Unmanned aerial vehicle highway inspection method

Abstract

The invention relates to the technical field of traffic control systems, and discloses an unmanned aerial vehicle highway inspection method, which comprises the following steps: monitoring link quality parameters through a basic communication link between a dormant unmanned aerial vehicle and a roadside unit, calculating a communication entropy value of the link quality parameters, and triggering the unmanned aerial vehicle to accurately wake up to execute inspection when a phase change early warning condition is met; according to the method, traditional active inspection is converted into an event-driven passive response mechanism by utilizing internal association of traffic flow phase change and communication entropy, so that unmanned aerial vehicle resources are naturally focused at a traffic state critical point, and meanwhile, stable operation and efficient energy utilization of the system under complex working conditions are realized through environment adaptive compensation and responsibility area overlapping driving.

Description

TECHNICAL FIELD

[0001] The present application relates to a kind of unmanned aerial vehicle expressway inspection method, belong to traffic control system technical field. BACKGROUND

[0002] Currently, the fixed route unmanned aerial vehicle cruising mode is generally used in the field of expressway inspection, and its core logic is to periodically cover and scan the road network as a static space. Although this mode can obtain road condition snapshots, it is difficult to capture the key critical state of traffic flow dynamic evolution. For example, in the scenario of continuous accumulation of vehicle flow during Friday evening peak, when the vehicle spacing in the bottleneck section enters the nonlinear reduction stage, the existing unmanned aerial vehicle just completes the scanning of the area and marks it as normal. However, at this time, the traffic flow has entered the eve of the phase transition from free flow to synchronous flow, and a long congestion chain reaction will occur in the area within ten minutes. Traditional inspection systematically misses the hidden evolutionary trend by focusing on the dominant congestion state, thus missing the timely intervention window.

[0003] Specifically, the existing technology has three structural defects: 1. Uniform spatial coverage is taken as the inspection target, rather than responding to the internal evolution needs of traffic flow, resulting in 90% of the endurance being consumed in low-value stable sections; 2. Visible light video analysis lags behind the vehicle flow dynamics, and cannot capture the micro-disturbance signals of phase transition precursors; 3. Fixed frequency cruising causes the synchronization of energy depletion in the unmanned aerial vehicle cluster in the area of sudden disturbance, inducing a monitoring vacuum period.

[0004] Although the industry has tried to improve the identification accuracy by increasing the density of sensors, it has increased the system complexity and deployment cost, and still has not solved the core contradiction between static resource allocation and dynamic traffic demand. It is particularly worth noting that the wireless channel quality data generated by the existing Road Side Unit (RSU) is only used for communication operation and maintenance, and the value of the traffic flow disturbance information contained is completely ignored. This systematic neglect of cross-domain associated signals makes traffic state perception always stay in the surface dimension. Therefore, how to establish a real-time perception mechanism for traffic flow phase transition precursors and drive the unmanned aerial vehicle resources to respond accurately on demand, while optimizing the system energy consumption, has become a technical problem to be solved by the present application. SUMMARY

[0005] The present application provides an unmanned aerial vehicle expressway inspection method, which mainly aims to solve the problem of real-time perception of traffic flow phase transition critical state and dynamic adaptation of inspection resources.

[0006] To achieve the above-mentioned purpose, the present application provides an unmanned aerial vehicle expressway inspection method, comprising the following steps:

[0007] Step a: The unmanned aerial vehicle deployed along the expressway is in a low-power sleep state, and establishes and maintains a basic communication heartbeat link with at least one ground fixed infrastructure through the communication module on the unmanned aerial vehicle.

[0008] Step b, the ground fixed infrastructure continuously receives the heartbeat signal of the basic communication heartbeat link, and calculates a communication entropy value of the basic communication heartbeat link based on a time series of link quality parameters of the heartbeat signal, wherein the link quality parameters include a communication round-trip delay;

[0009] Step c, determining whether the communication entropy value meets a determined traffic flow phase transition early warning condition, the determined traffic flow phase transition early warning condition being that the communication entropy value is continuously higher than a baseline threshold dynamically adjusted according to historical data within a given time window;

[0010] Step d, if the determined traffic flow phase transition early warning condition is met, the ground fixed infrastructure sends a wake-up instruction to the unmanned aerial vehicle corresponding to the basic communication heartbeat link;

[0011] Step e, after receiving the wake-up instruction, the unmanned aerial vehicle switches from a low-power sleep state to a task execution state, conducts low-altitude reconnaissance on the physical area where the unmanned aerial vehicle is located, and transmits the reconnaissance information to the traffic control center in real time.

[0012] Preferably, in step c, the determined traffic flow phase transition early warning condition further includes dynamically compensating the baseline threshold according to an environmental factor representing the weather condition of the physical area where the unmanned aerial vehicle is located obtained from the ground fixed infrastructure, wherein the environmental factor includes visibility, and the dynamic compensation is realized by the following formula: wherein, represents an environment-adaptive wake-up threshold, represents a compensation coefficient, represents the baseline threshold.

[0013] Preferably, the compensation coefficient is determined segmentally according to the visibility: when the visibility is greater than or equal to five hundred meters, the compensation coefficient is one point zero; when the visibility is greater than or equal to two hundred meters and less than five hundred meters, the compensation coefficient is one point five; and when the visibility is less than two hundred meters, the compensation coefficient is two point five.

[0014] Preferably, in step a, the ground fixed infrastructure is a cooperative vehicle infrastructure system roadside unit.

[0015] Preferably, in step b, the calculation of the communication entropy value is completed on an edge computing unit of the cooperative vehicle infrastructure system roadside unit.

[0016] Preferably, in step a, the unmanned aerial vehicle is deployed on a charging parking apron along the expressway.

[0017] Preferably, in step e, the reconnaissance information includes video data and traffic situation confirmation information.

[0018] Preferably, after completing the inspection task, if the communication entropy value of the physical area where the unmanned aerial vehicle is located returns to below the baseline threshold, the unmanned aerial vehicle automatically returns to the low-power sleep state.

[0019] Preferably, when multiple unmanned aerial vehicles are woken up and in a task execution state, each of the multiple unmanned aerial vehicles broadcasts its remaining battery percentage information, calculates the spatial overlap rate between its task responsibility area and the task responsibility area of its neighboring unmanned aerial vehicle according to the position information and the remaining battery percentage information of the neighboring unmanned aerial vehicle, and judges whether a given rotation charging condition is met. If the given rotation charging condition is met, the unmanned aerial vehicle autonomously enters a charging process, and another unmanned aerial vehicle takes over its inspection task, thereby realizing continuous monitoring of the area where the communication entropy value meets the traffic flow phase change warning condition.

[0020] Preferably, the given rotation charging condition is that the spatial overlap rate between the task responsibility area of the unmanned aerial vehicle and the task responsibility area of the neighboring unmanned aerial vehicle is greater than a calibrated overlap threshold, and the remaining battery of the unmanned aerial vehicle is less than the average battery of the neighboring unmanned aerial vehicle minus a determined battery difference.

[0021] Compared with the prior art, the present application has the following advantages:

[0022] 1. By using the basic communication heartbeat link in the sleep state of the unmanned aerial vehicle, the time sequence uncertainty change of the link quality parameter is continuously monitored by the road side unit, the critical state of the traffic flow changing from free flow to synchronous flow is directly triggered to trigger the precise awakening of the unmanned aerial vehicle, the mechanism converts the time and space resource consumption of the traditional active inspection into passive response to the endogenous signal of the traffic system, and under the premise of maintaining the original communication architecture, the unmanned aerial vehicle resources are guided to the key nodes of the traffic state evolution.

[0023] 2. When the communication link is disturbed due to extreme weather, the existing meteorological sensing data of the road network is reused to quantitatively map the environmental factors, and the warning threshold of the communication entropy is dynamically corrected by a lightweight segmentation model. The mechanism automatically decouples the traffic flow phase change signal and the weather interference artifacts in the decision logic, maintains the core warning sensitivity, avoids additional perception load, and maintains the judgment reliability of the system through the combination of the communication mechanism and the environmental perception.

[0024] 3. In the multi-unmanned aerial vehicle cooperative operation scene, the responsibility area spatial overlap rate is autonomously calculated by broadcasting the battery and position information, the rotation charging behavior based on local rules is triggered, when the battery difference of the unmanned aerial vehicles in the high overlap area reaches a preset threshold, the low battery node autonomously exits and wakes up the standby node to replace it, so that the cluster monitoring coverage rate and the energy consumption rate form an asynchronous balance in the spatial coupling area, and the system resilience is established through the interactive calculation of the geometric relationship and the energy state.

[0025] 4, the communication entropy monitoring of the dormant UAV, the environmental compensation calculation of the road side unit and the energy relay triggered by the responsibility area overlap, jointly constitute a closed loop response chain, the physical layer fluctuation of the communication link maps the traffic state change, the meteorological data dynamically corrects the decision threshold, the geometric overlap relationship guides the energy relay, and the three form a decentralized coupling in the edge computing layer, which makes the perception, decision and execution links in the traditional inspection integrated into an adaptive system, and the resource mismatch problem in the prior art is effectively alleviated. BRIEF DESCRIPTION OF DRAWINGS

[0026] Fig. 1 A schematic diagram of the highway inspection method system architecture of the unmanned aerial vehicle of the present application;

[0027] Fig. 2 A curve graph of the correspondence between the traffic flow state and the communication entropy value of the present application;

[0028] Fig. 3 A signaling timing diagram of the highway inspection method of the unmanned aerial vehicle of the present application.

[0029] The purpose of the present application, the functional characteristics and the advantages will be further described with reference to the accompanying drawings. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical scheme and advantages of the present application more clear, the technical scheme of the present application will be described in detail, and it should be noted that the embodiments are only part of the embodiments of the present application but not all, and all equivalent modifications or replacements based on the embodiments without departing from the spirit and essence of the present application should be covered in the protection scope of the present application.

[0031] The system architecture of the unmanned aerial vehicle highway inspection method disclosed in the application starts from the unmanned aerial vehicle deployed along the highway entering a low-power sleep state, and a basic communication heartbeat link is established and maintained between the unmanned aerial vehicle and the ground-fixed vehicle-road cooperative roadside unit through the built-in communication module of the unmanned aerial vehicle. The edge computing unit of the roadside unit is configured to continuously process the heartbeat signals on the link, extract and analyze the time series of link quality parameters, and then calculate the communication entropy value reflecting the disturbance degree of the traffic flow. The communication entropy value is used for continuous comparison with a dynamically determined traffic flow phase change warning condition. Once the condition is met, the roadside unit immediately sends a wake-up instruction to the corresponding unmanned aerial vehicle, triggering the unmanned aerial vehicle to switch from the sleep state to the task execution state, and the unmanned aerial vehicle carries out low-altitude reconnaissance on the target area and returns the real-time reconnaissance information to the traffic control center. In the multi-unmanned aerial vehicle cooperative operation scene, the system also includes a set of decentralized rotation charging mechanism. Through the broadcast of power and position information between unmanned aerial vehicles, the task responsibility area overlap rate is autonomously calculated to trigger the low-power unmanned aerial vehicle to automatically return to charge and be replaced by the adjacent unmanned aerial vehicle, thereby realizing continuous monitoring of the key area and efficient use of system energy as a whole.

[0032] In the practice of highway traffic management, the core challenge is not to respond to the formed congestion, but to prospectively identify the critical state of traffic flow transition from free flow to congestion flow. This state is macroscopically manifested as a non-linear reduction of vehicle spacing while the vehicle speed has not yet significantly decreased. The traditional inspection mode relying on video analysis cannot capture the micro-disturbance signals in this stage due to its lagging nature. To address this challenge, the unmanned aerial vehicle deployed at the charging parking lot along the highway is preset to a low-power sleep state, and its communication module only maintains a basic communication heartbeat link with the nearest vehicle-road cooperative roadside unit. The link sends heartbeat signals at a very low data packet rate, for example, once per second. The edge computing unit of the roadside unit continuously collects the link quality parameters of the heartbeat signals, especially the time series data of the communication round-trip delay. Then, based on a fixed-length time sliding window, for example, the past 300 consecutive communication round-trip delay sampling points, the edge computing unit quantifies the fluctuation uncertainty of the time series by applying the information entropy calculation formula, and obtains a real-time communication entropy value. This entropy value directly reflects the stability of the wireless channel environment. When the road traffic increases, the wireless signal multipath effect and channel obstruction caused by the interaction between vehicles increase, the jitter and randomness of the communication round-trip delay also increase, leading to a significant increase in the calculated communication entropy value. Thus, the evolution trend of the micro-dynamics of the traffic flow is mapped to a communication physical layer parameter that can be accurately quantified.

[0033] To ensure the reliability of the wake-up mechanism and avoid false triggering caused by instantaneous communication interference, the determination of the traffic flow phase change warning condition is designed as a double-checking process. First, a baseline threshold The threshold is not a fixed value, but is dynamically adapted to the traffic mode by statistical analysis of the historical communication entropy value data of the specific road section in the same period, taking a certain quantile of the statistical distribution, for example, taking the mean plus three standard deviations, to form a baseline; secondly, the triggering rule of the warning condition is defined as: in a given time window, for example, thirty seconds, the calculated communication entropy value is continuously higher than the baseline threshold This setting aims to filter out accidental signal jumps and ensure that the captured entropy value is systematically lifted with statistical significance caused by the continuous evolution of traffic flow state. Only when this condition is met will the roadside unit send a wake-up instruction to the target UAV. This mechanism combines passive response with statistical process control, enabling the UAV resources to be automatically guided to the key nodes where the traffic state has truly changed. However, the complexity of the highway environment means that the disturbance sources of the communication link are not limited to traffic flow itself. Adverse weather such as heavy fog and heavy rain can also cause signal attenuation and instability, thereby producing high communication entropy artifacts similar to traffic congestion precursors. If not distinguished, it will lead to unnecessary wake-up of the UAV, significantly consuming system energy. In view of this, the system integrates an environmental adaptive compensation procedure. The procedure reuses the environmental factors representing the weather conditions of the physical area where the UAV is located obtained by the existing environmental sensors of the roadside unit, especially the visibility data, and dynamically compensates the baseline threshold according to the data. The compensation logic is implemented by the formula wherein, is the environmental adaptive wake-up threshold, and the compensation coefficient is determined by segmenting according to the visibility: when the visibility is greater than or equal to five hundred meters, it is determined that the environmental impact on communication can be ignored, the value is one point zero; when the visibility is greater than or equal to two hundred meters and less than five hundred meters, it is determined that there is a moderate impact, the value is one point five; when the visibility is less than two hundred meters, it is determined that there is a severe impact, the value is two point five. In this way, the system realizes the automatic decoupling of traffic flow signals and weather interference artifacts in the decision logic, ensuring the reliability of the wake-up decision under complex working conditions without adding additional sensing hardware.

[0034] When the UAV receives the wake-up instruction, its system state is switched from the low-power sleep state to the task execution state, the physical area corresponding to the source of the instruction, the roadside unit, is surveyed at low altitude, and the survey information, including high-definition video data and traffic situation confirmation information that structurally describes key events such as traffic congestion or accidents, is transmitted in real time to the traffic control center. After the inspection task is completed, the UAV does not immediately return, but continues to monitor the communication entropy value of the area it is in. If the value stabilizes below the baseline threshold within a certain period of time, it indicates that the traffic disturbance has been resolved, and the UAV automatically returns to the charging parking lot and enters the low-power sleep state, realizing a complete closed-loop response and resource recovery process. Further, when dealing with traffic events that last for a long time or have a wide range of impact, multiple UAVs often need to work together. This raises the challenge of how to avoid a vacuum period in the monitoring area due to the concentrated depletion of power by UAVs. To this end, the system has a set of decentralized rotation charging protocols based on local rules. When multiple UAVs are awakened and in the task execution state, each UAV will broadcast its precise position and remaining power percentage information to the neighboring UAVs within its communication range. At the same time, each UAV calculates the spatial overlap rate between its task responsibility area and the received position information of neighboring UAVs in real time, and determines whether the given rotation charging condition is met. The given rotation charging condition is set to meet two sub-conditions simultaneously: first, the spatial overlap rate between the UAV's own task responsibility area and the task responsibility area of a neighboring UAV is greater than a calibrated overlap threshold. This threshold can be determined in advance through a combination of offline simulation and field testing, based on the field of view of the sensors mounted on the UAV and the standard inspection height, for example, set to fifty percent. Second, the UAV's own remaining power is less than the average power of all neighboring UAVs in the overlapping area minus a determined power difference. The difference is intended to reserve sufficient power margin for the return and charging process, and its value is calibrated according to the discharge curve of the UAV battery model and the distance to the charging facility, for example, set to fifteen percent. Once the rotation charging condition is met, the UAV with lower power automatically detaches from the current task and enters the return and charging process, while the area it was originally responsible for is seamlessly taken over by a neighboring UAV with higher power that overlaps with its responsibility area, thereby achieving asynchronous balance between continuous monitoring capability of key areas and cluster energy consumption rate, and improving the resilience and endurance of the entire inspection system.

[0035] Example 1: During an evening rush hour, on a highway in a city, three kilometers before an exit ramp, traffic volume continuously increases over time. Unmanned aerial vehicles (UAVs) deployed along this section are all in a low-power sleep state. Their communication modules maintain a basic communication heartbeat link with the corresponding vehicle-to-infrastructure (V2I) roadside units. The edge computing units of the roadside units continuously record and analyze the time series of communication round-trip delays. Initially, traffic is in a free-flow state, vehicle spacing is large, and the communication channel is stable. The calculated communication entropy value is within a dynamically set baseline threshold. The flow remained relatively stable initially; however, as the time approached 5:30 PM, the vehicle density entering this section reached a critical point. Frequent lane changes and acceleration / deceleration among vehicles began to cause microscopic disturbances in the wireless channel environment, leading to an increase in the randomness of communication round-trip delay. At this moment, the communication entropy value calculated by the roadside unit began to show a continuous unidirectional increase, gradually approaching the baseline threshold. Just then, the weather in the area changed, and a thin layer of fog began to spread. The environmental sensors of the roadside unit detected that the visibility had dropped from one kilometer to about four hundred meters. The environmental adaptive compensation procedure within the system was then activated, and the compensation coefficient was adjusted according to the segmentation rules corresponding to the visibility. The value was determined to be 1.5, and then the environmentally adaptive wake-up threshold was calculated. This process quantifies and removes communication disturbances introduced by weather factors from the decision-making criteria, aiming to balance the sensitivity and reliability of the inspection system.

[0036] Subsequently, despite the wake-up threshold being dynamically increased, the communication entropy continued to climb due to the ongoing deterioration of traffic flow conditions, eventually remaining above a certain level for a given time window. , the traffic flow phase transition warning condition is formally triggered, and the roadside unit immediately sends a wake-up instruction to the dormant drone in the area; after receiving the instruction, the drone enters the task execution state and conducts low-altitude reconnaissance of the bottleneck section. The real-time video data and traffic situation confirmation information show that the vehicle speed on this section has begun to decrease synchronously, and the vehicle aggregation effect is obvious, which is on the eve of forming a long-chain congestion. The traffic control center obtains the warning window accordingly and immediately takes intervention measures such as upstream ramp signal lamp timing adjustment and variable information sign release of detour prompt. The intervention of this mechanism is not through visual analysis of the congestion form, but through monitoring of the communication physical layer entropy value, changing the nature of the inspection task from identifying the lagging event of formed congestion to perceiving the critical state of the traffic system on the brink of instability, thereby moving the intervention time from passive post-congestion response to active pre-congestion diversion. Finally, during the entire evolution of the traffic event, due to the timely intervention of the intervention measures, the bottleneck section does not form the expected serious congestion, but maintains a low-speed but continuous flow synchronous flow state, avoiding the occurrence of traffic paralysis. After the evening peak, the communication entropy value of the section returns to below the baseline threshold, and the drone that completes the task automatically returns to the low-power dormant state. The energy consumption of the entire system and the deployment of the inspection resources are driven by the endogenous state of the traffic flow, rather than fixed cruising plans. The system couples the three originally fragmented technical links of physical layer fluctuations of communication links, dynamic compensation of environmental factors, and on-demand wake-up of drones to form a self-adaptive closed-loop response chain, enabling the inspection resources to achieve precise matching with traffic risks in terms of spatial and temporal distribution.

[0037] Example 2: To objectively verify the internal correlation between communication entropy value and evolution of highway traffic flow state, and to test the effectiveness of the environmental adaptive compensation mechanism under complex working conditions, this embodiment constructs a hardware-in-the-loop simulation test platform. The platform includes an entity vehicle infrastructure cooperation roadside unit as a ground fixed infrastructure, an entity drone communication module as a basic communication heartbeat link terminal, and a software system integrating traffic flow simulation and wireless channel simulation. The traffic flow simulation software is used to generate vehicle motion trajectories under different density and speed distributions, and the wireless channel simulation software dynamically calculates the signal propagation path loss, multipath effect and Doppler shift between the drone and the roadside unit according to the real-time position and motion state of the vehicle, and injects the signal superimposed with the corresponding channel influence into the entity communication transceiver device, thereby reproducing the communication physical layer dynamics in the highway scene in a controllable and repeatable laboratory.

[0038] In the experiment, the setting of several key parameters follows a clear engineering logic. The setting of the sampling period of the communication round-trip delay takes into account the balance between the tracking ability of the channel dynamic changes and the processing load of the edge computing unit. Considering that the coherence time of the channel caused by the movement of vehicles in the highway scenario is usually in the order of milliseconds, in order to avoid signal aliasing and capture the details of the channel changes, the sampling frequency needs to be significantly higher than the characteristic frequency of the channel changes. However, too high a frequency will bring unnecessary computational overhead. In this experiment, the sampling period is set to 10 milliseconds, i.e., the sampling frequency is 100 Hz. The length of the time window used to calculate the communication entropy value needs to be balanced between the stationarity of the statistical results and the timeliness of the state change detection. A too short window is easily affected by random noise, and a too long window will smooth out the details of the state changes, causing early warning delays. In this experiment, the window length is set to 300 sampling points, i.e., 3 seconds. This duration is long enough to include several channel slow fading periods, ensuring the statistical stability of the entropy value calculation, while having sufficient response speed for minute-level traffic state evolution.

[0039] The first phase of the experiment aims to verify the representation ability of the communication entropy value for the phase transition process of the traffic flow. The simulation program starts from a low-density free-flow state, and then continuously increases the number of vehicles entering the simulated road section at a set rate until the traffic flow enters the congested synchronous flow state. In this process, the roadside unit continuously records the changes of the communication entropy value and compares it with the macroscopic traffic parameters output by the traffic flow simulator. Table 1 shows the key state data points recorded in this phase. As shown in Table 1, in the free-flow state with low traffic density, the communication entropy value maintains a low level with limited fluctuations. When the traffic density exceeds the critical value and the traffic flow begins to phase transition to the synchronous flow, the communication entropy value starts to rise sharply and continuously, and is significantly higher than the value in the free-flow state, even if the macroscopic average vehicle speed has not yet decreased dramatically. After entering the congestion state, the entropy value tends to saturate at a high level. The internal mechanism of this phenomenon lies in the fact that as the vehicle density increases, the probability and complexity of the wireless signal propagation path between the unmanned aerial vehicle and the roadside unit being blocked and reflected by vehicles are greatly increased, resulting in the distribution of the communication round-trip delay evolving from a narrow Gaussian distribution to a wide distribution with multiple peaks and long tails, which increases the uncertainty. The communication entropy directly quantifies the uncertainty of this distribution, thereby achieving sensitive capture of the micro-dynamics of the traffic flow.

[0040] Table 1: Correspondence table of traffic flow state and communication entropy value.

[0041]

[0042] The second phase of the experiment focuses on the effectiveness verification of the environmental adaptive compensation mechanism. The experiment selects a state with medium and stable traffic density, at which the communication entropy value is stable at 2.10. The baseline threshold is set to 2.20, which is 10% higher than the baseline value. As shown in Table 2, the communication entropy value of the test group is significantly lower than that of the control group, and the average value is 2.06, which is 4.76% lower than the baseline value. This result indicates that the environmental adaptive compensation mechanism can effectively reduce the communication entropy value and improve the communication quality in the test group. is set to 2.5, on this basis, the channel simulation software is configured to simulate a dense fog event with sharp visibility reduction, which is realized by injecting additional fixed attenuation and phase noise into the channel model, Table 2 is the core data record at this stage, see Table 2, when the compensation mechanism is closed, the channel deterioration caused by dense fog leads to the communication entropy value jumping to 3.1, which exceeds the baseline threshold and generates a wake-up trigger; as a control group, the environmental adaptive compensation mechanism is turned on while keeping all other conditions unchanged, after the roadside unit receives the simulated environmental factor with visibility less than 200 meters, it automatically calculates the compensation coefficient is 2.5, and the wake-up threshold is dynamically adjusted to At this time, although the communication entropy value also jumps to 3.1, it does not reach the adjusted wake-up threshold , so the system does not generate a wake-up instruction.

[0043] Table 2: Environmental adaptive compensation mechanism effectiveness verification table.

[0044]

[0045] The data shows that by monitoring the communication entropy value of the basic communication heartbeat link, a precursor index with high sensitivity to the critical state of traffic flow from free flow to synchronous flow phase transition can be established, at the same time, by introducing a wake-up threshold dynamic compensation mechanism based on environmental factors, the decline in communication link quality caused by traffic state evolution and severe weather can be distinguished, thereby reducing the false positive rate of the system while ensuring that critical events are not missed.

[0046] Embodiment 3: This embodiment combines Figs. 1 to 3 to realize a kind of unmanned aerial vehicle highway inspection method. As Fig. 1As shown, the drones deployed along the highway (in hibernation mode) are in a low-power hibernation state and establish and maintain a basic communication heartbeat link with the ground-based roadside units (RSUs) through their built-in communication modules. The edge computing units of the RSUs continuously process the heartbeat signals on this link, extract and analyze the time series of link quality parameters, and then calculate the communication entropy value reflecting the degree of traffic flow disturbance, and determine whether the warning conditions are met. If the determined traffic flow phase change warning conditions are met, the RSUs immediately send a wake-up command to the corresponding drone. After receiving the wake-up command, the drone emerges from low-power hibernation. The system switches to mission execution mode, conducts low-altitude reconnaissance of the physical area where the drone is located, and transmits the reconnaissance information back to the traffic control center in real time. In multi-drone collaborative operation scenarios, the system also includes a decentralized rotation charging mechanism. Drones in mission execution mode broadcast their own power and location information. When the rotation charging mechanism is triggered, the drone autonomously enters the charging process, and another drone takes over its inspection task. This enables continuous monitoring of areas where the communication entropy value meets the traffic flow phase change early warning conditions. After the drone completes its mission and the communication entropy value returns to normal, it automatically returns to the charging dock and enters hibernation mode.

[0047] like Fig. 2 As shown in the figure, this diagram illustrates the relationship between simulation time (s) and communication entropy (nats), where the communication entropy is represented by a solid line and the baseline threshold by a dashed line. In the first phase of the simulation experiment, the aim was to verify the ability of communication entropy to characterize the traffic flow phase transition process. The simulation program started from a low-density free-flow state, and then the number of vehicles continuously increased until the traffic flow entered a congested synchronous flow state. During this process, the roadside unit continuously recorded the changes in communication entropy and compared them synchronously with the macroscopic traffic parameters output by the traffic flow simulator. The data shows that in the low-density free-flow state, the communication entropy remains at a low level with limited fluctuations. When the traffic density exceeds the threshold... At the critical point, when traffic flow begins to transition to synchronous flow, even if the macroscopic average vehicle speed has not yet decreased drastically, the communication entropy value has already begun to rise sharply and continuously, significantly higher than the value in the free flow state. After entering the congested state, the entropy value tends to saturate at a high level. This phenomenon indicates that as vehicle density increases, the probability and complexity of the wireless signal propagation path between drones and roadside units being blocked and reflected by vehicles increase significantly. This causes the distribution of communication round-trip delay to gradually evolve from a narrow band distribution that is approximately Gaussian to a broadband distribution with multi-peak and long-tail characteristics. Its uncertainty increases, and communication entropy is the direct quantification of this uncertainty, thereby enabling sensitive capture of the micro-dynamics of traffic flow.

[0048] like Fig. 3As shown, firstly, the UAV is in a low-power sleep state and continuously sends heartbeat signals. The roadside unit monitors the heartbeat and transmits the link quality parameters to the edge computing unit. After receiving the link quality parameters, the edge computing unit calculates the communication entropy value and determines whether the warning conditions are met. If the communication entropy value is continuously higher than the threshold, a wake-up decision is triggered. The roadside unit sends a wake-up command to the UAV. After receiving the wake-up command, the UAV switches to the mission execution state, performs low-altitude reconnaissance, and sends the reconnaissance information, including the transmitted video data and the transmitted traffic situation information, to the traffic control center in real time. The traffic control center receives and processes the inspection information.

[0049] Example 4: To ensure stable and reliable operation of the method of the present invention under different geographical locations, hardware configurations, and traffic environments, its deployment process includes a systematic parameter self-calibration and algorithm initialization procedure. This procedure aims to eliminate uncertainties in the setting of key parameters and transform the initial configuration process of the system into a deterministic engineering step. When deploying this system on a newly built highway section, the primary technical challenge is how to set an effective set of initial parameters for calculating communication entropy, judging traffic flow phase change early warning conditions, and multi-UAV collaborative operation protocols in the absence of long-term operational data. To address this challenge, before the system is officially put into operation, the specific calculation process for communication entropy is first standardized and defined, and the edge computing unit deployed on the vehicle-road cooperative roadside unit is... The unit is configured to execute a four-stage processing pipeline: The first stage is data buffering and smoothing. The edge computing unit stores the raw round-trip delay data of the received basic communication heartbeat link into a fixed-size circular buffer and applies a moving average filter with a window size of five to the data in the buffer to suppress isolated noise points introduced by instantaneous jitter in the communication network. The second stage is the construction of a statistical distribution. The processing unit discretizes the continuous round-trip delay values ​​into several data intervals using a histogram statistical method on a continuously updated time sliding window (i.e., 300 latest smoothed sampling points) and calculates the frequency of data points appearing in each interval, thus obtaining a probability distribution characterizing the current channel state. The third stage is entropy calculation. Based on this probability distribution, the entropy is calculated using the Shannon entropy formula. Calculate the communication entropy value for quantizing channel uncertainty, where For the data point to fall on the 1st The probability of each interval; the fourth stage is data reporting, where the entropy value is recorded at a frequency of one hertz along with the timestamp and used for subsequent judgment.

[0050] Then, the system initiates a seven-day offline learning and model building cycle for baseline thresholding. During the adaptive calibration period, the UAVs maintain a dormant state and do not perform wake-up, while all the roadside units continuously record the communication entropy values of each time period to form an initial historical database; after the learning period, the system processes the data in the database, divides each day into ninety-six time segments with a duration of fifteen minutes, and for each time segment, calculates the average of all corresponding communication entropy values in the past seven days and the standard deviation , then the initial baseline threshold value of the time segment is determined as ; after the system enters formal operation, the baseline threshold model is not static, but is updated daily. The newly generated data each day is included in the historical database, and the average and standard deviation of the corresponding time segment are updated in a weighted manner, so that the baseline threshold can continuously track and adapt to the long-term evolution of the traffic mode caused by seasonal or road network structure changes; for the given replacement charging condition in the multi-UAV cooperative operation, the determination of the spatial overlap rate overlap threshold and the power difference also follows a set of calibration procedures based on hardware specifications and operating environment. The overlap threshold of the spatial overlap rate is calculated through geometric relationships according to the field of view of the reconnaissance equipment mounted on the UAV, the legal minimum safe inspection height, and the redundancy range required to deal with unstable factors such as crosswinds, to ensure that the monitoring area of any UAV returning to charge can be seamlessly and completely covered by the adjacent UAV. The purpose of the power difference calibration is to ensure that any UAV initiating replacement charging has sufficient power to safely return to its charging parking lot. The calibration process is as follows: during the deployment phase, the UAV is tested for maximum range with full power, and the percentage of power consumed to return directly from the farthest end of the responsibility area to the charging parking lot is recorded. Multiply this percentage by a preset safety factor, which is greater than one, and the specific value is determined by the comprehensive risk level of the operating area. In a typical configuration, the safety factor can be taken as 1.5. Through this systematic initialization and parameter calibration, the core decision logic of the inspection method of the present application is based on the objective data quantification of the current environment and device performance in any specific implementation scenario.

[0051] Embodiment 5: Before the method of the present application is scaled up, the accuracy and applicability of the unified and enhanced environmental adaptive compensation mechanism need to be established and enhanced by establishing a decentralized mutual monitoring network. Before the method of the present application is scaled up, a standardized environmental impact feature database needs to be established to unify and enhance the accuracy and applicability of the environmental adaptive compensation mechanism. The construction procedure is as follows: in a controlled electromagnetic shielding laboratory, a set of benchmark unmanned aerial vehicle communication modules and vehicle-road cooperative roadside units are deployed, and a stable and measurable basic communication heartbeat link is established between them. Then, through special equipment, a single or combined environmental stress is systematically applied to the communication link, including using a high-precision spraying system to simulate different intensities of rainfall, using a temperature and humidity control and aerosol generator to simulate different densities of fog, and using an adjustable wind tunnel to simulate the disturbance of different wind speeds on the antenna posture. Under each combination of environmental stress, the offset of the communication entropy value is continuously recorded, and the offset is associated with the corresponding environmental sensor readings to generate a lookup table or regression model that maps the multi-dimensional environmental factor vector to the optimal compensation coefficient. This database is finally solidified into the edge computing unit of all deployed roadside units, so that the compensation of environmental factors is no longer limited to visibility, but is based on a more comprehensive and empirically verified physical model.

[0052] To ensure the reliability of system decision-making during long-term operation, each roadside unit is equipped with a periodic self-checking and consistency checking protocol. This protocol is configured to be automatically executed during the period of lowest daily traffic flow. During the execution period, each roadside unit cross-compares the real-time communication entropy value calculated by itself with the communication entropy values calculated by its adjacent roadside units upstream and downstream at the same time section. In a statistical sense, the communication entropy values of adjacent road sections in an extremely low traffic state should be highly correlated and distributed in a very small confidence interval. If the absolute value of the difference between the mean value of the entropy values reported by a particular roadside unit and the mean value of the synchronous values of adjacent units for consecutive multiple checking periods is continuously greater than a dynamic tolerance threshold determined based on historical data, the system will mark the sensing or calculation module of the roadside unit as a potential abnormal state. Once marked, the unmanned aerial vehicle wake-up decision-making authority of the roadside unit will be automatically suspended, and a maintenance request will be sent to the traffic control center, thereby achieving health state monitoring and early warning of fault risks for key nodes of the system.

[0053] ​Embodiment 6: Before applying the UAV highway patrol method of the present application to a specific highway network, a set of standardized deployment optimization and operating parameter calibration procedures need to be performed to ensure that the system performance matches the geographical and traffic characteristics of the road network; the first step of this procedure is the site optimization of the UAV charging parking lot, which is not equally deployed along the highway, but based on geographic information systems, the digital map of the road section, the effective patrol radius calculated by the UAV based on its battery capacity and safety margin, and the congested bottleneck sections identified through historical accident data and road geometric characteristics, are jointly analyzed as inputs, and through an optimization algorithm, a set of deployment points that can achieve full coverage of all high-risk road sections with the least number of UAVs is solved, and the spatial overlap rate between adjacent UAV task responsibility areas is not less than the preset overlap threshold, the output of this set is the final construction coordinates of each charging parking lot.

[0054] The second step of the procedure is to localize the time window parameters in the traffic flow phase change early warning condition during the initial learning period after system deployment. During the aforementioned seven-day offline learning period, the system will build a baseline threshold for the communication entropy value of each time period At the same time, it will also record all potential events where the entropy value continuously breaks through the preliminary threshold, and analyze the time length distribution between the threshold breaking and the significant deterioration of the traffic flow macroscopic parameters; by statistically analyzing this time distribution, taking its specific quantile, an optimal warning time window can be determined for the road section, which can effectively filter out benign traffic fluctuations with too short duration, and ensure that the warning can provide the required reaction time for the intervention of traffic control measures.

[0055] The third step of the procedure is to set the autonomous fault tolerance protocol of the UAV when it encounters extreme communication interference during task execution. When the UAV is awakened and tries to return the reconnaissance information to the traffic control center, but the high-bandwidth video data link fails to establish due to unpredictable strong interference, the UAV will automatically switch to the preset low-bandwidth confirmation mode; in this mode, the UAV no longer attempts to transmit video, but flies along the predetermined route over the target area, uses the on-board processor to analyze the collected images in real time to determine whether there is a condition where vehicles are below a preset speed threshold for a long period of time, and only sends back a data packet containing its own position information and congestion status confirmation information through a more robust low-power basic communication heartbeat link, after completing the core information transmission, the UAV automatically returns and reports the event code of the communication link anomaly, this mechanism ensures that even in a severe electromagnetic environment, the system can still complete the basic traffic situation confirmation task, and avoids the complete failure of the patrol function due to communication interruption.

[0056] It is apparent for a person skilled in the art that the application is not limited to the details of the above-described exemplary embodiments, but that the application can be implemented in other concrete forms without departing from the spirit or essential characteristics of the application.

[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not to limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for highway inspection by a UAV, characterized in that, The method comprises the following steps: Step a, the unmanned aerial vehicle deployed along the highway is in a low-power sleep state, and establishes and maintains a basic communication heartbeat link with at least one ground fixed infrastructure through a communication module on the unmanned aerial vehicle; Step b, the ground fixed infrastructure continuously receives heartbeat signals of the basic communication heartbeat link, and calculates a communication entropy value of the basic communication heartbeat link based on a time sequence of link quality parameters of the heartbeat signals, wherein the link quality parameters include communication round-trip delay; Step c, it is judged whether the communication entropy value meets a determined traffic flow phase change early warning condition, and the determined traffic flow phase change early warning condition is that the communication entropy value is continuously higher than a baseline threshold dynamically adjusted according to historical data within a given time window; Step d, if the determined traffic flow phase change early warning condition is met, the ground fixed infrastructure sends a wake-up instruction to the unmanned aerial vehicle corresponding to the basic communication heartbeat link; Step e, after receiving the wake-up instruction, the unmanned aerial vehicle switches from the low-power sleep state to a task execution state, conducts low-altitude reconnaissance on the physical region where the unmanned aerial vehicle is located, and returns reconnaissance information to the traffic control center in real time.

2. The method of claim 1, wherein, In step c, the determined traffic flow phase transition warning condition further comprises dynamically compensating the baseline threshold according to an environmental factor characterizing weather conditions of the physical area where the UAV is located, acquired from ground fixed infrastructure, wherein the environmental factor comprises visibility, and the dynamic compensation is achieved by the following formula: wherein, represents the environment-adaptive wake-up threshold, represents the compensation coefficient, represents the baseline threshold.

3. The method of claim 2, wherein, compensation coefficient The value of the compensation coefficient is determined in segments based on visibility: when visibility is greater than or equal to 500 meters, the compensation coefficient is... The compensation coefficient is 1.0; when the visibility is greater than or equal to 200 meters and less than 500 meters, the compensation coefficient is... The compensation coefficient is 1.5; when visibility is less than 200 meters, the compensation coefficient is... It is 2.

5.

4. The method of claim 1, wherein, In step a, the ground fixed infrastructure is a cooperative vehicle infrastructure system roadside unit.

5. The method of claim 4, wherein, In step b, the calculation of the communication entropy value is completed on an edge computing unit of the cooperative vehicle infrastructure system roadside unit.

6. The method of claim 1, wherein, In step a, the unmanned aerial vehicle is deployed on a charging parking apron along the highway.

7. The method of claim 1, wherein, In step e, the reconnaissance information includes video data and traffic situation confirmation information.

8. The method of claim 1, wherein, After completing the inspection task, if the communication entropy value of the physical region where the unmanned aerial vehicle is located returns to below the baseline threshold, the unmanned aerial vehicle automatically returns to the low-power sleep state.

9. The method of claim 1, wherein, When multiple unmanned aerial vehicles are woken up and in the task execution state, each of the multiple unmanned aerial vehicles broadcasts its remaining battery percentage information, calculates a spatial overlap rate between its task responsibility areas according to the position information and remaining battery percentage information of its neighboring unmanned aerial vehicles, and judges whether a given rotation charging condition is met, if the given rotation charging condition is met, the unmanned aerial vehicle autonomously enters a charging process, and another unmanned aerial vehicle takes over its inspection task.

Images