All-weather high-precision automatic parking and guiding system

Abstract

The invention is suitable for the technical field of unmanned aerial vehicles, and provides an all-weather high-precision automatic parking and guiding system, and the system comprises an environment perception and signal enhancement module which is used for carrying out the communication with a ground base station array through an airborne ultra-wideband tag so as to obtain distance measurement data, and collecting an environment image through an airborne visual sensor; a self-adaptive multi-source fusion positioning module which is connected with the environment perception and signal enhancement module and is used for carrying out tight coupling fusion on an original distance measurement value and a pose increment calculated by a visual odometer based on a self-adaptive extended Kalman filtering architecture so as to generate high-precision state estimation of the unmanned aerial vehicle; and the path planning and guiding control module is used for generating a landing track by adopting a model prediction control algorithm according to the high-precision state estimation, activating a precise visual servo mode in a tail end parking stage, and driving the unmanned aerial vehicle to complete parking. On the premise of guaranteeing the robustness of the system, high-precision and zero-intervention automatic parking of the unmanned aerial vehicle under complex working conditions such as wind, rain, snow, fog and the like is realized.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, specifically to an all-weather, high-precision automatic parking and guidance system. Background Technology

[0002] Existing autonomous homing and parking technologies for unmanned aerial vehicles (UAVs) primarily rely on satellite navigation and positioning systems (GNSS / RTK) or end-point guidance schemes based on visual landmarks. Among these, GNSS / RTK differential positioning is currently the fundamental method for achieving UAV waypoint flight. This scheme calculates absolute coordinates using satellite signals and has good usability in open environments. However, in complex scenarios such as around automated hangars, between urban buildings, or with large metal structures, satellite signals are highly susceptible to severe interference from obstruction or multipath effects, leading to positioning data drift or even signal interruption. Especially during the vertical descent phase, where high accuracy is required, the volatility of satellite positioning often makes it difficult to ensure that the UAV lands accurately within the charging port or landing pad.

[0003] On the other hand, vision-based assisted landing technology uses onboard cameras to identify specific optical beacons, QR codes (such as AprilTags), or infrared dots on the ground, and calculates relative position and attitude using image geometry. While this method can provide high relative accuracy at close range, its stability and reliability are highly dependent on environmental visibility and weather conditions. In all-weather scenarios such as rain, snow, dense fog, or drastic changes in lighting, optical lenses are prone to blurring, water vapor obscuring, or overexposure, leading to feature extraction failure. In addition, the field of view (FOV) of visual sensors is usually narrow. When the UAV is affected by strong crosswinds and tilts significantly, the target can easily leave the field of view, resulting in guidance mission failure.

[0004] In summary, most current mainstream technologies are based on single sensor sources or simple, loose combinations, lacking deep multimodal fusion mechanisms. Existing technologies struggle to simultaneously maintain positioning accuracy, continuity, and robustness under the dual challenges of radio frequency signal interference and visual image degradation, failing to meet the practical application requirements of high-precision, safe parking for UAVs in complex, all-weather environments. Therefore, there is a need to provide an all-weather, high-precision automatic parking and guidance system to address these issues. Summary of the Invention

[0005] In view of the shortcomings of existing technologies, the purpose of this invention is to provide an all-weather, high-precision automatic parking and guidance system to solve the problem of high-reliability guidance for UAVs in all-weather environments.

[0006] This invention is implemented as follows: an all-weather, high-precision automatic parking and guidance system, the system comprising: The environmental perception and signal enhancement module is used to communicate with the ground base station array through an airborne ultra-wideband tag to obtain ranging data, to collect environmental images through an airborne visual sensor, and to perform anti-interference preprocessing on the ranging data and environmental images. An adaptive multi-source fusion positioning module, connected to the environmental perception and signal enhancement module, is used to receive preprocessed ranging data and environmental images. Based on an adaptive extended Kalman filter architecture, it tightly couples and fuses the original ranging value with the pose increment calculated by the visual odometry to generate a high-precision state estimate of the UAV. The path planning and guidance control module, connected to the adaptive multi-source fusion positioning module, is used to generate a landing trajectory based on high-precision state estimation and model predictive control algorithm, and to activate the precision visual servo mode during the terminal parking phase to drive the UAV to complete precise parking.

[0007] As a further embodiment of the present invention, the environmental perception and signal enhancement module includes: The image restoration unit is used to enhance images degraded by rain and fog by employing a lightweight image restoration algorithm based on deep residual networks, combined with dark channel priors and multi-scale Retinex algorithms, and output clear images. The UWB signal analysis unit is used to analyze the received signal strength and head-to-diameter power ratio of the UWB signal in order to identify and mark non-line-of-sight obstructions in real time.

[0008] As a further aspect of the present invention, the steps performed in the adaptive multi-source fusion localization module include: defining the system state vector X. t This includes the position, velocity, attitude, and zero bias of the UAV's inertial measurement unit; and the construction of the dynamic observation noise covariance matrix R. t Based on the image quality assessment results and non-line-of-sight discrimination results output by the environmental perception and signal enhancement module, the confidence weights of the visual sensor and the UWB sensor are dynamically adjusted; based on the dynamic observation noise covariance matrix R... t Calculate the gain K of the extended Kalman filter. t This is to integrate the raw UWB ranging values ​​with the pose increments from the visual odometry.

[0009] As a further aspect of the present invention, the image restoration unit is specifically used to analyze the feature blurring degree and contrast reduction of the input image, and apply an enhancement algorithm combining dark channel prior and multi-scale Retinex to output a clear image after denoising.

[0010] As a further aspect of the present invention, the UWB signal analysis unit, based on UWB channel impulse response characteristic analysis, determines in real time whether non-line-of-sight occlusion exists, and the determination result is transmitted through an indicator function. Represented and used in ranging models The corresponding ranging error compensation term is introduced in the process. This is the current location of the drone. For the first The coordinates of each base station This is an inherent bias in the system. This is the positive ranging error caused by non-line-of-sight.

[0011] As a further aspect of the present invention, the dynamic observation noise covariance matrix R t The adjustment is achieved by adjusting the visual observation noise covariance R. vis and UWB observation noise covariance R uwb Implemented by assigning dynamic weighting factors α and β, Kalman gain K t The calculation formula is: .

[0012] As a further aspect of the present invention, the path planning and guidance control module employs a model predictive control algorithm to generate control commands u by solving the following optimization problem. t : , where X t For the current state estimate, X ref Using the reference trajectory, Q and R are weight matrices, and the optimization process is constrained by the maximum tilt angle of the UAV and the saturation constraint of the motor thrust.

[0013] As a further aspect of the present invention, the precision visual servoing mode is used when the distance D between the UAV and the parking point is less than a threshold D. th When activated, the system identifies the cooperative markers on the docking platform, uses the perspective n-point algorithm to calculate the relative pose error of the drone, and directly drives the drone to make fine adjustments to its position and attitude until it completes the net-touch or lock-on action.

[0014] As a further aspect of the present invention, the confidence weight of the visual sensor is dynamically evaluated based on changes in image entropy, and the confidence weight of the UWB sensor is based on a non-line-of-sight discrimination indicator function. Dynamically evaluate the corresponding error characteristics.

[0015] As a further aspect of the present invention, the model predictive control algorithm of the path planning and guidance control module integrates a disturbance observer to estimate and compensate for the impact of wind field disturbances on the attitude of the UAV, thereby improving the robustness of trajectory tracking.

[0016] Compared with the prior art, the beneficial effects of the present invention are: A tightly coupled fusion architecture combining UWB radio frequency and visual features is adopted. The strong penetration and high refresh rate of UWB signals effectively fill the perception gap in adverse weather conditions, while the high angular resolution of vision corrects the ranging bias of UWB, achieving complementary enhancement. Furthermore, an adaptive weighting mechanism based on environmental confidence is introduced, breaking the limitations of loose sensor combinations or simple switching in traditional solutions. The fusion weights can be dynamically adjusted in real time according to the environmental signal-to-noise ratio (such as rain / fog concentration and electromagnetic interference intensity), achieving a deep closed loop from the signal processing layer to the control decision layer. This ensures that the UAV maintains centimeter-level guidance accuracy and smooth control trajectory even when a single signal source is damaged, enabling high-precision, zero-intervention automatic parking of the UAV in complex conditions such as wind, rain, snow, and fog. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of a high-precision, all-weather automatic parking and guidance system. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0020] like Figure 1 As shown, this embodiment of the invention provides an all-weather, high-precision automatic parking and guidance system, the system comprising: The environmental perception and signal enhancement module 100 is used to communicate with the ground base station array via an airborne ultra-wideband (UWB) tag to obtain ranging data, acquire environmental images through an airborne visual sensor, and perform anti-interference preprocessing on the ranging data and environmental images. Located at the front end of the system, it undertakes the tasks of raw data acquisition and preprocessing. The system communicates with the ground base station array at high frequency (50Hz+) via the airborne UWB tag, while simultaneously acquiring environmental images using a downward / forward-looking binocular camera. Considering the physical damage to the sensors caused by severe weather, this module introduces a lightweight image restoration algorithm based on a deep residual network (ResNet) and a UWB channel impulse response (CIR) analysis mechanism.

[0021] The adaptive multi-source fusion positioning module 200 is connected to the environment perception and signal enhancement module 100. It is used to receive preprocessed ranging data and environmental images. Based on the adaptive extended Kalman filter (EKF) architecture, it tightly couples and fuses the raw UWB ranging value with the pose increment calculated by visual odometry (VIO) to generate a high-precision state estimate of the UAV.

[0022] The path planning and guidance control module 300 is connected to the adaptive multi-source fusion positioning module 200. It is used to generate a landing trajectory based on high-precision state estimation and model predictive control (MPC) algorithm, and to activate the precision visual servo mode in the terminal parking stage to drive the UAV to complete precise parking.

[0023] It should be noted that the embodiments of this invention include ultra-wideband (UWB) wireless positioning, computer vision perception, multi-sensor fusion, and anti-interference signal processing in complex environments. These technologies are key to achieving accurate "last-mile" recovery of UAVs in GNSS-denied or weak-signal environments, and are also core technical means to solve the bottleneck of all-weather automated operation of UAVs. Multi-sensor fusion positioning technology refers to the spatiotemporal registration and complementary fusion of sensor data with different physical characteristics to obtain more accurate and robust position and attitude information than a single sensor. In the scenario of automatic UAV parking, single visual navigation is easily affected by changes in lighting and lack of texture, while single UWB positioning, although highly penetrating, is susceptible to multipath interference and lacks attitude details. By fusing the high-precision ranging capability of UWB with the rich texture and angle information of visual sensors, a three-dimensional accurate positioning model is constructed, aiming to overcome the perception limitations of a single sensor in the near-ground parking stage and improve the accuracy and stability of the UAV guidance process. Anti-interference signal processing and adaptive correction technology mainly solve the problem of communication and perception reliability under adverse weather conditions. In extreme weather conditions such as rain, snow, and fog, radio signals attenuate and become distorted, while optical lenses face challenges of blurring and obstruction. This invention addresses these challenges by employing an adaptive signal enhancement algorithm and an environmental perception feedback mechanism to correct non-line-of-sight (NLOS) errors in UWB signals in real time and enhance the feature extraction capabilities of visual images. This ensures that the UAV maintains a highly reliable guidance link even under complex weather conditions. Through the synergistic effect of ultra-wideband and visual 3D fusion and adaptive signal processing technology in all-weather environments, a high-precision, robust system-level solution is provided for UAV automated hangars, mobile platform recovery, and automated parking in complex field environments. This demonstrates the significant potential for fully autonomous UAV recovery under all-weather conditions and is applicable to fields such as power line inspection, logistics transportation, and emergency rescue. The system's design emphasizes a closed-loop processing of "environmental perception - signal correction - tightly coupled fusion." By complementing multi-sensor information and dynamically adjusting weights, the system significantly improves guidance robustness and parking accuracy in non-line-of-sight (NLOS) and visual degradation scenarios.

[0024] In this embodiment of the invention, the environmental perception and signal enhancement module 100 includes: The image restoration unit employs a lightweight image restoration algorithm based on a deep residual network (ResNet) to address the contrast loss and feature blurring caused by rain and fog. It combines dark channel prior (DCP) and multi-scale Retinex algorithms to enhance images degraded by rain and fog, outputting a clear image. ; The UWB signal analysis unit is used to analyze the received signal strength (RSSI) and front-to-diameter power ratio (FPP) of UWB signals in order to identify and mark non-line-of-sight (NLOS) obstructions in real time.

[0025] In this embodiment of the invention, the steps performed by the adaptive multi-source fusion localization module 200 include: defining the system state vector X. t This includes the position, velocity, attitude, and inertial measurement unit (IMU) zero bias of the UAV; and the construction of the dynamic observation noise covariance matrix R. t Based on the image quality assessment results and NLOS discrimination results output by the environmental perception and signal enhancement module 100, the confidence weights of the visual sensor and the UWB sensor are dynamically adjusted; based on the dynamic observation noise covariance matrix R... t Calculate the gain K of the extended Kalman filter. t This is to integrate the raw UWB ranging values ​​with the pose increments from visual odometry (VIO).

[0026] In this embodiment of the invention, the image restoration unit is specifically used to analyze the feature blurring degree and contrast reduction of the input image, apply an enhancement algorithm combining dark channel prior and multi-scale Retinex, and output a clear image after denoising.

[0027] In this embodiment of the invention, the UWB signal analysis unit, based on UWB channel impulse response (CIR) characteristic analysis, determines in real time whether non-line-of-sight (NLOS) obstruction exists, and the determination result is expressed through an indicator function. Represented and used in ranging models A corresponding ranging error compensation term is introduced. Among them, Indicates in Time, drone tags and number The observed distance (measured value) between each ground base station. This is the current location of the drone. For the first The coordinates of each base station This is an inherent bias in the system. This represents random noise that follows a Gaussian (normal) distribution. This represents a noise mean of zero. The ranging noise variance of a UWB sensor reflects the measurement accuracy or dispersion of the sensor under line-of-sight (LOS) conditions. Non-line-of-sight Distinguish indicator function (0 or 1). To account for the positive ranging error caused by non-line-of-sight, this formula explicitly models environmental interference, providing a physically consistent observation model for subsequent fusion filtering.

[0028] In this embodiment of the invention, an environment-adaptive tightly coupled extended Kalman filter (AdaptiveEKF) architecture is constructed. Traditional loosely coupled filters only fuse position calculation results, while this embodiment directly inputs the raw UWB ranging values ​​and the pose increment calculated by visual odometry (VIO) into the filter. The system's state vector is defined as... This includes position, velocity, and attitude quaternions, as well as IMU bias. To address sensor performance fluctuations in all-weather environments, a dynamic observation noise covariance matrix is ​​introduced. When the visual algorithm detects a decrease in image entropy due to dense fog, or when UWB detects NLOS features, the system automatically adjusts the trust weights of the corresponding sensors. The dynamic observation noise covariance matrix R... t The adjustment is achieved by adjusting the visual observation noise covariance R. vis and UWB observation noise covariance R uwb Implemented by assigning dynamic weighting factors α and β, Kalman gain K t The calculation formula is: In the formula This refers to Kalman Gain. For visual observation noise covariance, For UWB observation noise covariance, , Represents dynamic weighting factors. Let represent the one-step prediction error covariance matrix, which describes the uncertainty of the current state estimate derived from the previous state before updating using the current observations. This represents the transpose of the observation matrix, which maps the system's state space (such as position and velocity) to the observation space (such as distance and pixel coordinates), and is used to describe the linearized relationship between state variables and observations.

[0029] In this embodiment of the invention, the path planning and guidance control module 300 uses a model prediction (MPC) control algorithm to generate control commands u by solving the following optimization problem. t : The above formula is based on a disturbance observer incorporated into the control design, taking into account the impact of wind field disturbances on the UAV's attitude. Estimate the current state. For the predicted state estimation, For reference trajectory, u tThis indicates the generation of control commands. and This is the weight matrix. and This represents a weighted quadratic form, with the specific mathematical meaning as follows: Its physical meaning is: to calculate the sum of squares of the errors and, based on the weight matrix... or Weighted penalties are applied to different state variables (such as position deviation vs. velocity deviation) or control variables (such as thrust magnitude). The optimization process is constrained by the maximum tilt angle of the UAV and the saturation constraint of the motor thrust.

[0030] In this embodiment of the invention, the precision visual servoing mode is used when the distance D between the UAV and the parking point is less than a threshold D. th When activated, the system identifies the cooperative markers (ArUco / LED array) on the parking platform, uses the perspective n-point (PnP) algorithm to calculate the relative pose error of the drone, and directly drives the drone to make fine adjustments to its position and attitude until it completes the net-touch or lock-on action.

[0031] In this embodiment of the invention, the confidence weight of the visual sensor is dynamically evaluated based on changes in image entropy, and the confidence weight of the UWB sensor is based on a non-line-of-sight discrimination indicator function. The system dynamically evaluates the corresponding error characteristics. The Model Predictive Control (MPC) algorithm of the path planning and guidance control module 300 integrates a disturbance observer to estimate and compensate for the impact of wind field disturbances on the UAV's attitude, thereby improving the robustness of trajectory tracking. This system drives the mid-level adaptive fusion algorithm through physical-level enhancement of the underlying signals and transmits high-confidence state information upwards to support MPC control, forming a highly closed-loop anti-disturbance feedback chain among the modules. This solution effectively solves the technical problem of "unclear visibility and inaccurate positioning" in traditional single guidance technologies under extreme weather and complex electromagnetic environments, providing key technical support for all-weather UAV automated hangars and mobile recovery platforms.

[0032] This invention uses a low-level time synchronization mechanism to perform pixel-level and signal-level spatiotemporal alignment of UWB high-frequency ranging data and visual sensor texture feature data, thereby constructing an automatic guidance hardware foundation with all-weather perception capabilities.

[0033] The adaptive multimodal fusion localization method in this invention is based on the real-time assessment of environmental interference using UWB channel impulse response (CIR) characteristics and image entropy values, and dynamically generates sensor confidence weights. Within the extended Kalman filter (EKF) framework, the observation noise covariance matrix is ​​automatically adjusted according to the confidence level to achieve a seamless and smooth switch from "visual-dominated" to "RF-dominated", thus solving the problem of single source failure under adverse weather conditions or metal obstruction.

[0034] This invention provides an image restoration algorithm combining dark channel priors and deep residual networks for rainy and foggy weather, and an NLOS bias culling algorithm based on signal feature analysis for complex electromagnetic environments. Through dual correction at the raw data level and the use of disturbance-resistant trajectories generated by model predictive control (MPC), centimeter-level precise automatic parking of UAVs is achieved under all-weather conditions.

[0035] The above description only details the preferred embodiments of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0036] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the disclosure in the specification and embodiments. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

Claims

1. An all-weather, high-precision automatic parking and guidance system, characterized in that: The system includes: The environmental perception and signal enhancement module is used to communicate with the ground base station array through an airborne ultra-wideband tag to obtain ranging data, to collect environmental images through an airborne visual sensor, and to perform anti-interference preprocessing on the ranging data and environmental images. An adaptive multi-source fusion positioning module, connected to the environmental perception and signal enhancement module, is used to receive preprocessed ranging data and environmental images. Based on an adaptive extended Kalman filter architecture, it tightly couples and fuses the original ranging value with the pose increment calculated by the visual odometry to generate a high-precision state estimate of the UAV. The path planning and guidance control module, connected to the adaptive multi-source fusion positioning module, is used to generate a landing trajectory based on high-precision state estimation and model predictive control algorithm, and to activate the precision visual servo mode during the terminal parking phase to drive the UAV to complete precise parking.

2. The all-weather high-precision automatic parking and guidance system according to claim 1, characterized in that, The environmental perception and signal enhancement module includes: The image restoration unit is used to enhance images degraded by rain and fog by employing a lightweight image restoration algorithm based on deep residual networks, combined with dark channel priors and multi-scale Retinex algorithms, and output clear images. The UWB signal analysis unit is used to analyze the received signal strength and head-to-diameter power ratio of the UWB signal in order to identify and mark non-line-of-sight obstructions in real time.

3. The all-weather high-precision automatic parking and guidance system according to claim 2, characterized in that, The steps performed in the adaptive multi-source fusion localization module include: defining the system state vector X. t This includes the position, velocity, attitude, and zero bias of the UAV's inertial measurement unit; and the construction of the dynamic observation noise covariance matrix R. t Based on the image quality assessment results and non-line-of-sight discrimination results output by the environmental perception and signal enhancement module, the confidence weights of the visual sensor and the UWB sensor are dynamically adjusted; based on the dynamic observation noise covariance matrix R... t Calculate the gain K of the extended Kalman filter. t This is to integrate the raw UWB ranging values ​​with the pose increments from the visual odometry.

4. The all-weather high-precision automatic parking and guidance system according to claim 2, characterized in that, The image restoration unit is specifically used to analyze the feature blurring degree and contrast reduction of the input image, and apply an enhancement algorithm combining dark channel prior and multi-scale Retinex to output a clear image after denoising.

5. The all-weather high-precision automatic parking and guidance system according to claim 2, characterized in that, The UWB signal analysis unit analyzes the UWB channel impulse response characteristics to determine in real time whether non-line-of-sight obstruction exists. The determination result is expressed through an indicator function. Represented and used in ranging models The corresponding ranging error compensation term is introduced in the process. Indicates in Time, drone tags and number The observed distance between ground base stations This is the current location of the drone. For the first The coordinates of each base station This is an inherent bias in the system. This represents random noise that follows a Gaussian distribution. This represents a noise mean of zero. The variance of the ranging noise of a UWB sensor reflects the measurement accuracy or dispersion of the sensor under line-of-sight conditions. Non-line-of-sight Discrimination indicator function, This is the positive ranging error caused by non-line-of-sight.

6. The all-weather high-precision automatic parking and guidance system according to claim 3, characterized in that, The dynamic observation noise covariance matrix R t The adjustment is achieved by adjusting the visual observation noise covariance R. vis and UWB observation noise covariance R uwb Implemented by assigning dynamic weighting factors α and β, Kalman gain K t The calculation formula is: In the formula For Kalman gain, For visual observation noise covariance, For UWB observation noise covariance, , Represents dynamic weighting factors. Let represent the one-step prediction error covariance matrix, which describes the uncertainty of the current state estimate derived from the previous state before updating using the current observations. This represents the transpose of the observation matrix, which maps the system's state space to the observation space and is used to describe the linearized relationship between state variables and observations.

7. The all-weather high-precision automatic parking and guidance system according to claim 1, characterized in that, The path planning and guidance control module employs a model predictive control algorithm that generates control commands u by solving the following optimization problem. t : ,in, Estimate the current state. For the predicted state estimation, For reference trajectory, u t This indicates the generation of control commands. and This is the weight matrix. and This represents a weighted quadratic form.

8. The all-weather high-precision automatic parking and guidance system according to claim 1, characterized in that, The precision visual servoing mode is used when the distance D between the drone and the parking point is less than a threshold D. th When activated, the system identifies the cooperative markers on the docking platform, uses the perspective n-point algorithm to calculate the relative pose error of the drone, and directly drives the drone to make fine adjustments to its position and attitude until it completes the net-touch or lock-on action.

9. The all-weather high-precision automatic parking and guidance system according to claim 3, characterized in that, The confidence weight of the visual sensor is dynamically evaluated based on changes in image entropy, while the confidence weight of the UWB sensor is based on a non-line-of-sight discrimination indicator function. Dynamically evaluate the corresponding error characteristics.

10. The all-weather high-precision automatic parking and guidance system according to claim 1, characterized in that, The model predictive control algorithm of the path planning and guidance control module integrates a disturbance observer to estimate and compensate for the impact of wind field disturbances on the attitude of the UAV, thereby improving the robustness of trajectory tracking.

Images