RID-based air-ground integrated unmanned aerial vehicle detection and obstacle avoidance system

Abstract

The application discloses a RID-based air-ground integrated unmanned aerial vehicle detection and obstacle avoidance system, which comprises an RID signal processing module, an electromagnetic evolution field construction module, a sensor point cloud processing module, a three-dimensional environment perception model construction module, an obstacle identification and dynamic updating module, a flight control and path planning module and a flight execution module.The RID signal processing module is used for collecting RID remote identification signal data; the electromagnetic evolution field construction module is used for constructing an electromagnetic evolution field data set; the sensor point cloud processing module is used for generating a sensor point cloud set; the three-dimensional environment perception model construction module is used for performing double-domain fusion of the electromagnetic evolution field data set and the sensor point cloud set; the obstacle identification and dynamic updating module is used for performing obstacle identification and dynamic updating; the flight control and path planning module is used for performing path planning and dynamic correction to generate an obstacle avoidance flight control instruction; and the flight execution module is used for adjusting the attitude and track of the unmanned aerial vehicle according to the obstacle avoidance flight control instruction.The application has the advantages of strong perception robustness and high flight safety by using RID signal fusion point cloud modeling, three-dimensional environment perception construction and dynamic obstacle avoidance.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) detection and obstacle avoidance technology, and in particular to an air-ground integrated UAV detection and obstacle avoidance system based on RID (Radar Identifier). Background Technology

[0002] With the widespread application of drones in urban low-altitude logistics, environmental monitoring, and emergency rescue, higher requirements are placed on stable detection and autonomous obstacle avoidance in complex low-altitude environments. Existing drones mostly rely on airborne lidar, millimeter-wave radar, and visual sensors for environmental perception and obstacle detection. Although they have a certain detection accuracy in conventional environments, they often suffer from incomplete perception and path planning failures in scenarios with insufficient lighting, severe obstruction, or strong electromagnetic interference, resulting in insufficient flight safety.

[0003] Current technologies primarily use Remote Identification (RID) as a tool for drone identity broadcasting and location sharing, limiting its application to compliance and regulatory purposes. They do not utilize its physical propagation characteristics for environmental perception, resulting in RID failing to provide effective obstacle avoidance support when the performance of sensors such as vision and radar deteriorates. Therefore, how to overcome the limitations of traditional sensors in complex low-altitude environments, utilize the propagation characteristics of RID signals to construct an electromagnetic evolution field, and fuse it with sensor point cloud data to generate a dynamically updated obstacle avoidance environment model has become a pressing issue that needs to be addressed in current technologies. Summary of the Invention

[0004] One objective of this invention is to propose an air-ground integrated UAV detection and obstacle avoidance system based on RID. This invention utilizes RID signal fusion point cloud modeling to construct three-dimensional environmental perception and dynamic obstacle avoidance, possessing the advantages of strong perception robustness and high flight safety.

[0005] According to an embodiment of the present invention, an air-to-ground integrated UAV detection and obstacle avoidance system based on RID includes:

[0006] The RID signal processing module is used to collect RID remote identification signal data of the UAV during flight, and to perform unified format processing on the signal data to generate a standardized RID signal data sequence.

[0007] The electromagnetic evolution field construction module is used to calculate propagation difference characteristics based on standardized RID signal data sequences and construct an electromagnetic evolution field dataset.

[0008] The sensor point cloud processing module is used to collect and process detection data from UAV-borne lidar, millimeter-wave radar, and visual sensors to generate sensor point cloud sets.

[0009] The 3D environment perception model construction module is used to fuse the electromagnetic evolution field data set and the sensor point cloud set in a dual domain to generate a signal-space coupled 3D environment perception model.

[0010] The obstacle recognition and dynamic update module is used to perform obstacle recognition and dynamic update on the signal-space coupled 3D environment perception model to generate an obstacle avoidance environment model.

[0011] The flight control and path planning module is used to input the obstacle avoidance environment model into the UAV flight control unit, perform path planning and dynamic correction, and generate obstacle avoidance flight control commands.

[0012] The flight execution module is used to adjust the UAV's attitude and trajectory according to obstacle avoidance flight control commands, so as to achieve stable and reliable obstacle detection and autonomous obstacle avoidance in complex low-altitude environments.

[0013] Optionally, modules can be integrated using the following methods:

[0014] Collect RID remote identification signal data of UAVs during flight, perform unified format processing, and generate standardized RID signal data sequences;

[0015] Based on standardized RID signal data sequences, signal propagation differences are calculated, and signal intensity attenuation features, phase shift features, and temporal difference features are extracted to construct an electromagnetic evolution field dataset that reflects the state of the surrounding space environment of the UAV.

[0016] Collect detection data from UAV-borne lidar, millimeter-wave radar, and visual sensors; perform noise filtering, coordinate registration, and multi-source time synchronization on the detection data; and generate a sensor point cloud set.

[0017] The electromagnetic evolution field data set and the sensor point cloud set are fused in two domains to generate a signal-space coupled three-dimensional environmental perception model.

[0018] Obstacle identification and dynamic updating are performed on the signal-space coupled 3D environment perception model to generate an obstacle avoidance environment model that includes the location, size and motion state of obstacles.

[0019] The obstacle avoidance environment model is input into the UAV flight control unit to perform path planning and dynamic correction, and generate obstacle avoidance flight control commands.

[0020] The drone's attitude and trajectory are adjusted according to obstacle avoidance flight control commands to achieve stable and reliable obstacle detection and autonomous obstacle avoidance in complex low-altitude environments.

[0021] Optionally, the generation of the standardized RID signal data sequence includes the following specific steps:

[0022] During the flight of the UAV, RID remote identification signal data is collected, which includes signal strength, phase information and timing information;

[0023] The acquired RID remote identification signal data is subjected to noise filtering, interference suppression and time-frequency synchronization processing to generate a purified RID remote identification signal data set.

[0024] The signal strength of the purified RID remote identification signal data set is normalized to generate a signal strength sequence.

[0025] Phase difference processing is performed on the phase information of the purified RID remote identification signal data set to generate a phase difference sequence;

[0026] The time-series information of the purified RID remote identification signal data set is subjected to time-series differential processing to generate a time-series differential sequence;

[0027] The signal strength sequence, phase difference sequence, and timing difference sequence are processed to unify their formats and then spliced ​​together to generate a standardized RID signal data sequence.

[0028] The standardized RID signal data sequence is written into the UAV's local storage unit.

[0029] Optionally, the construction of the electromagnetic evolution field dataset includes the following specific steps;

[0030] Signal strength attenuation characteristics, phase shift characteristics, and timing difference characteristics are calculated based on standardized RID signal data sequences. The signal strength attenuation characteristics are obtained by performing a logarithmic operation on the normalized signal strength. The phase shift characteristics are obtained by performing an accumulation operation on the phase difference information of historical sampling points. The timing difference characteristics are obtained by performing an averaging operation on the timing difference information within a preset time window.

[0031] The signal strength attenuation characteristics, phase shift characteristics, and timing difference characteristics are combined under the same time index to form a propagation difference feature vector;

[0032] Based on the propagation difference feature vector, an improved Gaussian core field mapping method is used to generate a preliminary electromagnetic evolution field dataset, specifically including:

[0033] The kernel width parameter is determined by jointly considering the signal strength attenuation characteristics and timing difference characteristics in the propagation difference feature vector;

[0034] The weighting factor is determined by combining the signal strength attenuation characteristics and the phase shift characteristics.

[0035] The spatial position vector of the point to be calculated, the spatial position vector of each sampling point, the kernel width parameter, and the weighting factor are input into the Gaussian kernel function to calculate the electromagnetic evolution field value of each spatial position under the global time index. The preliminary electromagnetic evolution field dataset is generated by performing a weighted summation on the calculation results of all sampling points.

[0036] Within the continuous time segments of the propagation difference feature vector, a temporal coherence threshold is set, and the preliminary electromagnetic evolution field values ​​corresponding to the time segments whose temporal coherence index exceeds the temporal coherence threshold are reweighted to obtain the final electromagnetic evolution field dataset.

[0037] Optionally, the generation of the sensor point cloud set includes the following specific steps:

[0038] During the flight of the UAV, point cloud data from airborne lidar, point cloud data from millimeter-wave radar, and visual depth data are collected respectively to form lidar point cloud sequences, millimeter-wave radar point cloud sequences, and visual depth sequences.

[0039] Noise is filtered out from the lidar point cloud sequence, millimeter-wave radar point cloud sequence, and visual depth sequence to obtain filtered lidar point cloud data, millimeter-wave radar point cloud data, and visual depth data.

[0040] The noise-filtered data are transformed into the UAV's body coordinate system to achieve coordinate unification, so that the three types of data can be represented under the same spatial reference.

[0041] Coordinate registration of the three types of data is performed in the UAV's body coordinate system;

[0042] Based on the completion of coordinate registration, time synchronization is performed to uniformly process the timestamps of lidar point cloud data, millimeter-wave radar point cloud data, and visual depth data.

[0043] The processed lidar point cloud data, millimeter-wave radar point cloud data, and visual depth data are fused to generate a sensor point cloud set in a unified format.

[0044] Optionally, the generation of the three-dimensional environment perception model includes the following specific steps:

[0045] Acquire electromagnetic evolution field data sets and sensor point cloud sets, and use them as inputs for dual-domain fusion;

[0046] Establish a correspondence between the electromagnetic evolution field dataset and the sensor point cloud dataset, calculate the Euclidean distance between the spatial position vector in the electromagnetic evolution field dataset and the spatial position vector in the sensor point cloud dataset, and use it as the spatial proximity.

[0047] The spatial proximity is compared with a preset spatial proximity threshold. All point pairs that are less than or equal to the spatial proximity threshold are retained as candidate matching point pairs, and a fusion weight is calculated for the candidate matching point pairs.

[0048] ;

[0049] in, The fusion weights are the fusion weights of the i-th sampling point and the j-th sensor point cloud. Let be the electromagnetic evolution field value of the i-th sampling point at the global time index t. For the first Electromagnetic evolution field values ​​at each sampling point under the global time index t. This is the sum of the electromagnetic evolution field values ​​at all n sampling points under the global time index t. Let i be the spatial location vector of the i-th sampling point. Let j be the spatial location vector of the j-th point in the sensor point cloud set. For spatial scale parameters, This represents the total number of sampling points in the electromagnetic evolution field dataset;

[0050] The electromagnetic evolution field values ​​are mapped to points in the sensor point cloud set using fusion weights, where the enhancement value corresponding to the j-th point in the sensor point cloud set is:

[0051] ;

[0052] in, The enhanced value of the j-th point in the sensor point cloud after fusing electromagnetic evolution field features;

[0053] The merged point cloud set is stored as a unified three-dimensional data structure;

[0054] A signal-space coupled 3D environment perception model is constructed based on a 3D data structure.

[0055] Optionally, the generation of the obstacle avoidance environment model includes the following specific steps:

[0056] Acquire a signal-space coupled 3D environment perception model, which includes the spatial location and augmentation value of each point in the sensor point cloud set;

[0057] Spatial segmentation is performed on the signal-space coupled 3D environment perception model, and the sensor point cloud set is divided into multiple candidate regions using a density-based clustering method;

[0058] Calculate the point density and electromagnetic consistency index in each candidate region;

[0059] The candidate region is determined as an obstacle region based on a combination of point density and electromagnetic consistency index. When both exceed a preset threshold, the candidate region is identified as an obstacle region.

[0060] The point set within the obstacle region is fitted with a three-dimensional boundary, and the bounding box of the obstacle in three-dimensional space is calculated to form the spatial geometric parameters of the obstacle.

[0061] Track the change in the center position of the obstacle region under continuous global time indexing, and calculate the motion velocity vector of the obstacle;

[0062] By combining the spatial geometric parameters and motion velocity vectors of obstacles, an obstacle avoidance environment model is generated, which includes the position, size, and motion state of the obstacles.

[0063] Optionally, the generation of the obstacle avoidance flight control command includes the following specific steps:

[0064] Obtain the obstacle avoidance environment model and use it as input to the UAV flight control unit;

[0065] In the UAV flight control unit, the current position vector of the UAV and the target position vector are defined, and the flight path position vector is also defined;

[0066] Construct a path cost function to simultaneously consider flight distance and obstacle avoidance safety:

[0067] ;

[0068] in, For path cost function, Let be the flight path position vector at global time index t. Let the target position vector be... Let m be the position of the m-th obstacle at global time index t. The total number of obstacles. Weighted by flight distance, For obstacle avoidance safety weights, This is the termination time index for path planning. This is the start time index for path planning;

[0069] The initial planned path is obtained by minimizing the path cost function.

[0070] During flight, the obstacle avoidance environment model is updated in real time, and the updated obstacle positions are obtained. The updated obstacle positions are then substituted into the path cost function to obtain the corrected cost function.

[0071] The local path is recalculated within a local range based on the modified cost function to obtain a dynamically modified path.

[0072] The dynamically corrected path is merged with the initial planned path to generate the final obstacle avoidance path, which is then discretized into a sequence of flight control commands.

[0073] The beneficial effects of this invention are:

[0074] This invention overcomes the limitations of existing technologies that only use RID as an identification and location sharing tool by performing in-depth processing on the RID remote identification signals generated during UAV flight. It transforms signal strength, phase information, and timing information into propagation difference features that can characterize environmental states and constructs an electromagnetic evolution field dataset. Compared with traditional methods that rely solely on visual sensors or radar, this invention can maintain environmental perception capabilities even under extreme conditions such as insufficient light, severe obstruction, or electromagnetic interference, thereby improving the survivability and flight reliability of UAVs in complex low-altitude environments.

[0075] In terms of multi-source information fusion, this invention performs dual-domain fusion of electromagnetic evolution field data sets with point cloud data from airborne lidar, millimeter-wave radar, and visual sensors to generate a signal-space coupled three-dimensional environmental perception model. This model enables the environmental representation to include both geometric features and electromagnetic properties. This fusion model not only improves the robustness of obstacle recognition but also dynamically reflects the spatial position, size, and motion state of obstacles, thus providing more comprehensive input conditions for path planning. Based on this, this invention introduces a dynamic obstacle update mechanism and a correction cost function to correct the planned path in real time, ensuring that the UAV can quickly make an evasive response when encountering dynamic obstacles.

[0076] Furthermore, the path planning and dynamic correction strategy proposed in this invention not only ensures the global rationality of the initial planned path, but also generates the final obstacle avoidance path by merging it point by point with the dynamically corrected path. This enables the UAV to maintain flight efficiency globally while achieving timely avoidance of sudden obstacles in local areas. The generated obstacle avoidance flight control command is then input into the UAV flight control unit to ensure the stable execution of attitude adjustment and trajectory correction. Attached Figure Description

[0077] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0078] Figure 1 This is a flowchart of a method for an air-to-ground integrated UAV detection and obstacle avoidance system based on RID proposed in this invention.

[0079] Figure 2This is a schematic diagram of the dual-domain fusion of the three-dimensional environmental perception model construction unit of the RID-based air-ground integrated UAV detection and obstacle avoidance system proposed in this invention. Detailed Implementation

[0080] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0081] refer to Figure 1-2 A ground-based air-to-ground integrated UAV detection and obstacle avoidance system based on RID, comprising:

[0082] The RID signal processing module is used to collect RID remote identification signal data of the UAV during flight, and to perform unified format processing on the signal data to generate a standardized RID signal data sequence.

[0083] The electromagnetic evolution field construction module is used to calculate propagation difference characteristics based on standardized RID signal data sequences and construct an electromagnetic evolution field dataset.

[0084] The sensor point cloud processing module is used to collect and process detection data from UAV-borne lidar, millimeter-wave radar, and visual sensors to generate sensor point cloud sets.

[0085] The 3D environment perception model construction module is used to fuse the electromagnetic evolution field data set and the sensor point cloud set in a dual domain to generate a signal-space coupled 3D environment perception model.

[0086] The obstacle recognition and dynamic update module is used to perform obstacle recognition and dynamic update on the signal-space coupled 3D environment perception model to generate an obstacle avoidance environment model.

[0087] The flight control and path planning module is used to input the obstacle avoidance environment model into the UAV flight control unit, perform path planning and dynamic correction, and generate obstacle avoidance flight control commands.

[0088] The flight execution module is used to adjust the UAV's attitude and trajectory according to obstacle avoidance flight control commands, so as to achieve stable and reliable obstacle detection and autonomous obstacle avoidance in complex low-altitude environments.

[0089] In this embodiment, the modules are interconnected using the following method:

[0090] Collect RID remote identification signal data of UAVs during flight, perform unified format processing, and generate standardized RID signal data sequences;

[0091] Based on standardized RID signal data sequences, signal propagation differences are calculated, and signal intensity attenuation features, phase shift features, and temporal difference features are extracted to construct an electromagnetic evolution field dataset that reflects the state of the surrounding space environment of the UAV.

[0092] Collect detection data from UAV-borne lidar, millimeter-wave radar, and visual sensors; perform noise filtering, coordinate registration, and multi-source time synchronization on the detection data; and generate a sensor point cloud set.

[0093] The electromagnetic evolution field data set and the sensor point cloud set are fused in two domains to generate a signal-space coupled three-dimensional environmental perception model.

[0094] Obstacle identification and dynamic updating are performed on the signal-space coupled 3D environment perception model to generate an obstacle avoidance environment model that includes the location, size and motion state of obstacles.

[0095] The obstacle avoidance environment model is input into the UAV flight control unit to perform path planning and dynamic correction, and generate obstacle avoidance flight control commands.

[0096] The drone's attitude and trajectory are adjusted according to obstacle avoidance flight control commands to achieve stable and reliable obstacle detection and autonomous obstacle avoidance in complex low-altitude environments.

[0097] In this embodiment, the generation of the standardized RID signal data sequence includes the following specific steps:

[0098] During the flight of the UAV, RID remote identification signal data is collected, which includes signal strength, phase information and timing information;

[0099] The acquired RID remote identification signal data is subjected to noise filtering, interference suppression and time-frequency synchronization processing to generate a purified RID remote identification signal data set.

[0100] The signal strength of the purified RID remote identification signal data set is normalized to generate a signal strength sequence.

[0101] Phase difference processing is performed on the phase information of the purified RID remote identification signal data set to generate a phase difference sequence;

[0102] The time-series information of the purified RID remote identification signal data set is subjected to time-series differential processing to generate a time-series differential sequence;

[0103] The signal strength sequence, phase difference sequence, and timing difference sequence are processed to unify their formats and then spliced ​​together to generate a standardized RID signal data sequence.

[0104] The standardized RID signal data sequence is written into the UAV's local storage unit and simultaneously transmitted to the ground processing system.

[0105] In this embodiment, the construction of the electromagnetic evolution field dataset includes the following specific steps;

[0106] Signal strength attenuation characteristics, phase shift characteristics, and timing difference characteristics are calculated based on standardized RID signal data sequences. The signal strength attenuation characteristics are obtained by performing a logarithmic operation on the normalized signal strength. The phase shift characteristics are obtained by performing an accumulation operation on the phase difference information of historical sampling points. The timing difference characteristics are obtained by performing an averaging operation on the timing difference information within a preset time window.

[0107] ;

[0108] in, In time Signal strength attenuation characteristics at any given time. In time The normalized signal strength at any given time is used to characterize the change in spatial distance between the drone and the obstacle;

[0109] ;

[0110] in, In time Phase shift characteristics at time, This represents the number of historical sampling points for the cumulative phase difference. In time Phase difference fraction at any given moment;

[0111] ;

[0112] in, In time Temporal difference characteristics at time points, The length of the time-series difference window used for computation. In time The temporal difference value at each moment;

[0113] The signal strength attenuation characteristics, phase shift characteristics, and timing difference characteristics are combined under the same time index to form a propagation difference feature vector;

[0114] Based on the propagation difference feature vector, an improved Gaussian core field mapping method is used to generate a preliminary electromagnetic evolution field dataset, specifically including:

[0115] The kernel width parameter is determined by jointly considering the signal strength attenuation characteristics and timing difference characteristics in the propagation difference feature vector:

[0116] ;

[0117] in, For the first The kernel width parameter for each sampling point For the first Signal intensity attenuation characteristics at each sampling point For the first Temporal difference features of each sampling point This is a normalization factor for the signal strength attenuation characteristics, used to map the signal strength attenuation characteristics at different sampling points to a uniform scale. This is a normalization factor for temporal difference features, used to map temporal difference features from different sampling points to a uniform scale. For the first The time index corresponding to each sampling point describes the specific sampling time in the standardized RID signal data sequence;

[0118] The sampling point is an observation unit obtained by discretely sampling the standardized RID signal data sequence in the time and space dimensions, including the signal strength attenuation characteristics, phase shift characteristics, and temporal differential characteristics at that time and location.

[0119] The weighting factor is determined based on a combination of signal strength attenuation characteristics and phase shift characteristics.

[0120] ;

[0121] in, For the first Weighting factors for each sampling point For the first Phase offset characteristics of each sampling point For the first Signal intensity attenuation characteristics at each sampling point This is a normalization factor for the phase shift features, used to map the phase shift features to a uniform scale;

[0122] The normalization factor for the signal strength attenuation feature is the mean of the signal strength attenuation feature, the normalization factor for the timing difference feature is the standard deviation of the timing difference feature, and the normalization factor for the phase shift feature is the maximum absolute value of the phase shift feature.

[0123] The spatial location vector of the point to be calculated, along with the spatial location vectors of each sampling point, the kernel width parameter, and the weighting factor, are input into the Gaussian kernel function to calculate the electromagnetic evolution field values ​​at each spatial location under the global time index. A preliminary electromagnetic evolution field dataset is then generated by performing a weighted summation on the calculation results for all sampling points.

[0124] ;

[0125] in, In spatial coordinates With time Preliminary electromagnetic evolution field values, Let be the spatial position vector of the point to be calculated. For the first The spatial location vector of each sampling point Indicates the point to be calculated and the first... The squared Euclidean distance between each sampling point This represents the natural exponential function. Indicates the total number of samples. This represents the global time index when calculating the electromagnetic evolution field;

[0126] A temporal coherence threshold is set within consecutive time segments of the propagation difference feature vector. The initial electromagnetic evolution field values ​​corresponding to time segments whose temporal coherence index exceeds the threshold are reweighted to obtain the final electromagnetic evolution field dataset. Specifically, this involves: dividing the propagation difference feature vector into several consecutive time segments according to time index order; each time segment containing a set of propagation difference features corresponding to adjacent sampling points; and calculating a temporal coherence index for each time segment. This index measures the stability of the propagation difference features in the time dimension within that time segment. Specifically, the mean and standard deviation of the signal strength attenuation feature, phase shift feature, and temporal difference feature within that time segment are calculated respectively, and the ratio of the standard deviation to the mean is used as the normalization factor. The fluctuation value is calculated, and then the average of the three types of normalized fluctuation values ​​is taken to obtain the temporal coherence index of the time segment. When the temporal coherence index is greater than the preset temporal coherence threshold, it is determined that there is abnormal fluctuation in the propagation difference characteristics within the time segment, and the electromagnetic evolution field value corresponding to the time segment is reweighted, that is, the electromagnetic evolution field value of the time segment is multiplied by a shrinkage factor to reduce its influence in the overall electromagnetic evolution field dataset. When the temporal coherence index is less than or equal to the preset temporal coherence threshold, the electromagnetic evolution field value corresponding to the time segment remains unchanged. Through this method, dynamic smoothing and anomaly suppression of propagation difference characteristics can be achieved during the construction of the electromagnetic evolution field, thereby ensuring the continuity and stability of the electromagnetic evolution field dataset in the time dimension.

[0127] The generation of the preset temporal coherence threshold includes: acquiring multiple standardized RID signal data sequences, extracting propagation difference feature vectors corresponding to consecutive time segments, and calculating the temporal coherence index of each time segment according to the time index order; performing statistical analysis on all temporal coherence indices, and calculating the mean and standard deviation of the overall distribution; and setting the preset temporal coherence threshold as the sum of the mean of the temporal coherence index and twice the standard deviation.

[0128] The contraction factor is generated by: calculating the difference between the temporal coherence index and the temporal coherence threshold as the overthreshold, and then using the overthreshold as the exponent to calculate the value of the negative exponent with base e to obtain the contraction factor.

[0129] In this embodiment, the generation of the sensor point cloud set includes the following specific steps:

[0130] During the flight of the UAV, point cloud data from airborne lidar, point cloud data from millimeter-wave radar, and visual depth data are collected respectively to form lidar point cloud sequences, millimeter-wave radar point cloud sequences, and visual depth sequences.

[0131] Noise is filtered out from the lidar point cloud sequence, millimeter-wave radar point cloud sequence, and visual depth sequence to obtain filtered lidar point cloud data, millimeter-wave radar point cloud data, and visual depth data.

[0132] The noise-filtered data are transformed into the UAV's body coordinate system to achieve coordinate unification, so that the three types of data can be represented under the same spatial reference.

[0133] Coordinate registration of the three types of data is performed in the UAV's body coordinate system;

[0134] Based on the completion of coordinate registration, time synchronization is performed to uniformly process the timestamps of lidar point cloud data, millimeter-wave radar point cloud data, and visual depth data.

[0135] The processed lidar point cloud data, millimeter-wave radar point cloud data, and visual depth data are fused to generate a sensor point cloud set in a unified format.

[0136] In this embodiment, the generation of the three-dimensional environment perception model includes the following specific steps:

[0137] Acquire electromagnetic evolution field data sets and sensor point cloud sets, and use them as inputs for dual-domain fusion;

[0138] Establish a correspondence between the electromagnetic evolution field dataset and the sensor point cloud dataset, calculate the Euclidean distance between the spatial position vector in the electromagnetic evolution field dataset and the spatial position vector in the sensor point cloud dataset, and use it as the spatial proximity.

[0139] The spatial proximity is compared with a preset spatial proximity threshold. All point pairs that are less than or equal to the spatial proximity threshold are retained as candidate matching point pairs, and a fusion weight is calculated for the candidate matching point pairs.

[0140] ;

[0141] in, The fusion weights are the fusion weights of the i-th sampling point and the j-th sensor point cloud. Let be the electromagnetic evolution field value of the i-th sampling point at the global time index t. For the first Electromagnetic evolution field values ​​at each sampling point under the global time index t. This is the sum of the electromagnetic evolution field values ​​at all n sampling points under the global time index t. Let i be the spatial location vector of the i-th sampling point. Let j be the spatial location vector of the j-th point in the sensor point cloud set. For spatial scale parameters, This represents the total number of sampling points in the electromagnetic evolution field dataset;

[0142] The electromagnetic evolution field values ​​are mapped to points in the sensor point cloud set using fusion weights, where the enhancement value corresponding to the j-th point in the sensor point cloud set is:

[0143] ;

[0144] in, The enhanced value of the j-th point in the sensor point cloud after fusing electromagnetic evolution field features;

[0145] The enhancement value refers to the signal attribute value obtained by fusing the features of each point in the sensor point cloud set with the electromagnetic evolution field data set. The enhancement value is added to the signal attribute value of the point in the sensor point cloud set as a fourth dimension to the three-dimensional spatial coordinates of the point cloud, so that each point not only has a position, but also has electromagnetic characteristic intensity. The enhancement value is used to characterize the response intensity of the spatial point in the electromagnetic propagation characteristics, so that the signal-enhanced point cloud set contains both geometric spatial information and electromagnetic evolution characteristics.

[0146] The merged point cloud set is stored as a unified 3D data structure. ;

[0147] The three-dimensional part of the three-dimensional data structure is used to store the spatial coordinates of each point in the sensor point cloud set. The enhancement value is stored as an additional attribute of the point and bound to the spatial position, ensuring that the fused point cloud set maintains three-dimensional consistency in spatial geometry and carries electromagnetic evolution field feature information at the attribute level.

[0148] A signal-space coupled 3D environment perception model is constructed based on a 3D data structure, so that the 3D environment perception model simultaneously includes the propagation characteristics in the electromagnetic evolution field data set and the geometric characteristics in the sensor point cloud set.

[0149] In this embodiment, the generation of the obstacle avoidance environment model includes the following specific steps:

[0150] Acquire a signal-space coupled 3D environment perception model, which includes the spatial location and augmentation value of each point in the sensor point cloud set;

[0151] Spatial segmentation is performed on the signal-space coupled 3D environment perception model, and the sensor point cloud set is divided into multiple candidate regions using a density-based clustering method;

[0152] Calculate the point density and electromagnetic consistency index in each candidate region:

[0153] ;

[0154] in, Let the point density be the k-th candidate region. The number of points within the candidate region. For the candidate region volume, The electromagnetic consistency index for the k-th candidate region is... Let j be the enhancement value of the j-th point within the candidate region. The number of points within the candidate region;

[0155] The candidate region is determined as an obstacle region based on a combination of point density and electromagnetic consistency index. When both exceed a preset threshold, the candidate region is identified as an obstacle region.

[0156] The preset thresholds when both exceed the preset thresholds include a point density threshold and an electromagnetic consistency threshold. The point density threshold is obtained by statistically analyzing the point density distribution of obstacle areas and non-obstacle areas in the flight test samples, and the mean plus twice the standard deviation is taken as the judgment criterion. The electromagnetic consistency threshold is obtained by statistically analyzing the electromagnetic consistency index of obstacle areas in the test samples, and the mean minus one standard deviation is taken as the threshold, thereby ensuring the accuracy and robustness of obstacle area judgment.

[0157] Perform 3D boundary fitting on the point set within the obstacle region, calculate the circumbound box of the obstacle in 3D space, and let the center of the circumbound box represent... The length, width, and height dimensions are respectively represented as follows: This forms the spatial geometric parameters of the obstacle;

[0158] Track the change in the center position of the obstacle region under continuous global time indexing, and calculate the motion velocity vector of the obstacle:

[0159] ;

[0160] in, Let m be the velocity vector of the obstacle. For obstacles in time The central location, For obstacles in time The central location, For continuous global time index;

[0161] By combining the spatial geometric parameters and motion velocity vectors of obstacles, an obstacle avoidance environment model is generated. The obstacle avoidance environment model includes the position, size, and motion state of the obstacles and is described and output in a unified structure for the UAV flight control unit to perform path planning and dynamic correction.

[0162] In this embodiment, the generation of the obstacle avoidance flight control command includes the following specific steps:

[0163] Obtain the obstacle avoidance environment model and use it as input to the UAV flight control unit;

[0164] In the UAV flight control unit, the current position vector and the target position vector of the UAV are defined, and the flight path position vector is defined. The flight path is a time continuous function that must satisfy the reachability from the current position vector of the UAV to the target position vector.

[0165] Construct a path cost function to simultaneously consider flight distance and obstacle avoidance safety:

[0166] ;

[0167] in, For path cost function, Let be the flight path position vector at global time index t. Let the target position vector be... Let m be the position of the m-th obstacle at global time index t. The total number of obstacles. Weighted by flight distance, For obstacle avoidance safety weights, This is the termination time index for path planning. This is the start time index for path planning;

[0168] The path cost function uses the flight path as the overall optimization variable and performs joint optimization on its position sequence under continuous global time indices. The result obtained after minimizing the path cost function is a path function containing multiple time index position points. This path function extends from the starting point to the target point, forming the initial planning path of the UAV.

[0169] The path cost function is minimized to obtain the initial planned path, which represents the set of path points optimized under the global time index.

[0170] During flight, the obstacle avoidance environment model is updated in real time, and the updated obstacle positions are obtained, with a time interval set. The predicted location of internal obstacles is represented as The updated obstacle positions are then substituted into the path cost function to obtain the corrected cost function.

[0171] The local path is recalculated within a local range based on the modified cost function to obtain the dynamic modified path, which represents the local optimal path under the condition of obstacle position update.

[0172] The modified cost function replaces the obstacle position with a dynamic position predicted based on the movement speed in the path cost function, so that the path optimization can adapt to the movement state of the obstacle in real time.

[0173] The dynamic correction path is merged with the initial planned path to generate the final obstacle avoidance path, and the final obstacle avoidance path is discretized into a sequence of flight control commands.

[0174] The generation of the final obstacle avoidance path includes: obtaining an initial planned path and a dynamically corrected path, both of which are discretized according to a global time index; comparing the corresponding path points of the two paths under each global time index, calculating the spatial distance difference between the initial planned path point and the dynamically corrected path point; when the spatial distance difference is lower than a preset safety threshold, the initial planned path point is directly used as the final obstacle avoidance path point; when the spatial distance difference is higher than the preset safety threshold, the final obstacle avoidance path point is calculated according to a weighted fusion rule, that is, the dynamically corrected path point is used as the primary reference and the initial planned path point is used as the secondary reference, and a weighted average position is obtained through weight allocation to ensure that the path point of the UAV under this time index can avoid obstacles; after completing the point-by-point processing of all time indices, the obtained path point sequence is sequentially connected to generate a continuous final obstacle avoidance path. This path inherits the overall rationality of the initial planned path in the global scope, while having the real-time obstacle avoidance capability of the dynamically corrected path in the local scope, thus taking into account both flight efficiency and safety.

[0175] Example 1:

[0176] To verify the feasibility of this invention in practice, it was applied to a verification environment of an urban low-altitude logistics route. This environment is located at the junction of the city's edge and core area, surrounded by numerous high-rise buildings, traffic overpasses, and moving obstacles, making it a scenario with high low-altitude environmental complexity. In this environment, the drone needs to perform cargo transportation tasks from the logistics warehousing center to the distribution point. The route traverses areas including narrow building passages, densely wooded green belts, and low-altitude corridors interfered with by traffic vehicles. Due to the complexity of natural lighting conditions, environmental obstructions, and electromagnetic interference sources, traditional detection methods relying on visual sensors and radar often suffer from reduced recognition accuracy, delayed path planning, or even failure, thus failing to guarantee the stability and safety of the drone during flight.

[0177] In this verification environment, according to the method of the present invention, the UAV first activates the RID remote identification signal broadcasting function during the takeoff phase. At the same time, the onboard processing unit begins to collect the intensity, phase and timing information of the RID signal in real time. After noise filtering and format unification processing, the generated standardized RID signal data sequence is input to the electromagnetic evolution field construction unit. During this process, the propagation difference characteristics of the RID signal are gradually extracted and mapped into an electromagnetic evolution field data set, enabling the UAV to obtain a stable representation of the space environment even in areas with insufficient light and radar interference. Unlike the traditional application method that only uses identity broadcasting, the extended use of the RID signal in the present invention gives the UAV an additional source of perception.

[0178] Meanwhile, airborne lidar, millimeter-wave radar, and visual sensors continuously collect data on the surrounding environment. After noise filtering, coordinate registration, and multi-source time synchronization, a unified format sensor point cloud set is generated. Subsequently, the electromagnetic evolution field data set and the sensor point cloud set are input into the three-dimensional environment perception model construction unit. Through dual-domain fusion, a signal-space coupled three-dimensional environment perception model is generated. Unlike the traditional method that relies solely on the geometric features of the point cloud, this model carries both spatial geometry and electromagnetic propagation characteristics information, enabling it to maintain highly robust perception of the environment within the limited area of ​​the sensors.

[0179] In the middle of the flight path, the UAV passes through narrow passages between buildings. Traditional radar and visual sensors are significantly affected by obstruction at this point, resulting in unstable obstacle boundary detection results. However, through the dual-domain fusion method of this invention, the system can combine electromagnetic evolution fields with point cloud features to identify hidden obstacles inside the passage and dynamically update the spatial position and motion state of the obstacles, generating an obstacle avoidance environment model. This model not only records the position and geometric dimensions of the obstacles but also tracks their movement speed in real time, providing reliable data for subsequent path planning.

[0180] During the flight control phase, the obstacle avoidance environment model is input into the UAV flight control unit. Obstacle avoidance flight control commands are generated through path planning and dynamic correction mechanisms. In global path planning, the system ensures the flight efficiency and overall trajectory rationality of the UAV. When encountering dynamic obstacles, the correction cost function is invoked. The system calculates the dynamically corrected path based on the predicted position of the obstacle and compares and weights it with the original path point by point to generate the final obstacle avoidance path that balances safety and continuity. In actual flight, this mechanism enables the UAV to quickly adjust its trajectory when encountering sudden moving obstacles, avoiding the risk of collision due to path rigidity.

[0181] Throughout the application process, the system of this invention exhibits significant advantages over existing technologies. Due to the introduction of an electromagnetic evolution field dataset constructed based on RID signals, even when insufficient lighting leads to a decline in visual recognition performance or radar detection is interfered with, the UAV can still rely on electromagnetic evolution field characteristics for stable obstacle recognition. Through fusion with point cloud geometric features, the three-dimensional environment perception model maintains high accuracy and high stability in complex environments. Experimental comparisons show that the UAV using the method of this invention is significantly superior to solutions that only use radar or vision in terms of obstacle recognition accuracy, path planning continuity, and flight safety.

[0182] Furthermore, the fusion mechanism of dynamic path correction and initial path enables the UAV to not only maintain a reasonable flight path globally, but also to respond quickly to environmental changes in a local area. Through the real-time generated obstacle avoidance environment model, the UAV can make avoidance actions in advance according to the movement trend of obstacles, avoiding sudden turns in emergency situations, thereby ensuring the stability of the flight process and energy consumption control.

[0183] To verify the performance of the present invention in practice, it was compared with traditional methods.

[0184] Table 1. Performance Comparison of RID-based Air-Ground Integrated UAV Detection and Obstacle Avoidance System with Traditional Methods

[0185]

[0186] As shown in Table 1, the method of this invention outperforms the traditional obstacle avoidance method combining radar and vision in complex low-altitude environments. The obstacle recognition accuracy reaches 92.6%, which is 10.3 percentage points higher than the 82.3% of the traditional method. This difference is due to the electromagnetic evolution field modeling of RID signals introduced in this invention, which enables the UAV to obtain stable obstacle representations in low light, strong interference and complex obstruction environments, thus making up for the perception blind spots of vision and radar under extreme conditions.

[0187] In terms of path planning continuity, this invention achieves 91.2%, which is 14.4 percentage points higher than the traditional method's 76.8%. This is thanks to the signal-space coupled three-dimensional environment perception model, which not only provides geometric information but also adds electromagnetic feature information, enabling UAVs to correct their paths more smoothly when the environment changes dynamically, avoiding frequent interruptions or unreasonable track jumps.

[0188] The flight safety index in this invention reaches 96.8%, which is higher than the 85.1% of the traditional method. This improvement comes from the introduction of obstacle dynamic update and correction cost function, which enables the UAV to adjust its trajectory in time when it encounters moving obstacles, thereby reducing the risk of collision.

[0189] In terms of average track deviation, the present invention has a deviation of only 0.45 meters, which is 0.48 meters less than the 0.93 meters of the traditional method. This shows that the UAV can maintain higher accuracy when performing global planning and local dynamic correction path fusion, the track is closer to the ideal trajectory, and the flight process is more stable.

[0190] In terms of energy consumption control, the energy consumption control rate of the present invention reaches 89.7%, which is 11.5% higher than the 78.2% of the traditional method. This is because the present invention avoids frequent sharp turns and ineffective detours, and combines a weighted fusion mechanism of dynamic path correction and initial planned path, so that the UAV can reduce additional energy consumption while ensuring safety.

[0191] In summary, this invention breaks through the reliance on a single sensor in traditional methods by fusing the propagation characteristics of RID signals with point clouds for modeling. This results in higher obstacle recognition accuracy, better path continuity, safer flight performance, and better energy consumption control, verifying the advantages of this invention in complex low-altitude environments.

[0192] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A ground-based air-to-air unmanned aerial vehicle (UAV) detection and obstacle avoidance system based on RID, characterized in that, include: The RID signal processing module is used to collect RID remote identification signal data of the UAV during flight, and to perform unified format processing on the signal data to generate a standardized RID signal data sequence. The electromagnetic evolution field construction module is used to calculate propagation difference characteristics based on standardized RID signal data sequences and construct an electromagnetic evolution field dataset. The sensor point cloud processing module is used to collect and process detection data from UAV-borne lidar, millimeter-wave radar, and visual sensors to generate sensor point cloud sets. The 3D environment perception model construction module is used to fuse the electromagnetic evolution field data set and the sensor point cloud set in a dual domain to generate a signal-space coupled 3D environment perception model. The obstacle recognition and dynamic update module is used to perform obstacle recognition and dynamic update on the signal-space coupled 3D environment perception model to generate an obstacle avoidance environment model. The flight control and path planning module is used to input the obstacle avoidance environment model into the UAV flight control unit, perform path planning and dynamic correction, and generate obstacle avoidance flight control commands. The flight execution module is used to adjust the attitude and trajectory of the UAV according to the obstacle avoidance flight control command, so as to achieve stable and reliable obstacle detection and autonomous obstacle avoidance in complex low-altitude environments; The generation of the standardized RID signal data sequence includes the following specific steps: During the flight of the UAV, RID remote identification signal data is collected, which includes signal strength, phase information and timing information; The acquired RID remote identification signal data is subjected to noise filtering, interference suppression and time-frequency synchronization processing to generate a purified RID remote identification signal data set. The signal strength of the purified RID remote identification signal data set is normalized to generate a signal strength sequence. Phase difference processing is performed on the phase information of the purified RID remote identification signal data set to generate a phase difference sequence; The time-series information of the purified RID remote identification signal data set is subjected to time-series differential processing to generate a time-series differential sequence; The signal strength sequence, phase difference sequence, and timing difference sequence are processed to unify their formats and then spliced ​​together to generate a standardized RID signal data sequence. Write the standardized RID signal data sequence into the drone's local storage unit; The construction of the electromagnetic evolution field dataset includes the following specific steps: Signal strength attenuation characteristics, phase shift characteristics, and timing difference characteristics are calculated based on standardized RID signal data sequences. The signal strength attenuation characteristics are obtained by performing a logarithmic operation on the normalized signal strength. The phase shift characteristics are obtained by performing an accumulation operation on the phase difference information of historical sampling points. The timing difference characteristics are obtained by performing an averaging operation on the timing difference information within a preset time window. The signal strength attenuation characteristics, phase shift characteristics, and timing difference characteristics are combined under the same time index to form a propagation difference feature vector; Based on the propagation difference feature vector, an improved Gaussian core field mapping method is used to generate a preliminary electromagnetic evolution field dataset, specifically including: The kernel width parameter is determined by jointly considering the signal strength attenuation characteristics and timing difference characteristics in the propagation difference feature vector; The weighting factor is determined by combining the signal strength attenuation characteristics and the phase shift characteristics. The spatial position vector of the point to be calculated, the spatial position vector of each sampling point, the kernel width parameter, and the weighting factor are input into the Gaussian kernel function to calculate the electromagnetic evolution field value of each spatial position under the global time index. The preliminary electromagnetic evolution field dataset is generated by performing a weighted summation on the calculation results of all sampling points. Within the continuous time segments of the propagation difference feature vector, a temporal coherence threshold is set, and the preliminary electromagnetic evolution field values ​​corresponding to the time segments whose temporal coherence index exceeds the temporal coherence threshold are reweighted to obtain the final electromagnetic evolution field dataset.

2. The air-to-ground integrated UAV detection and obstacle avoidance system based on RID according to claim 1, characterized in that, The modules are connected in the following way: Collect RID remote identification signal data of UAVs during flight, perform unified format processing, and generate standardized RID signal data sequences; Based on standardized RID signal data sequences, signal propagation differences are calculated, and signal intensity attenuation features, phase shift features, and temporal difference features are extracted to construct an electromagnetic evolution field dataset that reflects the state of the surrounding space environment of the UAV. Collect detection data from UAV-borne lidar, millimeter-wave radar, and visual sensors; perform noise filtering, coordinate registration, and multi-source time synchronization on the detection data; and generate a sensor point cloud set. The electromagnetic evolution field data set and the sensor point cloud set are fused in two domains to generate a signal-space coupled three-dimensional environmental perception model. Obstacle identification and dynamic updating are performed on the signal-space coupled 3D environment perception model to generate an obstacle avoidance environment model that includes the location, size and motion state of obstacles. The obstacle avoidance environment model is input into the UAV flight control unit to perform path planning and dynamic correction, and generate obstacle avoidance flight control commands. The drone's attitude and trajectory are adjusted according to obstacle avoidance flight control commands to achieve stable and reliable obstacle detection and autonomous obstacle avoidance in complex low-altitude environments.

3. The air-to-ground integrated UAV detection and obstacle avoidance system based on RID according to claim 2, characterized in that, The generation of the sensor point cloud set includes the following specific steps: During the flight of the UAV, point cloud data from airborne lidar, point cloud data from millimeter-wave radar, and visual depth data are collected respectively to form lidar point cloud sequences, millimeter-wave radar point cloud sequences, and visual depth sequences. Noise is filtered out from the lidar point cloud sequence, millimeter-wave radar point cloud sequence, and visual depth sequence to obtain filtered lidar point cloud data, millimeter-wave radar point cloud data, and visual depth data. The noise-filtered data are transformed into the UAV's body coordinate system to achieve coordinate unification, so that the three types of data can be represented under the same spatial reference. Coordinate registration of the three types of data is performed in the UAV's body coordinate system; Based on the completion of coordinate registration, time synchronization is performed to uniformly process the timestamps of lidar point cloud data, millimeter-wave radar point cloud data, and visual depth data. The processed lidar point cloud data, millimeter-wave radar point cloud data, and visual depth data are fused to generate a sensor point cloud set in a unified format.

4. The air-to-ground integrated UAV detection and obstacle avoidance system based on RID according to claim 2, characterized in that, The generation of the three-dimensional environment perception model includes the following specific steps: Acquire electromagnetic evolution field data sets and sensor point cloud sets, and use them as inputs for dual-domain fusion; Establish a correspondence between the electromagnetic evolution field dataset and the sensor point cloud dataset, calculate the Euclidean distance between the spatial position vector in the electromagnetic evolution field dataset and the spatial position vector in the sensor point cloud dataset, and use it as the spatial proximity. The spatial proximity is compared with a preset spatial proximity threshold. All point pairs that are less than or equal to the spatial proximity threshold are retained as candidate matching point pairs, and a fusion weight is calculated for the candidate matching point pairs. ； in, The fusion weights are the fusion weights of the i-th sampling point and the j-th sensor point cloud. Let be the electromagnetic evolution field value of the i-th sampling point at the global time index t. For the first Electromagnetic evolution field values ​​at each sampling point under the global time index t. This is the sum of the electromagnetic evolution field values ​​at all n sampling points under the global time index t. Let i be the spatial location vector of the i-th sampling point. Let j be the spatial location vector of the j-th point in the sensor point cloud set. For spatial scale parameters, This represents the total number of sampling points in the electromagnetic evolution field dataset; The electromagnetic evolution field values ​​are mapped to points in the sensor point cloud set using fusion weights, where the enhancement value corresponding to the j-th point in the sensor point cloud set is: ； in, The enhanced value of the j-th point in the sensor point cloud after fusing electromagnetic evolution field features; The merged point cloud set is stored as a unified three-dimensional data structure; A signal-space coupled 3D environment perception model is constructed based on a 3D data structure.

5. The RID-based integrated air-to-ground UAV detection and obstacle avoidance system according to claim 2, characterized in that, The generation of the obstacle avoidance environment model includes the following specific steps: Acquire a signal-space coupled 3D environment perception model, which includes the spatial location and augmentation value of each point in the sensor point cloud set; Spatial segmentation is performed on the signal-space coupled 3D environment perception model, and the sensor point cloud set is divided into multiple candidate regions using a density-based clustering method; Calculate the point density and electromagnetic consistency index in each candidate region; The candidate region is determined as an obstacle region based on a combination of point density and electromagnetic consistency index. When both exceed a preset threshold, the candidate region is identified as an obstacle region. The point set within the obstacle region is fitted with a three-dimensional boundary, and the bounding box of the obstacle in three-dimensional space is calculated to form the spatial geometric parameters of the obstacle. Track the change in the center position of the obstacle region under continuous global time indexing, and calculate the motion velocity vector of the obstacle; By combining the spatial geometric parameters and motion velocity vectors of obstacles, an obstacle avoidance environment model is generated, which includes the position, size, and motion state of the obstacles.

6. The RID-based integrated air-to-ground UAV detection and obstacle avoidance system according to claim 2, characterized in that, The generation of the obstacle avoidance flight control command includes the following specific steps: Obtain the obstacle avoidance environment model and use it as input to the UAV flight control unit; In the UAV flight control unit, the current position vector of the UAV and the target position vector are defined, and the flight path position vector is also defined; Construct a path cost function to simultaneously consider flight distance and obstacle avoidance safety: ； in, For path cost function, Let be the flight path position vector at global time index t. Let the target position vector be... Let m be the position of the m-th obstacle at global time index t. The total number of obstacles. Weighted by flight distance, For obstacle avoidance safety weights, This is the termination time index for path planning. This is the start time index for path planning; The initial planned path is obtained by minimizing the path cost function. During flight, the obstacle avoidance environment model is updated in real time, and the updated obstacle positions are obtained. The updated obstacle positions are then substituted into the path cost function to obtain the corrected cost function. The local path is recalculated within a local range based on the modified cost function to obtain a dynamically modified path. The dynamically corrected path is merged with the initial planned path to generate the final obstacle avoidance path, which is then discretized into a sequence of flight control commands.

Images