the unmanned delivery vehicle is deployed, the maximum operating risk of the key nodes in the cell determines the cell risk, and the maximum operating risk of the cell determines the maximum operating risk of the region. This solution transforms the regional risk assessment into a cell risk assessment problem.

[0271] Key node identification: Based on dynamic traffic information, candidate key node sections are screened according to the congestion probability of the section, key nodes (key roads and key intersections) are generated through spatial clustering, and their features are extracted to calculate the congestion probability of the key nodes.

[0272] Empirical data collection: Based on the selection of key nodes, traffic participant data and key node characteristics of typical roads and intersections are collected.

[0273] Micro-simulation environment configuration: Select the cellular road network and load the road network and traffic participant datasets.

[0274] When unmanned delivery vehicles are not deployed, the probability of congestion at key nodes is evaluated as the initial risk. When unmanned delivery vehicles are deployed, their operation process needs to be modeled (delivery routes, dwell time, deployment strategies, etc.), and a meso-modeling method based on scene detection and state probability transfer is used to build a capability model. The impact of key node operation risks is evaluated from the dimensions of safety and efficiency, and the comprehensive risk is obtained by fusion. Multiple experiments are conducted to obtain a data set of the corresponding relationship between the initial risk and the comprehensive risk, and then a risk model is obtained, and the risk level can be expressed in a discrete manner.

[0275] Regional risk assessment: Define the region and cellular structure, and set the deployment strategy for unmanned delivery vehicles. Divide the assessment batches, identify the key nodes in the cellular structure in each batch, calculate the distribution of unmanned delivery vehicles, update the comprehensive risk of key nodes, summarize the comprehensive risk of cellular and regions, and determine whether the deployment strategy needs to be adjusted or determine the deployment parameter plan that meets risk control based on the risk value.

[0276] Beneficial effects: Propose an innovative assessment framework, improve the risk assessment model, optimize the unmanned delivery vehicle capability modeling, improve the accuracy and objectivity of risk assessment, provide strong support for unmanned delivery vehicle deployment decisions, and expand the flexibility of risk assessment tasks.

[0277] Suppose we select a certain area of ​​a city as the evaluation area, and divide the area into three large areas (labeled as area A, area B, and area C respectively). Each large area is further divided into multiple cells, as shown in Table 1:

[0278] Table 1: Area division and cell settings

[0279]

[0280] Taking cell 1 as an example, after spatial clustering of the key node candidate sections selected based on dynamic traffic information, the following key nodes (including key roads and key intersections) are determined. Key node number K1 is a key intersection, key node number K2 is a key road, key node number K3 is a key intersection, etc. The congestion probability of each key node is calculated based on the collected data and related algorithms. For example, traffic participant data (such as the number of pedestrians, the number of vehicles of different types, etc.) and more detailed key node features (such as lane allocation in each direction of the intersection, traffic light duration, etc.) are collected at the key node K1. When configuring the micro-simulation environment, the road network of cell 1 is selected, and the road network dataset containing the above-mentioned key nodes and surrounding roads and the corresponding traffic participant dataset are loaded. The micro-simulation environment is constructed, and some basic parameters are set, such as the simulation time step, the total simulation time (simulated as 24 hours), etc. When the unmanned delivery vehicle is not deployed (initial risk assessment), after multiple simulation experiments, the initial congestion probability of the key node K1 when the unmanned delivery vehicle is not deployed is calculated. When the unmanned delivery vehicle is deployed (comprehensive risk assessment), N unmanned delivery vehicles are deployed in cell 1, and the meso-modeling method based on scene detection and state probability transfer is used to build its operation capability model, and different delivery routes, residence time and other strategies are set. Through multiple experiments, the corresponding relationship data set between initial risk and comprehensive risk under different deployment strategies is obtained, and then the risk model of the key node K1 is obtained. For example, a simple linear relationship model is fitted through data analysis: let the initial risk congestion probability be x, and the comprehensive risk congestion probability be y, and build a relationship model (different nodes, different time periods, etc. may have different models, usually nonlinear models), and discretize the risk level according to certain standards. For example, a congestion probability less than a1 is a low risk level, a1-a2 is a medium risk level, and greater than a2 is a high risk level. During the regional risk assessment, determine the total number of unmanned delivery vehicles to be deployed in Cell 1-Cell 5 of Region A, and deploy them in Cell 1-Cell 5 according to the deployment strategy. Divide into K assessment batches, identify the key nodes in each cell in each batch, calculate the distribution of unmanned delivery vehicles, update the comprehensive risk of key nodes, and summarize the comprehensive risks of cells and regions. Determine whether the deployment strategy needs to be adjusted based on the calculated regional comprehensive risk value. For example, if the comprehensive risk of Region A exceeds the set risk threshold, consider adjusting the deployment parameter plan such as the deployment quantity and delivery route to meet the risk control requirements.

[0281] The present invention has at least the following beneficial effects:

[0282] First, the present invention proposes a three-level assessment framework for the operation risk of unmanned delivery vehicles, namely "region-cellular structure-key nodes". By evaluating the congestion of road sections and intersections within the honeycomb structure, the key nodes of risk in urban traffic operation are automatically screened and identified. The static road network structure and dynamic traffic operation characteristics of the key nodes are referred to, and efficiency and safety are considered from the two dimensions of congestion and accidents. A regional risk assessment for the operation of unmanned delivery vehicles is proposed, thereby forming constraints on the deployment strategy of unmanned delivery vehicles.

[0283] Second, the present invention is based on the operating characteristics of unmanned delivery vehicles and independent honeycomb structures to flexibly adapt to the evaluation needs of different area sizes. The evaluation processes between different honeycombs are decoupled, thereby achieving the spatial independence of the algorithm. The comprehensive operating risk of the set area is obtained by integrating the comprehensive operating risks of the honeycomb structure, and the comprehensive operating risk of the honeycomb structure is obtained by the maximum comprehensive risk of the key nodes, ultimately realizing the formulation of rule plans after the unmanned delivery vehicles are deployed on open roads.

[0284] Third, the present invention collects road operation data with different road network structures and different traffic characteristics, and obtains empirical data on changes in road operation risks under different road structures, different traffic characteristics, and different unmanned delivery vehicle deployment strategies through micro-simulation methods. The above data are modeled using neural networks to construct risk models for key nodes.

[0285] Fourth, the present invention defines the scenarios in which unmanned delivery vehicles interact with other traffic participants, and based on the extraction of technical features of unmanned delivery vehicles, constructs a state transfer matrix of unmanned delivery vehicles in different scenarios, and realizes the assessment of risk changes after the unmanned delivery vehicles are deployed in risk scenarios. Once entering the preset risk scenario, the unmanned delivery vehicle is probabilistically converted to a new operating state according to the state transfer matrix, which further triggers changes in the traffic operation situation.

[0286] The number of devices and processing scales described here are used to simplify the description of the present invention. Applications, modifications and variations of the present invention will be obvious to those skilled in the art.

[0287] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the details shown and described herein.

Claims

1. A method for evaluating the operating risk of an unmanned delivery vehicle, characterized in that: The method comprises the following steps: Operation area division: The operation area of ​​the unmanned delivery vehicle is divided into multiple honeycomb-shaped plots to ensure that there is no overlap between the plots and that the operation area is fully covered; Key nodes determined: Step a: According to the static road network structure, obtain and determine the attribute data of each road in each honeycomb plot; Step b: Evaluate and determine the congestion probability of each road in each cell block based on the dynamic traffic operation characteristics; Step c: Set a congestion threshold ρ, classify the roads whose congestion probability exceeds the threshold ρ according to their attribute data, and merge the same type of roads according to the topological relationship between the roads to form one or more key node sets; Calculation of plot congestion probability: For each honeycomb plot, calculate the truncated mean of the congestion probability of all key nodes in it. The truncated mean of the congestion probability of each key node is the truncated mean of the congestion probability of the road section included in the key node, and select the maximum value from the truncated means of all key nodes as the overall congestion probability of the plot; Distribution resource allocation: Based on the congestion probability of each honeycomb plot, determine the appropriate number of unmanned delivery vehicles in the plot and the number of safety personnel responsible for handling potential accidents; The steps of determining the number of unmanned delivery vehicles and the number of safety personnel based on the congestion probability of the honeycomb plot are achieved through the following micro-simulation experiments: Honeycomb plot simulation modeling: Build a honeycomb plot simulation model that reflects the actual road network, traffic flow, and traffic rules; Unmanned delivery vehicle technical capability simulation: In the simulation model, the technical capabilities of the unmanned delivery vehicle are simulated, and the level of technical capabilities is adjusted by setting the state transition probability; When an unmanned delivery vehicle encounters a risk scenario, it will transfer from a risk state to a normal state or an abnormal state based on its technical capabilities, i.e., the state transition probability. The higher the technical capabilities, the greater the probability of transferring to a normal state. Simulation experiment running: In the constructed cellular plot simulation model, traffic flow, road attributes, and traffic signal data are input; Based on the technical capability simulation of the unmanned delivery vehicle, run simulation experiments to observe and record the key indicators of the unmanned delivery vehicle's operating efficiency and accident rate under different congestion probabilities; Resource allocation decisions: According to the results of simulation experiments, the impact of different congestion probabilities on the operating efficiency and safety of unmanned delivery vehicles is analyzed; Based on the analysis results, determine the appropriate number of unmanned delivery vehicles and the number of safety personnel required in honeycomb plots with different congestion probabilities; In the simulation of the technical capability of the unmanned delivery vehicle, the state transition probability is determined by the following steps: Abnormal scene monitoring: From the perspective of the individual unmanned delivery vehicle, the system monitors its own operating parameters and the operating parameters of potential interactive traffic participants in the surrounding environment in real time. Based on these parameters, it detects whether the unmanned delivery vehicle is in a set abnormal scene. The behavioral data in the abnormal scene includes: data on sudden deceleration, emergency stop, and collision of the unmanned delivery vehicle; Abnormal state variable setting: In simulation modeling, the position coordinates (x, y), speed (v_x, v_y) and acceleration (a_x, a_y) of the unmanned delivery vehicle under abnormal conditions are measured as basic state variables to form a state vector S(t) = [x(t), y(t), v_x(t), v_y(t), a_x(t), a_y(t)], where t represents the time; Expand abnormal state variables: further include variables such as the distance to surrounding vehicles or obstacles (d_{front}, d_{rear}), lane lane(t), traffic signals and road attributes to form an expanded state vector S(t) = [x(t), y(t), v_x(t), v_y(t), a_x(t), a_y(t), lane(t), d_{front}(t), d_{rear}(t)]; Discretization of abnormal state variables: Discretize the state variables with continuous values ​​to facilitate the subsequent construction of the state transfer matrix; Establishment of state transfer matrix: Based on the collected operating parameters of the unmanned delivery vehicle, construct the preliminary state transfer probability matrix elements, and combine the kinematic principles and traffic rules of the unmanned delivery vehicle to improve the state transfer matrix; Iterative calculation and matrix correction: Use the established state transfer matrix to perform iterative calculations in time series, gradually derive the operating state of the unmanned delivery vehicle at subsequent moments, use the reconstructed operating state as a new state transfer matrix element, recalculate the state transfer probability, and repeat this process to correct and optimize the state transfer matrix until a stable and accurate state transfer probability is obtained; Technical capability simulation: Based on the revised state transition matrix, simulate the state transition probability of the unmanned delivery vehicle at different technical capability levels.

2. The method for evaluating the operating risk of an unmanned delivery vehicle according to claim 1, characterized in that: The step of calculating the traffic congestion points in each cell block and taking them as key nodes is implemented based on dynamic traffic information, and specifically includes the following steps: Road network definition: Set the road network of each honeycomb plot in the area, including the set of all nodes and the set of all road sections. The attribute information collected on the road sections includes the number of lanes and whether non-motor vehicle lanes are set; Dynamic traffic information update: Update the congestion information of each road section in the road network at a set frequency, including the average driving speed and congestion status of the road section; Evaluation of road section congestion probability: For each road section in the road network, its congestion probability is calculated based on the traffic information updated during the evaluation period. Specifically, if the road section updates traffic information N times during the evaluation period, and if M of them show congestion, then the congestion probability P of the road section is M / N; Screening of candidate road sections for key nodes: setting a congestion probability threshold ρ, and taking road sections with congestion probability exceeding the threshold ρ as candidate road sections for key nodes; Spatial clustering operation: Perform spatial clustering on the candidate road segment set, and merge the candidate road segments with the same road characteristics into key nodes according to the topological relationship. The key nodes include key roads and key intersections. Key node feature extraction: Extract feature information of key nodes. The node features of key roads include the number of lanes and whether vehicles and non-motorized vehicles are mixed. The node features of key intersections include the type of intersection, whether there are traffic lights, and the number of lanes of the upstream and downstream roads of the intersection and whether vehicles and non-motorized vehicles are mixed. Definition of key nodes and calculation of congestion probability: A key node is defined as a structure containing a unique identifier, type, attribute information and road section information. For each key node, its congestion probability during the evaluation period is calculated. Specifically, suppose the key node contains M road sections, and the congestion probability of each road section is P_i. After removing extreme values, the congestion probability of the key node is the truncated mean of the congestion probabilities of each road section.

3. The method for evaluating the operating risk of an unmanned delivery vehicle according to claim 2, characterized in that: The method for determining the number of unmanned delivery vehicles and the number of safety personnel in a honeycomb plot comprehensively considers the risk integration of both safety and efficiency. Comprehensive risk assessment model: A comprehensive risk assessment model is constructed, which integrates the efficiency risk and safety risk after the deployment of unmanned delivery vehicles. The calculation formula is as follows: Comprehensive risk = α × efficiency risk + β × safety risk Among them, α and β are the weighted coefficients of efficiency risk and safety risk respectively, and α+β=1.

4. A system for evaluating the operating risk of an unmanned delivery vehicle for implementing the method of any one of claims 1 to 3, characterized in that: The system includes: Operation area division module: used to divide the operation area of ​​the unmanned delivery vehicle into multiple honeycomb-shaped plots to ensure that there is no overlap between the plots and that the operation area is fully covered; Key node determination module: Attribute data acquisition submodule: According to the static road network structure, the attribute data of each road in each honeycomb plot is acquired and determined; Congestion probability assessment submodule: evaluates and determines the congestion probability of each road in each cell plot based on dynamic traffic operation characteristics; Key node set generation submodule: set a congestion threshold ρ, classify the roads whose congestion probability exceeds the threshold ρ according to their attribute data, and merge the same type of roads according to the topological relationship between the roads to form one or more key node sets; Block congestion probability calculation module: For each honeycomb block, the truncated mean of the congestion probabilities of all key nodes in it is calculated. The truncated mean of the congestion probability of each key node is the truncated mean of the congestion probabilities of the road sections included in the key node, and the maximum value is selected as the overall congestion probability of the block; Distribution resource allocation module: Based on the congestion probability of each honeycomb plot, determine the appropriate number of unmanned delivery vehicles in the plot and the number of safety personnel responsible for handling potential accidents.

Images