Aerocar autonomous obstacle avoidance system based on vehicle-road cloud coordination

Abstract

The invention relates to a hovercar autonomous obstacle avoidance system based on vehicle-road cloud cooperation. The system comprises a vehicle end system, a road end system, a cloud end system and a system integration platform which work cooperatively. The vehicle end system comprises vehicle end hardware configuration and a single vehicle obstacle avoidance module, and the vehicle end hardware configuration comprises a vehicle-mounted calculation unit, a vehicle end sensor array and a V2X module; the road end system comprises a road end hardware configuration and a perception enhancement module, and the road end hardware configuration comprises a road side computing unit and a road end sensor array; the cloud system comprises a cloud hardware configuration and multi-vehicle cooperation module, and is realized through a vehicle-road-cloud three-level architecture: vehicle end deployment improves an artificial potential field algorithm, a dynamic potential field model and an emergency gradient field are fused, and local decision delay is reduced; the road end eliminates a sensing blind area through a high-precision radar array and a meteorological compensation model; the cloud constructs airspace digital twin bodies, performs multi-vehicle trajectory conflict prediction, dynamically optimizes an airspace traffic strategy, and improves low-altitude airspace traffic density.

Description

Technical Field

[0001] This invention relates to an autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud collaboration, specifically an autonomous obstacle avoidance system for flying cars that integrates vehicle-mounted edge computing, 5G-V2X communication and cloud-based collaborative control. It is applicable to multi-agent collaborative traffic management scenarios in urban low-altitude airspace and belongs to the field of intelligent transportation and low-altitude travel technology. Background Technology

[0002] The existing obstacle avoidance systems for flying cars have the following defects: (1) Limited perception: Traditional millimeter-wave radar has a blind zone rate of >40% in rainy and foggy weather, and lidar has a point cloud missing rate of ≥15% due to strong light interference. (2) Decision delay: In dynamic obstacle scenarios, the perception-planning time delay link is too long, resulting in a minimum obstacle avoidance distance of ≥400m when driving at high speed, which seriously restricts the utilization rate of low-altitude airspace. (3) Lack of a unified cloud scheduling mechanism: When obstacles suddenly appear, the single-vehicle potential field algorithm cannot predict the trajectory of multiple vehicles, making it difficult to cope with sudden group obstacle threats and easily leading to airspace conflicts. Summary of the Invention

[0003] To overcome the aforementioned technical bottlenecks faced by existing flying cars during low-altitude cruise, this invention provides an autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud collaboration. This system is implemented through a three-tiered vehicle-road-cloud architecture: The vehicle deploys an improved artificial potential field algorithm, integrating a dynamic potential field model with an emergency gradient field to reduce local decision-making latency; the roadside uses a high-precision radar array and a weather compensation model to eliminate perception blind spots; and the cloud constructs an airspace digital twin to perform multi-vehicle trajectory conflict prediction, dynamically optimize airspace traffic strategies, and increase low-altitude airspace traffic density.

[0004] An autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud cooperation includes a vehicle-end system, a road-end system, a cloud system, and a system integration platform that work together.

[0005] The vehicle-side system includes vehicle-side hardware configuration and a single-vehicle obstacle avoidance module. The vehicle-side hardware configuration includes an on-board computing unit, a vehicle-side sensor array, and a V2X module. The vehicle-side sensor array includes a lidar, a 4D millimeter-wave radar, and a TOF depth camera.

[0006] The roadside system includes roadside hardware configuration and perception enhancement module. The roadside hardware configuration includes a roadside computing unit and a roadside sensor array. The roadside sensor array includes a 4D millimeter-wave radar, a meteorological sensor, a temperature / humidity sensor, and a GPS / IMU positioning module.

[0007] The cloud system includes cloud hardware configuration and a multi-vehicle collaboration module;

[0008] The system integration platform includes a human-computer interaction module and a data management module.

[0009] The human-machine interaction module includes a driver interaction interface, a roadside monitoring interface, and a cloud management platform interface.

[0010] Driver interface. Augmented reality head-up display technology projects target point navigation arrows, obstacle locations, TTC time, and vertical climb commands onto the windshield; the instrument panel displays engine speed, speed, altitude, and power status, with audible and visual alarms in case of abnormalities; it supports voice commands, and the system interprets and provides key information.

[0011] Roadside monitoring interface. The digital twin 3D map displays the roadside coverage area, and the panoramic screen presents the perception situation; the device status panel displays the online status, latency, and calibration parameters of the sensors, and abnormal devices are highlighted in red to indicate maintenance.

[0012] The cloud-based management platform interface includes a GIS map displaying real-time low-altitude airspace trajectories, conflict hotspots, and traffic strategies; a multi-level response panel showing risk value distribution and vehicle handling status; and historical data queries supporting filtering of cases by time, region, and risk level, as well as viewing logs.

[0013] The data management module includes a data acquisition and processing module, a data fusion and analysis module, and a data security and sharing module.

[0014] Data acquisition and processing. Vehicle-side data acquisition includes LiDAR point clouds, 4D millimeter-wave radar velocity, TOF images, and vehicle status; roadside data acquisition includes networked millimeter-wave radar echoes, meteorological data, and roadside calibration parameters; cloud-based data acquisition includes multi-vehicle trajectory prediction, digital twin status, and system logs. Data undergoes real-time cleaning and processing to generate "perception-decision" correlated data.

[0015] Data fusion and analysis. By combining Kalman filtering and deep learning algorithms, data from vehicle-mounted LiDAR and roadside millimeter-wave radar are fused to improve obstacle localization accuracy; models are trained based on historical conflict data to predict high-risk areas, supporting cloud-based strategy optimization.

[0016] Data security and sharing. Tracking and privacy data are transmitted using the national standard SM4 encryption to prevent tampering; role-based access control is implemented. API integration with air traffic control and meteorological departments allows for uploading tracks and receiving meteorological data to optimize models.

[0017] The single-vehicle obstacle avoidance module introduces a velocity direction derivative by improving the gravitational field function in the artificial potential field method and introduces a collision risk index by introducing a repulsive field function. It triggers a vertical climb command by using an emergency gradient field when the TTC is less than a set threshold.

[0018] Artificial potential field method, a classic intelligent algorithm in path planning, achieves autonomous obstacle avoidance and path navigation for flying cars by simulating the gravitational and repulsive forces in a physical potential field. This invention improves upon the traditional artificial potential field method by considering the vertical obstacle avoidance algorithm of the flying car obstacle avoidance system and introducing an emergency gradient field function. When the TTC (Total Traction Time) is less than a set threshold, a vertical climb command is triggered, solving the failure problem of traditional horizontal obstacle avoidance in space-constrained or high-speed approach scenarios.

[0019] The improved artificial potential field method specifically includes:

[0020] Gravitational field function introduces partial derivatives in velocity direction Solving the oscillation problem of traditional APF in narrow channels; introducing a collision risk index e into the repulsive field function. -α·TTC Traditional repulsive field By relying solely on distance, the dynamic collision risk is transformed into repulsive force intensity, achieving an exponentially enhanced repulsive force against high-speed approaching obstacles and automatically weakening the repulsive force interference against obstacles far away.

[0021] Gravitational field function:

[0022]

[0023] In the formula, K is the proportionality coefficient of the gravitational potential field; P0 is the straight-line distance between the driving vehicle and the target point; The directional derivative of the velocity;

[0024] Repulsive field function:

[0025]

[0026] In the formula, ||P g || represents the real-time distance between the flying car and the obstacle, in meters; ρ represents the maximum range of the repulsive force exerted by the obstacle on the unmanned vehicle (default value: 200m); K rep α is the repulsion ratio coefficient (range: 0.5~2.0); α is the risk sensitivity factor (key parameter: α=0.8); TTC is the collision time (Time To Collision), calculated as follows:

[0027] The emergency gradient field specifically includes:

[0028] Acceleration from 10 m / s² is achieved through multi-stage coordination of the power system. 2 Up to 30m / s 2 Precise control solves the problem of failure of traditional horizontal obstacle avoidance in space-constrained or high-speed approach scenarios;

[0029] Emergency gradient field function:

[0030]

[0031] In the formula, g is the acceleration due to gravity; The direction of velocity is z; when TTC < 5s, vertical climb is forcibly triggered.

[0032] The perception enhancement module uses a meteorological compensation model to couple and model the three parameters of rainfall intensity, dust concentration, and visibility, thereby reducing compensation errors and improving roadside perception accuracy. It also integrates environmental sensor data through a roadside perception calibration mechanism to perform multi-dimensional attenuation compensation and distance inversion on radar echoes, calibrating detection values ​​to eliminate environmental interference and improve perception accuracy in complex scenarios.

[0033] The meteorological compensation model reduces compensation errors and improves roadside sensing accuracy by coupling three parameters: rainfall intensity, dust concentration, and visibility. The core formula of the meteorological compensation model is:

[0034]

[0035] In the formula, R raw k represents the raw echo intensity of the millimeter-wave radar. r is the rainfall attenuation coefficient, with a value of 0.015; RI is the rainfall intensity, in mm / h; k d is the dust attenuation coefficient, with a value of 0.02; DC is the dust concentration, in μg / m³. 3 MOR stands for meteorological optical distance, measured in meters (m). base The baseline visibility is set at 10000m.

[0036] The roadside perception calibration mechanism takes the original radar echo as input and integrates environmental parameters such as rainfall, dust concentration, and visibility. It eliminates environmental interference from the radar echo through three modules: rain attenuation compensation, dust attenuation compensation, and visibility compensation, generating a compensated echo. Based on the characteristics of the compensated signal, a distance inversion algorithm is performed to accurately analyze the actual distance information of the target object. Through cross-validation between the distance inversion results and the original radar detection values, a closed-loop calibration model is constructed to dynamically correct the radar detection data. This improves the target detection accuracy, distance measurement reliability, and scene adaptability of the roadside perception system under complex weather conditions, ensuring consistency between the perception results and the real traffic environment.

[0037] The multi-vehicle collaboration module uses a risk quantification model to calculate a comprehensive risk value based on the distance between aircraft and time to collision (TTC), combined with spatial distance sensitivity factors. This quantifies the collision risk level in multi-vehicle collaboration and uses a graded response mechanism to divide the risk value into four levels: monitoring, fine-tuning, replanning, and takeover. It matches the gradient execution time to achieve dynamic risk management of multi-level collaboration in the cloud.

[0038] The risk quantification model is based on the distance-to-collision-time (TTC) between aircraft, combined with a spatial distance sensitivity factor, to calculate a comprehensive risk value that quantifies the collision risk level in multi-vehicle collaboration. The core formula is:

[0039]

[0040] In the formula, β is a spatial distance sensitivity factor with a value of 0.1; The distance between the i-th pair of aircraft; TTC i Let i be the collision time of the i-th pair;

[0041] The hierarchical response mechanism specifically includes:

[0042] When the risk value exceeds the set threshold, global path replanning is triggered. When the risk value is in the range of 0-0.3, monitoring measures are implemented: when the risk value is in the range of 0.3-0.7, path fine-tuning suggestions are implemented with an execution time of ≤200ms; when the risk value is >0.7, replanning is forcibly executed with an execution time of ≤50ms; when the risk value is >0.9, control is taken over and airspace is urgently isolated with an execution time of ≤20ms.

[0043] The collaborative decision-making design principles for vehicle-side systems, road-side systems, cloud-based systems, and system integration platforms are as follows: (1) Dynamic allocation of responsibilities, dynamically adjusting decision-making leadership based on risk level; (2) Time sensitivity guarantee, enabling the shortest communication path in emergency scenarios; (3) Redundancy verification mechanism, with the cloud performing millisecond-level arbitration for high-risk decisions.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0045] This invention achieves efficient obstacle avoidance and airspace management for flying cars in complex low-altitude environments through a three-level collaborative architecture of vehicle-road-cloud: the vehicle deploys a dynamic potential field model and an emergency gradient field, combined with 5G ultra-low latency communication, to shorten local decision-making response time; the roadside eliminates perception blind spots through a weather compensation model and 4D radar networking; and the cloud-based digital twin dynamically predicts multi-vehicle trajectory conflicts and implements hierarchical risk control, thereby improving low-altitude airspace traffic density and multi-aircraft collaborative safety.

[0046] This invention improves the artificial potential field method by introducing a velocity direction partial derivative into the gravitational field function. Solving the oscillation problem of traditional APF in narrow channels; introducing a collision risk index e into the repulsive field function. -α·TTC Traditional repulsive field By relying solely on distance, the dynamic collision risk is transformed into repulsive force intensity, achieving an exponentially enhanced repulsive force against high-speed approaching obstacles and automatically weakening the repulsive force interference against obstacles far away.

[0047] The meteorological compensation model of this invention introduces rainfall intensity, dust concentration, and visibility parameters on the basis of the original echo intensity of millimeter-wave radar. Through three-parameter coupling modeling, the compensation error is reduced and the perception accuracy is improved.

[0048] The vehicle-road-cloud collaborative architecture for obstacle avoidance of flying cars in this invention achieves a comprehensive effect of reducing decision-making latency, eliminating perception blind spots, and improving airspace efficiency through cross-layer optimization of dynamic potential field model (vehicle-side), meteorological perception fusion (road-side), and digital twin prediction (cloud-side). Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a system composition diagram of the autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud collaboration of the present invention.

[0051] Figure 2 This diagram illustrates the roadside perception calibration mechanism of the autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud collaboration, as described in this invention.

[0052] Figure 3 This diagram illustrates the cloud-based hierarchical response mechanism of the autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud collaboration, as described in this invention.

[0053] Figure 4 This is a flowchart of the vehicle-road-cloud collaborative decision-making process for the autonomous obstacle avoidance system of flying cars based on vehicle-road-cloud collaboration of the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Reference Figure 1 A flying car autonomous obstacle avoidance system based on vehicle-road-cloud cooperation includes a vehicle-end system, a road-end system, a cloud system, and a system integration platform that work together.

[0056] The vehicle-side system includes vehicle-side hardware configuration and a single-vehicle obstacle avoidance module. The vehicle-side hardware configuration includes an on-board computing unit (computing power ≥20 TOPS, memory latency ≤80ns).

[0057] The vehicle-mounted sensor array includes: a 1550nm lidar (300m@10% reflectivity), a 4D millimeter-wave radar (velocity resolution 0.1m / s), and a TOF depth camera (anti-glare ≥120dB).

[0058] 5G-V2X module: Operating frequency band n78 (3.5GHz) + n79 (4.9GHz) dual-carrier aggregation, air interface latency: ≤1ms.

[0059] The single-vehicle obstacle avoidance module introduces a velocity direction derivative by improving the gravitational field function in the artificial potential field method and introduces a collision risk index by introducing a repulsive field function. It triggers a vertical climb command by using an emergency gradient field when the TTC is less than a set threshold.

[0060] Artificial potential field method, a classic intelligent algorithm in path planning, achieves autonomous obstacle avoidance and path navigation for flying cars by simulating the gravitational and repulsive forces in a physical potential field. This invention improves upon the traditional artificial potential field method by considering the vertical obstacle avoidance algorithm of the flying car obstacle avoidance system and introducing an emergency gradient field function. When the TTC (Total Traction Time) is less than a set threshold, a vertical climb command is triggered, solving the failure problem of traditional horizontal obstacle avoidance in space-constrained or high-speed approach scenarios.

[0061] The improved artificial potential field method specifically includes:

[0062] Gravitational field function introduces partial derivatives in velocity direction Solving the oscillation problem of traditional APF in narrow channels; introducing a collision risk index e into the repulsive field function. -α·TTC Traditional repulsive field By relying solely on distance, the dynamic collision risk is transformed into repulsive force intensity, achieving an exponentially enhanced repulsive force against high-speed approaching obstacles and automatically weakening the repulsive force interference against obstacles far away.

[0063] Gravitational field function:

[0064]

[0065] In the formula, K is the proportionality coefficient of the gravitational potential field; P0 is the straight-line distance between the driving vehicle and the target point; The directional derivative of the velocity;

[0066] Repulsive field function:

[0067]

[0068] In the formula, ||P g || represents the real-time distance between the flying car and the obstacle, in meters; ρ represents the maximum range of the repulsive force exerted by the obstacle on the unmanned vehicle (default value: 200m); K repα is the repulsion ratio coefficient (range: 0.5~2.0); α is the risk sensitivity factor (key parameter: α=0.8); TTC is the collision time (Time To Collision), calculated as follows:

[0069] The emergency gradient field specifically includes:

[0070] Acceleration from 10 m / s² is achieved through multi-stage coordination of the power system. 2 Up to 30m / s 2 Precise control solves the problem of failure of traditional horizontal obstacle avoidance in space-constrained or high-speed approach scenarios;

[0071] Emergency gradient field function:

[0072]

[0073] In the formula, g is the acceleration due to gravity; The direction of velocity is z; when TTC < 5s, vertical climb is forcibly triggered.

[0074] The roadside system includes roadside hardware configuration and perception enhancement module. The roadside hardware configuration includes a roadside computing unit (computing power ≥20TOPS, memory latency ≤80ns) and a roadside sensor array. The roadside sensor array includes 4D millimeter-wave radar, meteorological sensors (rain gauge, dust concentration sensor, visibility meter), temperature / humidity sensor, and GPS / IMU positioning module.

[0075] The perception enhancement module uses a meteorological compensation model to couple and model the three parameters of rainfall intensity, dust concentration, and visibility, thereby reducing compensation errors and improving roadside perception accuracy. It also integrates environmental sensor data through a roadside perception calibration mechanism to perform multi-dimensional attenuation compensation and distance inversion on radar echoes, calibrating detection values ​​to eliminate environmental interference and improve perception accuracy in complex scenarios.

[0076] The meteorological compensation model reduces compensation errors and improves roadside sensing accuracy by coupling three parameters: rainfall intensity, dust concentration, and visibility. The core formula of the meteorological compensation model is:

[0077]

[0078] In the formula, R raw k represents the raw echo intensity of the millimeter-wave radar. r is the rainfall attenuation coefficient, with a value of 0.015; RI is the rainfall intensity, in mm / h; k d is the dust attenuation coefficient, with a value of 0.02; DC is the dust concentration, in μg / m³. 3 MOR stands for meteorological optical distance, measured in meters (m).base The baseline visibility is set at 10000m.

[0079] Reference Figure 2 The roadside perception calibration mechanism takes the original radar echo as input and integrates environmental parameters such as rainfall, dust concentration, and visibility. It eliminates environmental interference from the radar echo through three modules: rain attenuation compensation, dust attenuation compensation, and visibility compensation, generating a compensated echo. Based on the characteristics of the compensated signal, a distance inversion algorithm is performed to accurately analyze the actual distance information of the target object. By cross-validating the distance inversion results with the original radar detection values, a closed-loop calibration model is constructed to achieve dynamic correction of the radar detection data. This improves the target detection accuracy, distance measurement reliability, and scene adaptability of the roadside perception system under complex weather conditions, ensuring the consistency between the perception results and the real traffic environment.

[0080] The cloud system includes cloud hardware configuration (GPU nodes: NVIDIA A100×8; CPU nodes: Xeon Platinum 83380×16, capable of handling a cluster of 5000 flying cars) and a multi-vehicle collaboration module;

[0081] The multi-vehicle collaboration module uses a risk quantification model to calculate a comprehensive risk value based on the distance between aircraft and time to collision (TTC), combined with spatial distance sensitivity factors. This quantifies the collision risk level in multi-vehicle collaboration and uses a graded response mechanism to divide the risk value into four levels: monitoring, fine-tuning, replanning, and takeover. It matches the gradient execution time to achieve dynamic risk management of multi-level collaboration in the cloud.

[0082] The risk quantification model is based on the distance-to-collision-time (TTC) between aircraft, combined with a spatial distance sensitivity factor, to calculate a comprehensive risk value that quantifies the collision risk level in multi-vehicle collaboration. The core formula is:

[0083]

[0084] In the formula, β is a spatial distance sensitivity factor with a value of 0.1; The distance between the i-th pair of aircraft; TTC i Let i be the collision time of the i-th pair;

[0085] Reference Figure 3 The hierarchical response mechanism specifically includes:

[0086] When the risk value exceeds the set threshold, global path replanning is triggered. When the risk value is in the range of 0-0.3, monitoring measures are implemented: when the risk value is in the range of 0.3-0.7, path fine-tuning suggestions are implemented with an execution time of ≤200ms; when the risk value is >0.7, replanning is forcibly executed with an execution time of ≤50ms; when the risk value is >0.9, control is taken over and airspace is urgently isolated with an execution time of ≤20ms.

[0087] The system integration platform includes a human-computer interaction module and a data management module.

[0088] The human-machine interaction module includes a driver interaction interface, a roadside monitoring interface, and a cloud management platform interface.

[0089] Driver interface. Augmented reality head-up display technology projects target point navigation arrows, obstacle locations, TTC time, and vertical climb commands onto the windshield; the instrument panel displays engine speed, speed, altitude, and power status, with audible and visual alarms in case of abnormalities; it supports voice commands, and the system interprets and provides key information.

[0090] Roadside monitoring interface. The digital twin 3D map displays the roadside coverage area, and the panoramic screen presents the perception situation; the device status panel displays the online status, latency, and calibration parameters of the sensors, and abnormal devices are highlighted in red to indicate maintenance.

[0091] The cloud-based management platform interface includes a GIS map displaying real-time low-altitude airspace trajectories, conflict hotspots, and traffic strategies; a multi-level response panel showing risk value distribution and vehicle handling status; and historical data queries supporting filtering of cases by time, region, and risk level, as well as viewing logs.

[0092] The data management module includes a data acquisition and processing module, a data fusion and analysis module, and a data security and sharing module.

[0093] Data acquisition and processing. Vehicle-side data acquisition includes LiDAR point clouds, 4D millimeter-wave radar velocity, TOF images, and vehicle status; roadside data acquisition includes networked millimeter-wave radar echoes, meteorological data, and roadside calibration parameters; cloud-based data acquisition includes multi-vehicle trajectory prediction, digital twin status, and system logs. Data undergoes real-time cleaning and processing to generate "perception-decision" correlated data.

[0094] Data fusion and analysis. By combining Kalman filtering and deep learning algorithms, data from vehicle-mounted LiDAR and roadside millimeter-wave radar are fused to improve obstacle localization accuracy; models are trained based on historical conflict data to predict high-risk areas, supporting cloud-based strategy optimization.

[0095] Data security and sharing. Tracking and privacy data are transmitted using the national standard SM4 encryption to prevent tampering; role-based access control is implemented. API integration with air traffic control and meteorological departments allows for uploading tracks and receiving meteorological data to optimize models.

[0096] Reference Figure 4 The collaborative decision-making design principles for vehicle-side systems, road-side systems, cloud systems and system integration platforms are as follows: (1) Dynamic allocation of responsibilities, dynamically adjusting the decision-making leadership according to the risk level; (2) Time sensitivity guarantee, using the shortest communication path in emergency scenarios; (3) Redundancy verification mechanism, the cloud performs millisecond-level arbitration for high-risk decisions.

[0097] An embodiment is provided to demonstrate the above technical solution.

[0098] (1) Scene setting:

[0099] Flying Car A: Cruises at a speed of 80m / s at an altitude of 100m, with the target point 5000m due north.

[0100] Static obstacles: 300m to the west and 400m to the north (200m tall commercial building).

[0101] Dynamic obstacle group: Drone B: 180m east, 320m north (60m / s westward); Flying car C: 150m southwest, 110m altitude (40m / s northeastward).

[0102] Weather conditions: Moderate rain (RI = 25 mm / h), visibility MOR = 3000 m.

[0103] (2) System end-to-end response process:

[0104]

[0105]

[0106] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. A flying car autonomous obstacle avoidance system based on vehicle-road-cloud cooperation, characterized in that: This includes vehicle-side systems, roadside systems, cloud systems, and system integration platforms that work together in a coordinated manner. The vehicle-side system includes vehicle-side hardware configuration and a single-vehicle obstacle avoidance module. The vehicle-side hardware configuration includes an on-board computing unit, a vehicle-side sensor array, and a V2X module. The roadside system includes roadside hardware configuration and perception enhancement module. The roadside hardware configuration includes a roadside computing unit and a roadside sensor array. The cloud system includes cloud hardware configuration and a multi-vehicle collaboration module; The system integration platform includes a human-machine interaction module and a data management module. The human-machine interaction module includes a driver interaction interface, a roadside monitoring interface, and a cloud management platform interface. The data management module includes a data acquisition and processing module, a data fusion and analysis module, and a data security and sharing module.

2. The autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud cooperation according to claim 1, characterized in that: The single-vehicle obstacle avoidance module introduces a velocity direction derivative by improving the gravitational field function in the artificial potential field method and introduces a collision risk index by introducing a repulsive field function. It triggers a vertical climb command by using an emergency gradient field when the TTC is less than a set threshold.

3. The autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud cooperation according to claim 2, characterized in that: The improved artificial potential field method specifically includes: Gravitational field function introduces partial derivatives in velocity direction The repulsive field function introduces a collision risk index e. -α·TTC ; Gravitational field function: In the formula, K is the proportionality coefficient of the gravitational potential field; P0 is the straight-line distance between the driving vehicle and the target point; The directional derivative of the velocity; Repulsive field function: In the formula, ||P g || represents the real-time distance between the flying car and the obstacle, in meters; ρ represents the maximum range of the repulsive force exerted by the obstacle on the unmanned vehicle; K rep α is the repulsion force proportionality coefficient; α is the risk sensitivity factor; TTC is the collision time, calculated as follows: The emergency gradient field specifically includes: Precise control of acceleration enhancement is achieved through multi-level coordination of the power system; Emergency gradient field function: In the formula, g is the acceleration due to gravity; The direction of velocity is z; when TTC < 5s, vertical climb is forcibly triggered.

4. The autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud cooperation according to claim 1, characterized in that: The perception enhancement module uses a meteorological compensation model to couple and model the three parameters of rainfall intensity, dust concentration and visibility, thereby improving the accuracy of roadside perception. It also uses a roadside perception calibration mechanism to fuse environmental sensor data, perform multi-dimensional attenuation compensation and distance inversion on radar echoes, and calibrate the detection values ​​to eliminate environmental interference and improve perception accuracy in complex scenarios.

5. The autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud cooperation according to claim 4, characterized in that: The meteorological compensation model specifically includes: The core formula of the meteorological supplementary model is: In the formula, R raw k represents the raw echo intensity of the millimeter-wave radar. r RI is the rainfall attenuation coefficient; RI is the rainfall intensity, in mm / h; k d DC is the dust attenuation coefficient; DC is the dust concentration, in μg / m³. 3 MOR stands for meteorological optical distance, measured in meters (m). base Baseline visibility: The roadside perception calibration mechanism takes the original radar echo as input and integrates environmental parameters such as rainfall, dust concentration, and visibility. It eliminates environmental interference from the radar echo through three modules: rain attenuation compensation, dust attenuation compensation, and visibility compensation, generating a compensated echo. Based on the characteristics of the compensated signal, a distance inversion algorithm is performed to accurately analyze the actual distance information of the target object. By cross-validating the distance inversion results with the original radar detection values, a closed-loop calibration model is constructed to achieve dynamic correction of the radar detection data. This improves the target detection accuracy, distance measurement reliability, and scene adaptability of the roadside perception system under complex weather conditions, ensuring the consistency between the perception results and the real traffic environment.

6. The autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud cooperation according to claim 1, characterized in that: The multi-vehicle collaboration module uses a risk quantification model to calculate a comprehensive risk value based on the distance between aircraft and the collision time, combined with spatial distance sensitivity factors. This quantifies the collision risk level in multi-vehicle collaboration and uses a graded response mechanism to divide the risk value into four levels: monitoring, fine-tuning, replanning, and takeover. It matches the gradient execution time to achieve dynamic risk management of multi-level collaboration in the cloud.

7. The autonomous obstacle avoidance system for flying cars based on vehicle-road-cloud cooperation according to claim 1, characterized in that: The core formula of the risk quantification model is: In the formula, β is a spatial distance sensitivity factor with a value of 0.1; The distance between the i-th pair of aircraft; TTC i Let i be the collision time of the i-th pair; The hierarchical response mechanism specifically includes: When the risk value exceeds the set threshold, global path replanning is triggered. When the risk value is in the range of 0-0.3, monitoring measures are implemented: when the risk value is in the range of 0.3-0.7, path fine-tuning suggestions are implemented with an execution time of ≤200ms; when the risk value is >0.7, replanning is forcibly executed with an execution time of ≤50ms; when the risk value is >0.9, control is taken over and airspace is urgently isolated with an execution time of ≤20ms.

Images