A dynamic path optimization method and device for a patrol vehicle
By using a large model to understand the multidimensional features of the inspection vehicle's environment and status information, a dynamic path cost function is constructed and corrected by real-time energy consumption data. This solves the shortcomings of path planning in existing technologies, enables accurate prediction and efficient utilization of the inspection vehicle's energy consumption, and improves the reliability and economy of the mission.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 北京北创芯通科技有限公司
- Filing Date
- 2026-03-09
- Publication Date
- 2026-06-02
Smart Images
Figure CN122130112A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of path planning technology, and in particular to a dynamic path optimization method and apparatus for an inspection vehicle. Background Technology
[0002] In the field of intelligent inspection, path planning for inspection vehicles with energy consumption optimization as the goal has become a key technology to improve operational efficiency and endurance. Existing methods mostly rely on traditional graph search algorithms, whose path costs are typically composed of distance, time, or energy consumption estimates based on simplified physical models with fixed weights. These approaches have the following limitations: first, the energy consumption models used lack expressive power, making it difficult to accurately quantify the combined impact of complex road conditions, slope, weather, and vehicle load; second, they lack the ability to understand and integrate high-level environmental semantics such as "slippery road surface" and "frequent starts and stops," resulting in insufficient planning basis. Summary of the Invention
[0003] This specification provides a method and apparatus for dynamic path optimization of an inspection vehicle to solve at least one of the technical problems mentioned above.
[0004] To solve the above-mentioned technical problems, the embodiments in this specification are implemented as follows: According to a first aspect of the embodiments of this specification, a dynamic path optimization method for an inspection vehicle is provided, comprising: Collect inspection environment information and vehicle current status information, including geospatial attributes and dynamic environmental parameters; Based on the inspection environment information, a candidate path set is constructed, and the feature information of each road segment in the candidate path set is input into a pre-trained large model. The large model outputs the energy consumption judgment value or energy consumption distribution parameter of each road segment. Based on the energy consumption determination results output by the large model, a dynamic path cost function is constructed, and the cost of each candidate path is evaluated based on the path cost function. The candidate path with the lowest path cost is selected from all candidate paths as the energy-optimal path, and the inspection vehicle is controlled to perform the inspection task. The actual energy consumption data of the inspection vehicle when executing the energy-optimal path is collected, and the output of the large model is corrected based on the deviation between the actual energy consumption data and the energy consumption judgment result output by the large model. When the energy consumption deviation is detected to exceed the preset threshold or the vehicle status does not meet the preset conditions, path replanning is triggered, and the energy consumption determination and path optimization steps based on the large model are re-executed to generate a new energy-optimal path.
[0005] In some optional implementations, the step of constructing a candidate path set based on the inspection environment information specifically includes: The inspection area is modeled as a directed graph. ,in, This represents a set of intersections or key inspection nodes. Represents a set of road segments. , Indicates from node To the node The section of road; Each section Corresponding to a set of basic feature vectors , is represented as: ;in, Indicates road segment Length, Indicates road segment The slope, Indicates road segment Road condition category, Indicates road segment obstacle density, Indicates road segment Weather conditions, This represents the vehicle state vector.
[0006] In some optional implementations, the step of outputting energy consumption determination values or energy consumption distribution parameters for each road segment from the large model specifically includes: The basic feature vector Input large model Obtain the road section Predicted energy consumption judgment value The calculation method is as follows: ,in, Indicates the section of road the inspection vehicle passed through. Predicted energy consumption; or, The large model Output section The energy consumption distribution parameters are expressed as: ,in, This represents the average of the predicted energy consumption. This represents the variance of the predicted energy consumption.
[0007] In some alternative implementations, for candidate paths Its total predicted energy consumption The calculation method is as follows: ,in, Representing a path The Middle Predicted energy consumption values for each road segment.
[0008] In some alternative implementations, the dynamic path cost function Represented as: ;in, Representing a path The cost of travel time, Representing a path The cost of risk or uncertainty, , , These are dynamic weight parameters.
[0009] In some optional implementations, the step of correcting the output of the large model based on the deviation between the actual energy consumption data and the energy consumption determination result output by the large model specifically includes: When the inspection vehicle is executing the energy-optimal path, it passes through the following road sections. Actual energy consumption ; Calculate the actual energy consumption The predicted energy consumption value output by the large model Energy consumption deviation between The calculation method is as follows: ; Based on the energy consumption deviation The output of the large model is corrected, and the corrected predicted energy consumption value is determined. The calculation method is as follows: ;in, The preset learning rate parameter; The conditions for triggering path replanning include: The absolute value of the energy consumption deviation is greater than a preset threshold. That is, satisfying ;or, The remaining battery power of the inspection vehicle Below the preset battery threshold That is, satisfying .
[0010] In some alternative implementations, the dynamic weight parameters The adjustment strategy is based on the vehicle's current remaining battery power. and task urgency indicators The adaptive determination method is as follows: Based on the remaining power range and the urgency of the task, the basic weight vector is obtained by looking up a table or through a fuzzy logic controller. ; Based on this, the total energy consumption is estimated according to the route. Uncertainty measurement Make minor adjustments; the formula is adjusted as follows: in, For the first Energy consumption variance of each road segment This represents the uncertainty factor.
[0011] In some alternative implementations, the risk or uncertainty cost Calculate using the following steps: Based on the energy consumption distribution parameters output by the large model Calculation path The total energy consumption distribution is ; Define risk cost To achieve a given confidence level The energy consumption condition risk value CVaR under the following conditions is calculated using the following formula: in, Indicates the total energy consumption distribution along the path. Inverse function under quantiles It represents the mathematical expectation.
[0012] In some optional implementations, the optimization process of the path cost function is constructed as a constrained multi-objective optimization problem. The objective is to balance energy consumption, time, and risk while ensuring task completion time and power safety. Its mathematical expression is: in, The maximum time allowed for the task. This is the initial charge level. To ensure a safe minimum power consumption, Given an energy consumption budget, The probability is the safety threshold. By solving the Pareto front of this multi-objective optimization problem, and dynamically selecting the optimal compromise solution based on the current power level and task stage, the energy-optimal path is obtained.
[0013] According to a second aspect of the embodiments of this specification, a dynamic path optimization device for an inspection vehicle is provided, the device comprising: The information acquisition module is used to collect inspection environment information and vehicle current status information, including geospatial attributes and dynamic environmental parameters. The path modeling and energy consumption determination module is used to construct a candidate path set based on the inspection environment information, and input the feature information of each road segment in the candidate path set into a pre-trained large model, and the large model outputs the energy consumption determination value or energy consumption distribution parameter of each road segment. The path evaluation module is used to construct a dynamic path cost function based on the energy consumption determination results output by the large model, and to evaluate the cost of each candidate path based on the path cost function. The path selection and control module is used to select the candidate path with the lowest path cost from the candidate paths as the energy-optimal path, and control the inspection vehicle to perform the inspection task. The feedback correction module is used to collect the actual energy consumption data of the inspection vehicle when executing the energy-optimal path, and correct the output of the large model based on the deviation between the actual energy consumption data and the energy consumption judgment result output by the large model. The route replanning module is used to trigger route replanning when the energy consumption deviation is detected to exceed a preset threshold or the vehicle status does not meet the preset conditions. This triggers the re-execution of the energy consumption determination and route optimization steps based on the large model to generate a new energy-optimal route.
[0014] One embodiment of this specification can achieve at least the following beneficial effects: In this application's technical solution, the dynamic route optimization method for inspection vehicles does not rely solely on traditional fixed distance or time costs when generating the energy-optimal route. Instead, it first uses a large model to understand and calculate multi-dimensional features composed of environmental and vehicle states, outputting energy consumption judgment information for each road segment. Then, based on this judgment result, a path cost function integrating energy consumption, time, and risk is dynamically constructed, and the path with the minimum overall cost is selected as the execution basis. Because the large model can comprehensively analyze complex semantic information such as road conditions, slope, and weather, its accuracy in energy consumption judgment is significantly improved compared to traditional simplified models. Furthermore, during the actual execution of the inspection vehicle, the solution continuously collects real energy consumption data and compares and corrects it with the model's judgment values, while simultaneously monitoring key states and deviations. Once a significant deviation of the energy consumption prediction from reality or insufficient vehicle battery power is detected, a new round of optimization is immediately triggered. This closed-loop mechanism, based on high-precision semantic judgment of large models and combined with real-time feedback and condition triggering, enables path planning to closely match the dynamic environment and actual vehicle status. This allows for more accurate prediction and efficient utilization of the inspection vehicle's energy consumption, as well as more effective avoidance of mission interruption risks, thereby improving the overall reliability and economy of the inspection mission. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1This is a flowchart of a dynamic path optimization method for an inspection vehicle provided in the embodiments of this specification; Figure 2 For corresponding Figure 1 A schematic diagram of the structure of a dynamic path optimization device for an inspection vehicle. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of one or more embodiments of this specification clearer, the technical solutions of one or more embodiments of this specification will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of them. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of one or more embodiments of this specification.
[0018] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another.
[0019] Figure 1 This is a flowchart of a dynamic path optimization method for an inspection vehicle provided in the embodiments of this specification; Figure 2 For corresponding Figure 1 A schematic diagram of the structure of a dynamic path optimization device for an inspection vehicle.
[0020] This application provides a dynamic path optimization method for inspection vehicles, such as... Figure 1 As shown, the method may include: Step 102: Collect inspection environment information and vehicle current status information, including geospatial attributes and dynamic environmental parameters.
[0021] In the embodiments of this specification, an inspection vehicle can refer to an intelligent vehicle capable of autonomous or semi-autonomous driving within an inspection area and performing tasks such as monitoring and inspection. For inspection tasks, the inspection vehicle needs to safely and efficiently complete the inspection work along a designated route in a complex and ever-changing environment, while optimizing energy consumption as much as possible to ensure the continuity and reliability of the task. The collection of inspection environment information, including geospatial attributes and dynamic environmental parameters, in this step can refer to the inspection vehicle acquiring environmental data related to the inspection task in real time through various onboard sensors and communication devices. Geospatial attributes may include static spatial information such as the road topology, key node locations, road segment lengths, and slopes within the inspection area. Dynamic environmental parameters may include information such as real-time road condition categories, obstacle distribution density, and current weather conditions that change over time or in different contexts. This environmental information collectively constitutes the external basis for the inspection vehicle to make route decisions and assess energy consumption.
[0022] Meanwhile, the vehicle's current status information can refer to the real-time parameters of the inspection vehicle itself during operation, such as the remaining battery power, vehicle load weight, drive system efficiency, tire condition, etc. This status information can be used to reflect the inspection vehicle's current energy reserves and mobility.
[0023] Step 104: Construct a candidate path set based on the inspection environment information, and input the feature information of each road segment in the candidate path set into a pre-trained large model. The large model outputs the energy consumption judgment value or energy consumption distribution parameter of each road segment.
[0024] In the embodiments described in this specification, for inspection tasks, the inspection vehicle needs to travel safely and efficiently from its current location to the target location, during which it may encounter multiple different feasible routes. The candidate route set can refer to the set of all reasonable driving routes from the starting point to the destination planned based on known inspection environment information (such as road network and traffic restrictions).
[0025] After obtaining the candidate path set, each path can be further divided into continuous road segment units. Each road segment unit corresponds to a set of feature information describing its own attributes, such as the length, slope, road surface material condition, real-time number of obstacles, and current weather impact. In the technical solution of this application, these multi-dimensional road segment feature information are input into a large model that has been pre-trained with a large amount of data. This model can infer the energy consumption that a vehicle will generate when passing through the road segment based on the complex patterns and relationships it has learned. Its output is the energy consumption judgment value of the road segment, that is, a specific predicted energy consumption value; or, the output is energy consumption distribution parameters, such as the mean and variance used to represent the possible range and fluctuation of predicted energy consumption, thereby providing a quantitative basis for evaluating the energy consumption characteristics of different paths.
[0026] Step 106: Based on the energy consumption determination results output by the large model, construct a dynamic path cost function, and evaluate the cost of each candidate path based on the path cost function.
[0027] In the embodiments of this specification, the path cost function can refer to a mathematical evaluation model used to comprehensively quantify and compare the advantages and disadvantages of different driving routes. For the route decision of inspection vehicles, this function can integrate the performance of a route in multiple dimensions such as energy consumption, time required, and driving risk into a single comprehensive cost value, thereby providing a direct basis for selecting the most suitable route from multiple candidate routes.
[0028] Specifically, the dynamic path cost function in this application can be constructed based on the energy consumption judgment results of each road segment output by the large model. The core components of this function can include an energy consumption cost term that reflects the total predicted energy consumption of the entire path, a cost term that characterizes the time required for the vehicle to pass through the path, and a risk cost term that measures the uncertainty of path energy consumption. Each term is equipped with a dynamic weight parameter. These weights can be flexibly adjusted according to the real-time status of the inspection vehicle (such as the remaining battery power) and task requirements, so that the evaluation focus of the cost function can adapt to different scenarios.
[0029] Step 108: Select the candidate path with the lowest path cost from the candidate paths as the energy-optimal path, and control the inspection vehicle to perform the inspection task.
[0030] In the embodiments of this specification, the energy-optimal path refers to the driving route with the smallest dynamic path cost function calculation result among all candidate paths. For inspection tasks, selecting this path means that, after comprehensively weighing energy consumption, time cost, and driving risks, it is currently considered the most economical and reliable task execution plan. Specifically, in the technical solution of this application, after completing the comprehensive evaluation and quantitative ranking of all candidate paths through the dynamic path cost function, the path with the smallest cost value will be automatically selected as the execution route for this inspection. Subsequently, based on the spatial sequence and travel instructions of the selected path, corresponding vehicle control signals are generated to drive the physical inspection vehicle mechanism, such as the drive unit and steering unit, to begin autonomously driving along the route and executing the preset inspection operations, thereby transforming the previous intelligent planning results into actual vehicle movement and task execution.
[0031] Step 110: Collect the actual energy consumption data of the inspection vehicle when executing the energy-optimal path, and correct the output of the large model based on the deviation between the actual energy consumption data and the energy consumption judgment result output by the large model.
[0032] In the embodiments of this specification, actual energy consumption data refers to the numerical records collected in real time by the energy consumption monitoring unit on the inspection vehicle during its journey along the energy-optimal path, reflecting the actual energy consumed by the vehicle as it passes through each road segment. For the continuous optimization of the large model's energy consumption judgment capability, this energy consumption data from real-world operating scenarios can provide crucial feedback information for verifying prediction accuracy and driving model self-improvement.
[0033] In practical applications, the actual energy consumption of the inspection vehicle can be continuously recorded as it passes through each road segment during its mission. These measured data are then compared one by one with the energy consumption predictions previously generated by the larger model for the same road segment, and the specific deviation between the two is calculated. Based on this deviation, and through preset correction rules or algorithms, the internal parameters or output logic of the larger model can be adaptively adjusted, enabling subsequent energy consumption predictions to more accurately reflect actual operating conditions and achieving iterative improvements in model performance.
[0034] Step 112: When the energy consumption deviation is detected to exceed the preset threshold or the vehicle status does not meet the preset conditions, route replanning is triggered, and the energy consumption determination and route optimization steps based on the large model are re-executed to generate a new energy-optimal route.
[0035] In the embodiments described in this specification, the conditions for triggering route replanning can refer to preset judgment criteria used to initiate a new round of route decision-making procedures in order to ensure the continuity of inspection tasks and the safety of vehicle operation. For dynamically changing inspection environments and the vehicle's own state, setting such conditions can ensure that when the actual execution situation deviates significantly from expectations or the vehicle's energy reserves are insufficient, timely intervention and adjustment of the driving plan can be achieved.
[0036] In practical applications, two key indicators can be continuously monitored: the real-time energy consumption deviation calculated in the aforementioned steps and the current remaining battery power of the inspection vehicle. When the absolute value of the newly calculated energy consumption deviation for any road segment exceeds a preset fault tolerance threshold, or when the remaining battery power is detected to be below a preset safety lower limit, the replanning trigger condition is determined to be met. Once the condition is met, the reliance on the original path is immediately interrupted, and the entire optimization process—from environmental information collection and energy consumption determination based on a large model to path cost assessment and selection—is restarted. This generates a new energy-optimal path that better reflects the current situation, and the inspection vehicle is controlled to switch to this new path to continue performing its mission.
[0037] In this application's technical solution, the dynamic route optimization method for inspection vehicles does not rely solely on traditional fixed distance or time costs when generating the energy-optimal route. Instead, it first uses a large model to understand and calculate multi-dimensional features composed of environmental and vehicle states, outputting energy consumption judgment information for each road segment. Then, based on this judgment result, a path cost function integrating energy consumption, time, and risk is dynamically constructed, and the path with the minimum overall cost is selected as the execution basis. Because the large model can comprehensively analyze complex semantic information such as road conditions, slope, and weather, its accuracy in energy consumption judgment is significantly improved compared to traditional simplified models. Furthermore, during the actual execution of the inspection vehicle, the solution continuously collects real energy consumption data and compares and corrects it with the model's judgment values, while simultaneously monitoring key states and deviations. Once a significant deviation of the energy consumption prediction from reality or insufficient vehicle battery power is detected, a new round of optimization is immediately triggered. This closed-loop mechanism, based on high-precision semantic judgment of large models and combined with real-time feedback and condition triggering, enables path planning to closely match the dynamic environment and actual vehicle status. This allows for more accurate prediction and efficient utilization of the inspection vehicle's energy consumption, as well as more effective avoidance of mission interruption risks, thereby improving the overall reliability and economy of the inspection mission.
[0038] Based on the technical solutions described above, this specification also provides some specific implementation schemes, which are described below.
[0039] In an optional embodiment, the step of constructing a candidate path set based on the inspection environment information may specifically include: The inspection area is modeled as a directed graph. ,in, This represents a set of intersections or key inspection nodes. Represents a set of road segments. , Indicates from node To the node The section of road; Each section Corresponding to a set of basic feature vectors , is represented as: ;in, Indicates road segment Length, Indicates road segment The slope, Indicates road segment Road condition category, Indicates road segment obstacle density, Indicates road segment Weather conditions, This represents the vehicle state vector.
[0040] In this embodiment, the process of constructing a candidate path set begins with a digital abstract model of the inspection area. Specifically, the physical road network within the inspection area can be abstracted into a directed graph model. In this diagram, the set Each element in the set represents a road intersection or a critical node in an inspection task that must be visited; This represents the drivable road segments connecting these nodes, where each directed edge... It is clearly stated that from the node Drive to the node A specific road segment. Based on this graph model, path search algorithms (such as Dijkstra's algorithm or A* algorithm) can be used to generate multiple connected paths from the starting point to the ending point. The set of these paths constitutes a candidate path set for subsequent evaluation and selection.
[0041] To analyze each specific segment of the candidate paths To provide a more detailed description, a set of multi-dimensional basic feature vectors is defined in the technical solution of this embodiment. This vector can integrate environmental and vehicle status information, where The geometric length of a road segment is the basis for calculating travel distance; This indicates the average gradient of a road segment, which directly affects the work done by a vehicle against gravity. It can characterize road conditions, such as different surface states like smooth, potholes, or slippery. The density of obstacles on a given road segment reflects the smoothness of traffic flow and the likelihood of frequent starts and stops. Weather conditions, such as sunny or rainy days, affect the road surface adhesion coefficient and air resistance; while It is a comprehensive vehicle state vector that can contain real-time parameters such as remaining battery power and vehicle load. This structured feature vector can characterize the internal and external conditions faced by the vehicle when passing through the road segment, providing a complete input data foundation for subsequent large-scale models to perform accurate energy consumption semantic understanding and judgment.
[0042] In an optional embodiment, the step of outputting energy consumption determination values or energy consumption distribution parameters for each road segment from the large model may specifically include: The basic feature vector Input large model Obtain the road section Predicted energy consumption judgment value The calculation method is as follows: ,in, Indicates the section of road the inspection vehicle passed through. Predicted energy consumption; or, The large model Output section The energy consumption distribution parameters are expressed as: ,in, This represents the average of the predicted energy consumption. This represents the variance of the predicted energy consumption.
[0043] In the technical solution of this embodiment, the road segment can be described. Basic feature vectors of various conditions As input, it is fed into a pre-trained large model. This model, through complex internal nonlinear mapping and calculation, directly outputs a specific numerical value, namely the predicted energy consumption value. Their relationship can be expressed as Here This represents the optimal single-point prediction of the energy consumed by the inspection vehicle when passing through a specific road segment, after the model comprehensively considers all features such as length, slope, and road conditions. This value can provide a direct quantitative basis for the subsequent accumulation of total energy consumption and cost comparison.
[0044] In another alternative implementation, the large model can provide a measure of the uncertainty in energy consumption prediction. In this case, the model is targeted at road segments. The output is its energy consumption distribution parameters, specifically expressed as follows: This means that the model predicts the energy consumption of this road segment to follow a mean of . variance is It follows a normal distribution. The mean is... This represents the central estimate of predicted energy consumption, compared to the point estimate. The function is similar; while variance This quantifies the uncertainty or confidence level of the prediction. The larger the variance, the lower the model's confidence in predicting the energy consumption of the road section under given conditions.
[0045] In the optional embodiment technical solution, for candidate paths Its total predicted energy consumption The calculation method is as follows: ,in, Representing a path The Middle Predicted energy consumption values for each road segment.
[0046] In this embodiment, a candidate path is formed by the orderly connection of several continuous road segments. Its total predicted energy consumption The calculation is achieved through accumulation. Specifically, each segment in the path... Each corresponds to a predicted energy consumption value output by a large model based on the characteristic information of that road segment. This value represents the estimated energy consumption required for the inspection vehicle to traverse this single road segment. The total predicted energy consumption of the entire route is the sum of the predicted energy consumption values of all n road segments constituting the route, mathematically expressed as: .
[0047] In an optional embodiment, the dynamic path cost function Represented as: ;in, Representing a path The cost of travel time, Representing a path The cost of risk or uncertainty, , , These are dynamic weight parameters.
[0048] In this embodiment, a dynamic path cost function is used to evaluate and compare the merits of different candidate paths. It is constructed as a linear weighted combination, and the function is specifically expressed as ,in This represents the total predicted energy consumption of the path. This represents the time cost required for a vehicle to travel along this path. This represents the risk cost associated with the uncertainty of path energy consumption. The elements in the function are three dynamic weight parameters. , and These weights correspond to the relative importance of the three optimization objectives: energy consumption, time, and risk. These weights are not fixed but can be adaptively adjusted according to real-time circumstances (such as the vehicle's current battery level or the urgency of the task). In the technical solution of this embodiment, through the calculation of this cost function, each candidate path is mapped to a comprehensive cost value, thereby transforming the multi-objective path optimization problem into a comparable single-objective numerical optimization problem.
[0049] In an optional embodiment, correcting the output of the large model based on the deviation between the actual energy consumption data and the energy consumption determination result output by the large model may specifically include: When the inspection vehicle is executing the energy-optimal path, it passes through the following road sections. Actual energy consumption ; Calculate the actual energy consumption The predicted energy consumption value output by the large model Energy consumption deviation between The calculation method is as follows: ; Based on the energy consumption deviation The output of the large model is corrected, and the corrected predicted energy consumption value is determined. The calculation method is as follows: ;in, The preset learning rate parameter; The conditions for triggering path replanning include: The absolute value of the energy consumption deviation is greater than a preset threshold. That is, satisfying ;or, The remaining battery power of the inspection vehicle Below the preset battery threshold That is, satisfying .
[0050] In this embodiment, to improve the long-term accuracy of the large model in predicting energy consumption and enhance the system's adaptability, a feedback correction mechanism based on actual operating data is designed. The specific details of this mechanism are as follows: while the inspection vehicle travels along the selected optimal energy consumption path, onboard sensors collect real-time data on the vehicle's passage through specific road sections. The actual energy consumed during the process is recorded as the actual energy consumption. Subsequently, this actual measurement value was compared with the predicted energy consumption value output by the previous large model for this road segment. Compare the two and calculate the deviation between them. This bias quantifies the error in the model's prediction during the actual passage. Based on this bias, the model's output can be corrected online to generate an updated prediction value for that road segment. Its calculation formula is ,in It is a preset learning rate parameter used to control the magnitude and speed of correction. Through this continuous, incremental correction based on measured data, the energy consumption judgment capability of the large model can be continuously improved as execution experience accumulates, thereby gradually reducing the systematic deviation between the predicted values and the actual situation.
[0051] Furthermore, to ensure the safety and efficiency of the inspection task, this embodiment's technical solution sets clear path replanning trigger conditions. The first condition is the monitoring of the aforementioned energy consumption deviation, specifically, when the absolute value of the energy consumption deviation of any most recently calculated road segment reaches a certain threshold... Exceeding the preset tolerance threshold When this occurs, it indicates a significant discrepancy between the actual situation and the model predictions, suggesting that the original route's energy consumption estimates may no longer be reliable, thus triggering a replanning. The second condition is the monitoring of the vehicle's own energy state, specifically, when the remaining battery power of the inspection vehicle is detected... The power level dropped below the preset safe power threshold. This means that continuing to drive as originally planned may pose a risk of insufficient battery power. In order to ensure that the vehicle can return safely or complete the critical task, route replanning must be triggered immediately to find a new route that is safer and has more controllable energy consumption in the current low battery state.
[0052] In an optional embodiment, the dynamic weighting parameter The adjustment strategy is based on the vehicle's current remaining battery power. and task urgency indicators The adaptive determination method is as follows: Based on the remaining power range and the urgency of the task, the basic weight vector is obtained by looking up a table or through a fuzzy logic controller. ; Based on this, the total energy consumption is estimated according to the route. Uncertainty measurement Make minor adjustments; the formula is adjusted as follows: in, For the first Energy consumption variance of each road segment This represents the uncertainty factor.
[0053] In the technical solution of this embodiment, the dynamic weight parameter It is not fixed, but rather adaptively adjusted based on the real-time status of the inspection vehicle and the task requirements. The core basis of its adjustment strategy is the vehicle's current remaining battery power. And an urgency indicator that comprehensively reflects the timeliness requirements of the task. The specific determination process is divided into two stages: First, based on... The corresponding power range (e.g., high, medium, low) and The degree of urgency is represented by a set of matching basic weight vectors obtained by querying a preset weight configuration table or by using a well-designed fuzzy logic controller. This mechanism ensures that when there is sufficient power and the task is not demanding, the focus can be on optimizing energy consumption or time. When power is scarce or the task is urgent, the optimization focus can be dynamically shifted to ensuring battery life or improving execution speed.
[0054] After obtaining the basic weights, the uncertainty of the path energy consumption prediction itself can be further considered, and the weights can be fine-tuned. This is based on the estimated total energy consumption of the path. Uncertainty measurement This value represents the variance of energy consumption across all road segments in the path. The larger the variance, the lower the model's accuracy in predicting energy consumption along that path, and the higher the risk. Fine-tuning can be achieved using the following formula: as well as ,in This is a preset uncertainty factor. The purpose of this formula is to account for the uncertainty in path energy consumption prediction. When the value increases, the emphasis on the "predicted energy consumption" and "risk cost" items will automatically increase (i.e., increase the weight of these items). and This makes the final path selection more inclined to avoid routes with unreliable predictions and high potential risks. In this embodiment, through this two-stage adjustment mechanism, the path cost function can more intelligently and realistically reflect the current complex operating state and environmental uncertainties.
[0055] In optional embodiments, the risk or uncertainty cost The calculation can be performed using the following steps: Based on the energy consumption distribution parameters output by the large model Calculation path The total energy consumption distribution is ; Define risk cost To achieve a given confidence level The energy consumption condition risk value CVaR under the following conditions is calculated using the following formula: in, Indicates the total energy consumption distribution along the path. Inverse function under quantiles It represents the mathematical expectation.
[0056] In this embodiment, in order to reasonably quantify and avoid the operational risks that may be caused by energy consumption uncertainty in path planning, a risk or uncertainty cost is defined. The specific calculation method. This calculation is based on a large model of each road segment in the path. Output energy consumption distribution parameters That is, predicting energy consumption follows a certain pattern. For the mean, The variance follows a normal distribution. Due to the path... Composed of multiple independent road segments connected in series, its total predicted energy consumption is based on probability and statistics principles. It also follows a normal distribution, and its mean is the sum of the average energy consumption of each road segment. The variance is the sum of the energy consumption variances of each road segment. ,Right now This quantitative description allows the total energy consumption of a path to no longer be a fixed value, but a random variable with specific probabilistic characteristics, thus providing a mathematical basis for assessing the risk of energy consumption fluctuations.
[0057] Based on the probability distribution of the total energy consumption mentioned above, the risk cost Defined as at a given confidence level The Conditional Value at Risk (CVaR) is determined first. Specifically, the total energy consumption of the path is first determined as distributed across... Inverse function value under quantiles This value represents the value at the time when the value is less than 100%. The maximum threshold that energy consumption may reach under a given probability level. CVaR measures the conditional expected value when total energy consumption exceeds this threshold, i.e., the average energy overrun level under the worst-case scenario (tail risk). Its calculation formula is expressed as follows: ,in This represents the expected value in mathematics. In this embodiment, by introducing CVaR as a risk cost, the path optimization process can proactively take into account the extreme risk of energy consumption overruns, thereby guiding the system to select paths that are more robust in terms of energy consumption even under adverse conditions, which can enhance the safety and reliability of the inspection task.
[0058] In an optional embodiment, the optimization process of the path cost function is constructed as a constrained multi-objective optimization problem. The objective is to balance energy consumption, time, and risk while ensuring task completion time and power safety. Its mathematical expression is: in, The maximum time allowed for the task. This is the initial charge level. To ensure a safe minimum power consumption, Given an energy consumption budget, The probability is the safety threshold. By solving the Pareto front of this multi-objective optimization problem, and dynamically selecting the optimal compromise solution based on the current power level and task stage, the energy-optimal path is obtained.
[0059] In this embodiment, the optimization process of the path cost function is constructed as a constrained multi-objective optimization problem. The objective of this problem is to predict the total energy consumption of the path while ensuring task feasibility and vehicle operation safety. Estimated travel time and the cost of energy consumption uncertainty These three indicators are comprehensively weighed and synergistically optimized. The mathematical model includes the following constraints: firstly, the travel time of the path. The time must not exceed the maximum allowed time for the task. To meet timeliness requirements; secondly, from the initial power consumption... After deducting the predicted total energy consumption of the route, the remaining power must not be lower than the set safe power limit. This is the minimum power requirement to ensure the vehicle can safely return or continue operation; finally, a probabilistic safety constraint is introduced, namely, the actual total energy consumption of the route does not exceed the given energy budget. The probability must be higher than the set safety threshold. This constraint can clearly define the risk control requirements for energy consumption uncertainty.
[0060] To solve the aforementioned optimization problem involving multiple competing objectives, this embodiment employs a solution strategy based on the Pareto optimality concept. Through algorithmic solving, a set of Pareto optimal solutions is obtained. Each solution in this set represents a compromise among the three objectives that cannot be further optimized simultaneously; all these solutions constitute the Pareto front. Subsequently, instead of selecting a fixed solution, the optimal compromise solution best suited to the current state is dynamically selected from the Pareto front based on the vehicle's current real-time remaining battery power and the different stages of the task (e.g., the initial stage, middle stage, or near-end stage). This dynamic selection mechanism ensures that the ultimately determined energy-optimal path is not only theoretically a superior solution under multi-objective trade-offs but also adaptively matches the vehicle's actual energy status with the stage-specific requirements of the task.
[0061] It should be understood that in the methods described in one or more embodiments of this specification, the order of some steps may be adjusted according to actual needs, or some steps may be omitted.
[0062] Based on the foregoing technical solutions, the present invention also provides a dynamic path optimization device for inspection vehicles, such as... Figure 2 As shown, the device, from a macroscopic perspective, may include the following modules: Information acquisition module 202 is used to collect inspection environment information and vehicle current status information, including geospatial attributes and dynamic environmental parameters; The path modeling and energy consumption determination module 204 is used to construct a candidate path set based on the inspection environment information, and input the feature information of each road segment in the candidate path set into a pre-trained large model, and output the energy consumption determination value or energy consumption distribution parameter of each road segment from the large model. The path evaluation module 206 is used to construct a dynamic path cost function based on the energy consumption determination result output by the large model, and evaluate the cost of each candidate path based on the path cost function. The path selection and control module 208 is used to select the candidate path with the lowest path cost from the candidate paths as the energy-optimal path, and control the inspection vehicle to perform the inspection task. The feedback correction module 210 is used to collect the actual energy consumption data of the inspection vehicle when executing the energy-optimal path, and correct the output of the large model based on the deviation between the actual energy consumption data and the energy consumption judgment result output by the large model. The route replanning module 212 is used to trigger route replanning when the energy consumption deviation is detected to exceed a preset threshold or the vehicle status does not meet the preset conditions. This triggers the re-execution of the energy consumption determination and route optimization steps based on the large model to generate a new energy-optimal route.
[0063] Those skilled in the art will understand that the modules in the apparatus of the foregoing embodiments can be distributed in the apparatus of the embodiments as described in the embodiments, or they can be located in one or more devices different from this embodiment with corresponding changes. The modules of the above embodiments can be combined into one module, or they can be further divided into multiple sub-modules, that is, the module division can be flexibly performed to implement the method embodiments described above.
[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A dynamic path optimization method for inspection vehicles, characterized in that, include: Collect inspection environment information and vehicle current status information, including geospatial attributes and dynamic environmental parameters; Based on the inspection environment information, a candidate path set is constructed, and the feature information of each road segment in the candidate path set is input into a pre-trained large model. The large model outputs the energy consumption judgment value or energy consumption distribution parameter of each road segment. Based on the energy consumption determination results output by the large model, a dynamic path cost function is constructed, and the cost of each candidate path is evaluated based on the path cost function. The candidate path with the lowest path cost is selected from all candidate paths as the energy-optimal path, and the inspection vehicle is controlled to perform the inspection task. The actual energy consumption data of the inspection vehicle when executing the energy-optimal path is collected, and the output of the large model is corrected based on the deviation between the actual energy consumption data and the energy consumption judgment result output by the large model. When the energy consumption deviation is detected to exceed the preset threshold or the vehicle status does not meet the preset conditions, path replanning is triggered, and the energy consumption determination and path optimization steps based on the large model are re-executed to generate a new energy-optimal path.
2. The dynamic path optimization method for inspection vehicles as described in claim 1, characterized in that, The construction of the candidate path set based on the inspection environment information specifically includes: The inspection area is modeled as a directed graph. ,in, This represents a set of intersections or key inspection nodes. Represents a set of road segments. , Indicates from node To the node The section of road; Each section Corresponding to a set of basic feature vectors , is represented as: ;in, Indicates road segment Length, Indicates road segment The slope, Indicates road segment Road condition category, Indicates road segment obstacle density, Indicates road segment Weather conditions, This represents the vehicle state vector.
3. The dynamic path optimization method for inspection vehicles as described in claim 2, characterized in that, The energy consumption determination values or energy consumption distribution parameters of each road segment output by the large model specifically include: The basic feature vector Input large model Obtain the road section Predicted energy consumption judgment value The calculation method is as follows: ,in, Indicates the section of road the inspection vehicle passed through. Predicted energy consumption; or, The large model Output section The energy consumption distribution parameters are expressed as: ,in, This represents the average of the predicted energy consumption. This represents the variance of the predicted energy consumption.
4. The dynamic path optimization method for inspection vehicles as described in claim 3, characterized in that, For candidate paths Its total predicted energy consumption The calculation method is as follows: ,in, Representing a path The Middle Predicted energy consumption values for each road segment.
5. The dynamic path optimization method for inspection vehicles as described in claim 4, characterized in that, The dynamic path cost function Represented as: ;in, Representing a path The cost of travel time, Representing a path The cost of risk or uncertainty, , , These are dynamic weight parameters.
6. The dynamic path optimization method for inspection vehicles as described in claim 3, characterized in that, The step of correcting the output of the large model based on the deviation between the actual energy consumption data and the energy consumption determination result output by the large model specifically includes: When the inspection vehicle is executing the energy-optimal path, it passes through the following road sections. Actual energy consumption ; Calculate the actual energy consumption The predicted energy consumption value output by the large model Energy consumption deviation between The calculation method is as follows: ; Based on the energy consumption deviation The output of the large model is corrected, and the corrected predicted energy consumption value is determined. The calculation method is as follows: ;in, The preset learning rate parameter; The conditions for triggering path replanning include: The absolute value of the energy consumption deviation is greater than a preset threshold. That is, satisfying ;or, The remaining battery power of the inspection vehicle Below the preset battery threshold That is, satisfying .
7. The dynamic path optimization method for inspection vehicles as described in claim 5, characterized in that, The dynamic weight parameters The adjustment strategy is based on the vehicle's current remaining battery power. and task urgency indicators The adaptive determination method is as follows: Based on the remaining power range and the urgency of the task, the basic weight vector is obtained by looking up a table or through a fuzzy logic controller. ; Based on this, the total energy consumption is estimated according to the route. Uncertainty measurement Make minor adjustments; the formula is adjusted as follows: in, For the first Energy consumption variance of each road segment This represents the uncertainty factor.
8. The dynamic path optimization method for inspection vehicles as described in claim 5, characterized in that, The cost of the aforementioned risks or uncertainties Calculate using the following steps: Based on the energy consumption distribution parameters output by the large model Calculation path The total energy consumption distribution is ; Define risk cost To achieve a given confidence level The energy consumption condition risk value CVaR under the following conditions is calculated using the following formula: in, Indicates the total energy consumption distribution along the path. Inverse function under quantiles It represents the mathematical expectation.
9. The dynamic path optimization method for inspection vehicles as described in claim 5, characterized in that, The optimization process of the path cost function is constructed as a constrained multi-objective optimization problem. The objective is to balance energy consumption, time, and risk while ensuring task completion time and power safety. Its mathematical expression is: in, The maximum time allowed for the task. This is the initial charge level. To ensure a safe minimum power consumption, Given an energy consumption budget, The probability is the safety threshold. By solving the Pareto front of this multi-objective optimization problem, and dynamically selecting the optimal compromise solution based on the current power level and task stage, the energy-optimal path is obtained.
10. A dynamic path optimization device for an inspection vehicle, characterized in that, include: The information acquisition module is used to collect inspection environment information and vehicle current status information, including geospatial attributes and dynamic environmental parameters. The path modeling and energy consumption determination module is used to construct a candidate path set based on the inspection environment information, and input the feature information of each road segment in the candidate path set into a pre-trained large model, and the large model outputs the energy consumption determination value or energy consumption distribution parameter of each road segment. The path evaluation module is used to construct a dynamic path cost function based on the energy consumption determination results output by the large model, and to evaluate the cost of each candidate path based on the path cost function. The path selection and control module is used to select the candidate path with the lowest path cost from the candidate paths as the energy-optimal path, and control the inspection vehicle to perform the inspection task. The feedback correction module is used to collect the actual energy consumption data of the inspection vehicle when executing the energy-optimal path, and correct the output of the large model based on the deviation between the actual energy consumption data and the energy consumption judgment result output by the large model. The route replanning module is used to trigger route replanning when the energy consumption deviation is detected to exceed a preset threshold or the vehicle status does not meet the preset conditions. This triggers the re-execution of the energy consumption determination and route optimization steps based on the large model to generate a new energy-optimal route.