LiDAR-based autonomous obstacle avoidance method and device for unmanned aerial vehicles

Abstract

This application provides a method and apparatus for autonomous obstacle avoidance of unmanned aerial vehicles (UAVs) based on lidar, relating to the field of UAV technology. By extracting the optical features of the surface material of the current area and matching the optimal band interval based on a pre-built feature fingerprint database, the emission wavelength of the lidar can adapt to the main constituent materials of the environment, significantly improving the signal-to-noise ratio and feature recognition of the echo signal. A programmable gain circuit is introduced at the receiving end to dynamically compensate for signal strength changes caused by band switching or different distances, improving the reliability and consistency of point cloud data generation. This ensures that the generated "optimized obstacle avoidance path" not only effectively avoids marked obstacle points but also better conforms to actual environmental constraints, significantly improving the perception reliability and obstacle recognition accuracy of the UAV lidar in complex environments containing various heterogeneous materials, thereby comprehensively enhancing the safety of autonomous obstacle avoidance and the environmental adaptability of the overall system.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and in particular to a method and apparatus for autonomous obstacle avoidance of UAVs based on lidar. Background Technology

[0002] With the rapid development of infrastructure construction such as mountain roads and railways, a large amount of waste generated from these projects needs to be dumped in specialized waste disposal sites with complex terrain. These sites have varied surrounding environments, including natural and man-made obstacles such as asphalt pavements, exposed rocks, metal fences, and vegetation. Regular inspections are necessary to ensure the safety of these waste disposal sites. Drones, with their flexibility and efficiency, have become an important tool for these inspections. However, achieving reliable autonomous obstacle avoidance for drones in complex environments still faces significant challenges.

[0003] Currently, most lidar-based autonomous obstacle avoidance technologies for drones employ fixed-band and fixed-gain signal processing modes. This approach ignores the significant differences in the reflection characteristics of laser signals from different environmental surface materials. For example, different materials such as asphalt, rock, metal, and vegetation exhibit varying reflectivity, absorptivity, and scattering characteristics under different laser bands. Under fixed parameters, lidar may saturate due to excessively strong signals for some materials, while attenuating due to insufficient signals for others, resulting in unstable echo signal quality. This directly leads to false detections and missed detections of obstacles, reducing perception reliability and consequently affecting the accuracy of drone obstacle avoidance path planning and the ultimate flight safety.

[0004] Therefore, how to enable the lidar system of UAVs to adapt to the optical properties of different surface materials and dynamically optimize detection parameters in order to obtain stable and high-quality environmental perception data has become a key technical challenge to improve the autonomous obstacle avoidance performance of UAVs in complex environments. Summary of the Invention

[0005] The purpose of this invention is to provide an autonomous obstacle avoidance method and device for unmanned aerial vehicles (UAVs) based on lidar. By extracting the optical features of the surface material in the current area and matching the optimal band interval based on a pre-built feature fingerprint database, adaptive optimization of the transmission wavelength is achieved, thereby improving the quality of the echo signal from the source. By performing programmable gain adjustment of the echo signal at the receiving end according to the matched band, the stability and reliability of the point cloud data are ensured. Based on the obtained high-quality point cloud, the obstacle recognition accuracy is significantly improved, thereby making the environmental model on which path planning depends more accurate.

[0006] In a first aspect, the present invention provides an autonomous obstacle avoidance method for unmanned aerial vehicles (UAVs) based on lidar, applied to the control system of the UAV. The transmitting end of the lidar of the UAV includes a band selection circuit; the receiving end of the lidar includes a programmable gain circuit; the method includes:

[0007] The drone's current location is scanned globally using lidar to obtain the initial global echo signal.

[0008] The global initial echo signal is clustered, and the first cluster echo signal is determined based on the signal proportion of multiple cluster echo signals in the clustering results.

[0009] The real-time feature vector of the first cluster echo signal is extracted as the optical feature of the surface material of the current region;

[0010] Based on the optical characteristics of the surface material in the current region, search for the matching band interval of the current region in the preset surface material optical characteristic fingerprint database;

[0011] Adjust the band selection circuit to make the lidar transmit the first test signal in the matching band interval;

[0012] Receive the first echo signal of the current area corresponding to the first test signal, and perform gain processing on the first echo signal based on the programmable gain circuit to obtain the first gain echo signal of the current area;

[0013] Perform point cloud analysis on the first gain echo signal of the current area to obtain the point cloud analysis dataset of the current area;

[0014] Based on the point cloud parsing dataset, the obstacle clusters in the current area are determined, and the expected planned path is optimized based on the obstacle clusters to obtain the optimized obstacle avoidance path.

[0015] In some preferred embodiments of the present invention, the step of clustering the global initial echo signal and determining the first clustered echo signal based on the signal proportion of multiple clustered echo signals in the clustering results includes:

[0016] Clustering of the initial global echo signal yields multiple clustered echo signals;

[0017] Calculate the percentage of echo points contained in each cluster of echo signals out of the total number of echo points;

[0018] The cluster echo signal with the largest percentage was determined as the first cluster echo signal.

[0019] In some preferred embodiments of the present invention, the method further includes:

[0020] The feature vector samples of different surface materials under different test bands are obtained; the feature vector samples include echo intensity distribution, echo signal attenuation ratio and multiple echo characteristics.

[0021] Based on the band response sensitivity samples and the preset optimization objectives, the matching band intervals corresponding to each surface material sample are determined; wherein, the optimization objectives include at least one of the following: signal-to-noise ratio objective, feature enhancement objective, and material discrimination objective;

[0022] Establish a mapping relationship between different surface material samples, optimized targets and matching band intervals to construct a surface material optical feature fingerprint library.

[0023] In some preferred embodiments of the present invention, the step of searching for a matching band interval of the current region in a preset surface material optical feature fingerprint database based on the surface material optical features of the current region includes:

[0024] The similarity between the optical features of the surface material in the current region and the feature vectors of each surface material sample pre-stored in the surface material optical feature fingerprint database is calculated.

[0025] The matching feature vector is determined based on the similarity calculation results;

[0026] Based on the matching feature vector and the currently selected optimization objective, the corresponding matching band interval is found from the mapping relationship.

[0027] In some preferred embodiments of the present invention, the method further includes:

[0028] If the rate of change of the signal proportion is greater than the preset threshold, the first cluster echo signal is redefined.

[0029] Based on the surface material optical characteristics corresponding to the redefined first cluster echo signal, the corresponding matching band interval is searched again.

[0030] Adjust the band selection circuit according to the redefined matching band range.

[0031] In some preferred embodiments of the present invention, the programmable gain circuit includes: a programmable bandpass filter; the parameters of the programmable bandpass filter correspond one-to-one with the matched band intervals; the step of performing gain processing on the first echo signal based on the programmable gain circuit includes:

[0032] Determine the target parameters of the programmable bandpass filter based on the matched band interval;

[0033] Configure a programmable bandpass filter based on target parameters so that the programmable bandpass filter can filter the first echo signal; wherein, the target parameters include the center frequency and bandwidth corresponding to the matching band interval.

[0034] In some preferred embodiments of the present invention, the programmable gain circuit further includes a variable gain amplifier; the step of performing gain processing on the first echo signal based on the programmable gain circuit further includes:

[0035] A variable gain amplifier is used to amplify the filtered first echo signal; the target gain factor of the variable gain amplifier is determined based on the following steps:

[0036] First-order gain is calculated based on the signal amplitude, signal-to-noise ratio and saturation of the first echo signal, and the first-order gain factor is output.

[0037] The first-order gain factor was applied to the variable gain amplifier for testing, and the second echo signal was obtained.

[0038] Determine whether the signal amplitude of the second echo signal meets the preset target peak window;

[0039] If satisfied, the first-order gain factor is determined as the target gain factor of the variable gain amplifier;

[0040] If the condition is not met, the first-order gain factor is iteratively adjusted based on the error between the signal amplitude of the second echo signal and the target peak window until the output echo signal meets the target peak window, and the finally adjusted gain factor is determined as the target gain factor.

[0041] In some preferred embodiments of the present invention, the step of determining obstacle clusters in the current region based on a point cloud parsing dataset and optimizing the expected planned path based on the obstacle clusters to obtain an optimized obstacle avoidance path includes:

[0042] Perform obstacle clustering based on the point cloud parsing dataset, and output at least one obstacle cluster;

[0043] Calculate the conflict probability between each obstacle cluster and the expected planned path of the drone, and mark the obstacle clusters with a conflict probability greater than the preset conflict probability as obstacle avoidance points;

[0044] By constraining the avoidance of all obstacle points, the expected planned path is replanned to generate an optimized obstacle avoidance path.

[0045] In some preferred embodiments of the present invention, the step of calculating the conflict probability between each obstacle cluster and the expected planned path of the UAV, and marking obstacle clusters with conflict probabilities greater than a preset conflict probability as obstacle avoidance points, includes:

[0046] Calculate the shortest spatial distance between the outer contour of the obstacle cluster and the expected planned path;

[0047] The collision probability is calculated based on the shortest spatial distance, the safe radius of the drone, the relative speed of the drone, the positioning error of the drone, and the control error of the drone.

[0048] Obstacle clusters with a collision probability greater than the preset conflict probability are marked as obstacle avoidance points.

[0049] Secondly, the present invention provides an autonomous obstacle avoidance device for unmanned aerial vehicles (UAVs) based on lidar, applied to the control system of a UAV. The transmitting end of the UAV's lidar includes a band selection circuit; the receiving end of the lidar includes a programmable gain circuit; the device includes:

[0050] The initial echo signal determination module is used to perform a global scan of the area where the UAV is currently located based on the lidar to obtain the global initial echo signal.

[0051] The first cluster echo signal determination module is used to cluster the global initial echo signal and determine the first cluster echo signal based on the signal proportion of multiple cluster echo signals in the clustering results.

[0052] The surface material optical feature determination module is used to extract the real-time feature vector of the first cluster echo signal as the surface material optical feature of the current region;

[0053] The matching band interval determination module is used to search for the matching band interval of the current region in a preset surface material optical feature fingerprint database based on the surface material optical features of the current region.

[0054] The radar signal transmission module is used to adjust the band selection circuit so that the lidar transmits the first test signal in the matching band interval;

[0055] The radar signal receiving module is used to receive the first echo signal of the current area corresponding to the first test signal, and to perform gain processing on the first echo signal based on the programmable gain circuit to obtain the first gain echo signal of the current area.

[0056] The point cloud analysis data processing module is used to perform point cloud analysis on the first gain echo signal of the current area to obtain the point cloud analysis dataset of the current area.

[0057] The obstacle avoidance path processing module is used to determine the obstacle clusters in the current area based on the point cloud parsing dataset, and to optimize the expected planned path based on the obstacle clusters to obtain the optimized obstacle avoidance path.

[0058] This invention brings the following beneficial effects:

[0059] This invention provides a method and apparatus for autonomous obstacle avoidance of unmanned aerial vehicles (UAVs) based on lidar, applied to the control system of a UAV. The transmitting end of the UAV's lidar includes a band selection circuit; the receiving end of the lidar includes a programmable gain circuit. The method includes: performing a global scan of the area currently occupied by the UAV based on the lidar to obtain a global initial echo signal; clustering the global initial echo signal and determining a first clustered echo signal based on the signal proportion of multiple clustered echo signals in the clustering results; extracting the real-time feature vector of the first clustered echo signal as the surface material optical feature of the current area; searching for a matching band interval of the current area in a preset surface material optical feature fingerprint database based on the surface material optical feature of the current area; adjusting the band selection circuit so that the lidar transmits a first test signal in the matching band interval; and receiving the corresponding signal of the first test signal. The first echo signal of the current region is obtained by performing gain processing on the first echo signal based on a programmable gain circuit, resulting in a first-gain echo signal of the current region. Point cloud analysis is then performed on the first-gain echo signal of the current region to obtain a point cloud analysis dataset. Based on the point cloud analysis dataset, obstacle clusters in the current region are identified, and the expected planned path is optimized based on these obstacle clusters to obtain an optimized obstacle avoidance path. By extracting the optical features of the surface material of the current region and matching the optimal band interval based on a pre-built feature fingerprint database, adaptive optimization of the transmission wavelength is achieved, thereby improving the echo signal quality from the source. By adjusting the programmable gain of the echo signal at the receiving end according to the matched band, the stability and reliability of the point cloud data are ensured. Based on the obtained high-quality point cloud, obstacle recognition accuracy is significantly improved, thus making the environmental model on which path planning relies more accurate. Attached Figure Description

[0060] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0061] Figure 1 A flowchart of an autonomous obstacle avoidance method for unmanned aerial vehicles based on lidar provided in an embodiment of the present invention;

[0062] Figure 2 A schematic diagram of a drone autonomous obstacle avoidance device based on lidar provided in an embodiment of the present invention;

[0063] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0064] Icons: 310 - Initial echo signal determination module; 320 - First cluster echo signal determination module; 330 - Surface material optical feature determination module; 340 - Matching band interval determination module; 350 - Radar signal transmission module; 360 - Radar signal reception module; 370 - Point cloud analysis data processing module; 380 - Optimized obstacle avoidance path processing module; 400 - Memory; 401 - Processor; 402 - Bus; 403 - Communication interface. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0066] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0067] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0068] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0069] Furthermore, terms such as "horizontal," "vertical," and "sag" do not imply that components must be absolutely horizontal or suspended, but rather that they can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal relative to "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0070] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0071] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0072] Example 1

[0073] This invention provides an autonomous obstacle avoidance method for unmanned aerial vehicles (UAVs) based on lidar, applied to the control system of the UAV. The transmitting end of the UAV's lidar includes a band selection circuit; the receiving end of the lidar includes a programmable gain circuit; see [link to related documentation]. Figure 1 The flowchart shown in this embodiment of the invention provides an autonomous obstacle avoidance method for unmanned aerial vehicles (UAVs) based on lidar. The method includes:

[0074] Step S102: Perform a global scan of the area where the UAV is currently located based on the lidar to obtain the global initial echo signal.

[0075] Specifically, in complex environments such as mountainous spoil heaps, there are various natural or man-made obstacles, including asphalt roads, exposed rocks, metal fences, and vegetation, with significantly different optical properties of their surface materials. To comprehensively perceive such heterogeneous environments, this embodiment of the invention controls a lidar mounted on a drone to perform a global scan of the current area. The lidar emitter emits laser pulses into the scanned area and receives signals reflected from different surfaces, thereby acquiring raw point cloud data frames containing information such as spatial location, echo intensity, and echo count—that is, the global initial echo signal. By actively emitting detection signals and receiving their reflections, it is possible to rapidly acquire three-dimensional geometric and reflection characteristics information of a large area non-contactly, overcoming the perception limitations of traditional visible light sensors in environments with changing lighting and missing textures.

[0076] Step S104: Cluster the global initial echo signal and determine the first cluster echo signal based on the signal proportion of multiple cluster echo signals in the clustering results.

[0077] Specifically, since the global initial echo signal contains mixed information from different surface materials, it needs to be decoupled and analyzed. This embodiment of the invention employs algorithms such as Euclidean distance clustering or density clustering (DBSCAN) to segment the global initial echo signal into multiple clustered echo signals based on the spatial proximity and echo feature similarity of the point cloud data. Each cluster represents a physically continuous region in the environment with similar material properties; for example, one cluster might correspond to an asphalt road surface, while another might correspond to a metal guardrail. Subsequently, the percentage of echo points contained in each cluster out of the total number of points is calculated, i.e., the signal proportion. By comparing the signal proportions of all clusters, the cluster with the largest proportion is determined as the first cluster echo signal. This quantitative screening process ensures that the dominant (largest or most significant) surface material type in the current field of view can be automatically and objectively identified, allowing subsequent optimization strategies to focus on environmental factors that have the greatest impact on overall perception performance, avoiding over-adaptation to minor details in complex scenes.

[0078] Furthermore, in some preferred embodiments of the present invention, the step of clustering the global initial echo signal and determining the first clustered echo signal based on the signal proportion of multiple clustered echo signals in the clustering results includes: clustering the global initial echo signal to obtain multiple clustered echo signals; calculating the percentage of echo points contained in each clustered echo signal to the total number of echo points; and determining the clustered echo signal with the largest percentage as the first clustered echo signal.

[0079] Specifically, clustering algorithms group laser points with similar spatial distribution and reflection characteristics together based on preset distance thresholds or density parameters, forming several independent data clusters. The process of calculating signal proportion is essentially a statistical analysis of the size of each data cluster; the largest cluster represents the object surface with the widest coverage within the drone's current "field of view." For example, when a drone flies over a spoil heap, if the area of ​​exposed rock is much larger than scattered metal structures and vegetation, the rock cluster will have the highest signal proportion and will be selected as the first cluster echo signal. This method replaces the limitations of manually preset environmental models with a data-driven approach, enabling the drone to capture dynamic changes in the environment in real time and accurately, providing a reliable basis for adaptive adjustment of perception parameters. By focusing on dominant materials, the most significant improvement in perception performance can be achieved with minimal computational overhead, improving the operational efficiency and environmental adaptability of the entire obstacle avoidance system.

[0080] Step S106: Extract the real-time feature vector of the first cluster echo signal as the optical feature of the surface material of the current region.

[0081] Specifically, after identifying the first cluster echo signal corresponding to the dominant material, it is necessary to extract a feature set that can quantitatively characterize the optical reflection behavior of the material. In this embodiment of the invention, a set of multidimensional feature values ​​is calculated and extracted from all data points contained in the first cluster echo signal, including but not limited to: average echo intensity (reflecting the overall reflectivity of the material), echo intensity distribution (such as variance, histogram, reflecting reflection uniformity), echo pulse width (related to material surface roughness and penetrability), and multiple echo characteristics (such as the proportion of pulses generating multiple echoes, reflecting material transmissivity or internal structure). These calculated feature values ​​are combined in a fixed order to generate a real-time feature vector representing the characteristics of the dominant material in the current region. This feature vector is a digital fingerprint of the interaction between the material surface and a laser of a specific wavelength, transforming complex physical reflection phenomena into a numerical sequence that can be used for computer matching and decision-making. This achieves a key transformation from "physical echo" to "digital feature," providing accurate input for subsequent intelligent parameter optimization based on a knowledge base.

[0082] Step S108: Based on the optical characteristics of the surface material in the current region, search for the matching band interval of the current region in the preset surface material optical characteristic fingerprint database.

[0083] Specifically, the surface material optical feature fingerprint database contains a rich set of material-band response relationships. Surface material optical features are used as query criteria and input into this fingerprint database for retrieval. The database stores a large number of standard feature vectors obtained from scanning known materials under different laser bands, along with their corresponding optimal detection band intervals. By running similarity calculations (such as cosine similarity) or pattern matching algorithms, one or more standard feature vectors that are closest to the current real-time feature vector are found in the database. Then, the matching band intervals associated with these standard feature vectors, targeting specific optimization objectives (such as highest signal-to-noise ratio or most prominent features), are retrieved. This enables UAVs to quickly access best practices from historical experimental data based on real-time perception, thereby making scientific band selection decisions and effectively overcoming the drawbacks of the traditional fixed-band mode's "one-size-fits-all" approach.

[0084] Furthermore, in some preferred embodiments of the present invention, the method further includes: acquiring feature vector samples of different surface material samples under different test bands; wherein the feature vector samples include echo intensity distribution, echo signal attenuation ratio, and multiple echo characteristics; determining the matching band interval corresponding to each surface material sample based on the band response sensitivity samples and preset optimization targets; wherein the optimization targets include at least one of the following: signal-to-noise ratio target, feature enhancement target, and material discrimination target; establishing a mapping relationship between different surface material samples, optimization targets, and matching band intervals to construct a surface material optical feature fingerprint database.

[0085] Specifically, the construction of a surface material optical feature fingerprint database is a prerequisite and core knowledge foundation for the adaptive sensing method of this paper. The construction of this database is an offline, systematic experimental and analytical process. First, in a controlled experimental environment, a tunable lidar system is used to scan and test a material sample database covering a wide range of material types (such as asphalt, concrete, various metals, glass, wood, plastics, different vegetation, water bodies, etc.). During testing, the lidar's emission wavelength is switched between several typical bands (e.g., 905nm, 1064nm, 1550nm, etc.) to cover different detection windows. For each "material-band" combination, its echo signal is acquired, and following a similar but more standardized process to step S106, a set of multi-dimensional feature vector samples is calculated and generated, specifically including: echo intensity distribution (describing the statistical characteristics of the reflected signal), echo signal attenuation ratio (quantifying the attenuation rate of signal energy), and multiple echo characteristics (characterizing the material's ability to penetrate laser light and its subsurface reflection capability).

[0086] Subsequently, for each material sample, the variation of its eigenvector with the test band is analyzed, i.e., "band response sensitivity analysis" is performed. For example, the analysis reveals that metals have extremely high echo intensity but attenuate rapidly in a certain band, while vegetation exhibits significant echo characteristics multiple times in another band. Through this analysis, response sensitivity samples for each material in different bands can be obtained. Next, based on different task optimization objectives, "matching band intervals" are selected for each material. For example, if the objective is "highest signal-to-noise ratio," the band with the highest mean echo intensity and lowest variance is selected for the material; if the objective is "feature enhancement" (such as better distinguishing wires), the band that makes the target profile features sharpest is selected; if the objective is "maximum material differentiation" (such as distinguishing asphalt roads from dirt roads), the band that maximizes the difference in echo characteristics between the two materials is selected.

[0087] Finally, a one-to-one mapping relationship is established between the "material sample identifier," the "optimization target," and the calculated "matching band interval," and these are stored together with their corresponding standard feature vector samples, thus constructing a structured surface material optical feature fingerprint library. This library encapsulates prior knowledge of laser-matter interactions, enabling UAVs to obtain optimal detection parameters with near-laboratory precision through efficient database queries during online operation, without the need for complex physical calculations.

[0088] Furthermore, in some preferred embodiments of the present invention, the step of searching for a matching band interval of the current region in a preset surface material optical feature fingerprint database based on the surface material optical features of the current region includes: calculating the similarity between the surface material optical features of the current region and the feature vectors of each surface material sample pre-stored in the surface material optical feature fingerprint database; determining a matching feature vector based on the similarity calculation result; and searching for the corresponding matching band interval from the mapping relationship based on the matching feature vector and the currently selected optimization objective.

[0089] Specifically, firstly, the UAV uses metrics such as cosine similarity to calculate in parallel the similarity score between the real-time acquired optical features of the surface material and each pre-stored standard feature vector sample in the fingerprint database. The goal is to find the known point in the feature space that is closest to the current feature point. After sorting all similarity scores, the standard feature vector with the highest score is determined as the "matching feature vector," meaning that the dominant material in the current environment is most similar to the known material in the fingerprint database in terms of optical properties.

[0090] After determining the matching feature vector, an optimization target is selected based on the current flight mission or strategy. The optimization target is a high-level decision; for example, during nighttime or long-distance inspections, a "signal-to-noise ratio target" might be prioritized to maximize detection range; when precise identification of power line contours is required, a "feature enhancement target" might be chosen; and when traversing mixed-material areas, a "material discrimination target" might be selected to avoid misjudgment. Once the optimization target is selected, the corresponding band interval is directly retrieved based on a pre-established mapping relationship of "matching feature vector - optimization target - matching band interval." For example, if the matching feature vector corresponds to "asphalt" and the current optimization target is "signal-to-noise ratio," the query result might be "1550nm band." This design decouples and integrates material identification with mission objectives, making band selection both precise and flexible, achieving intelligent linkage between perception and decision-making.

[0091] Step S110: Adjust the band selection circuit so that the lidar transmits the first test signal in the matching band interval.

[0092] Specifically, the band selection circuit is an electronic unit controlled by system digital commands. When it receives a matching band range command (such as "switch to 1550nm") from the control system, it immediately executes the corresponding hardware operation to change the output wavelength of the laser.

[0093] There are two main ways to implement a band selection circuit: one is based on a tunable laser diode design, which adjusts the drive current or the operating temperature of the laser to change its emission wavelength within a continuous range until it stabilizes in the target band; the other is to use a multi-laser diode gating scheme, which integrates multiple lasers that are fixed to work at different wavelengths (such as 905nm, 1064nm, and 1550nm). The band selection circuit turns on the drive circuit of the target laser according to the instruction, while turning off the other lasers.

[0094] After the switch is complete, the lidar emits its first test signal (a series of laser pulses) into the current area at a new, optimized wavelength. From this moment on, the characteristics of the "light source" used by the UAV for environmental detection have been customized according to the reflectivity of the main materials in the environment. This adaptive capability at the signal source is a fundamental measure to improve the signal-to-noise ratio and feature recognition of the entire sensing link. It makes the emitted detection signal itself more "suitable" for the current environment, laying the foundation for subsequent reception of high-quality echoes.

[0095] Furthermore, in some preferred embodiments of the present invention, the method further includes: if the data change rate of the signal proportion is greater than a preset threshold, then redetermine the first clustered echo signal; based on the optical characteristics of the surface material corresponding to the redetermined first clustered echo signal, re-find the corresponding matching band interval; and adjust the band selection circuit according to the redetermined matching band interval.

[0096] Specifically, the drone's flight environment is dynamic; for example, when moving from an open road into a forest, the dominant materials will change. To ensure that the adaptive system can continuously track environmental changes, this embodiment of the invention provides a dynamic reconfiguration mechanism. This involves real-time monitoring of the "data change rate" of the proportions of multiple clustered echo signals. This change rate is defined as the degree of difference (such as mean square error or maximum change) between the proportion of each clustered signal in the current scan frame and the corresponding proportion in the previous frame.

[0097] A threshold for the rate of change is preset. After each frame of data is processed, the current rate of change is calculated. If the rate of change is less than or equal to the threshold, it indicates that the environmental composition is relatively stable, and no reconfiguration is required; the current band settings are continued. If the rate of change is greater than the threshold, it indicates that the material composition of the environment is undergoing drastic changes (e.g., a drone is flying away from a rocky area and into a dense forest), and the currently used first cluster and its corresponding band configuration may have become invalid.

[0098] Once the reconfiguration condition is triggered, a complete rapid reconfiguration process is immediately initiated: First, based on the global initial echo signal acquired in the latest frame, cluster analysis is re-executed to obtain multiple new clusters representing the latest environment. Then, the signal proportion of each new cluster is recalculated, and the echo signal of the first cluster is redefined based on the new proportions. Next, new real-time feature vectors are extracted from the new first cluster and used to re-query the surface material optical feature fingerprint database to obtain a new matching band interval. Finally, the band selection circuit is driven to switch to this new band interval.

[0099] This closed-loop feedback mechanism greatly enhances the system's robustness and environmental adaptability. It ensures that when the UAV traverses heterogeneous environments, its perception system's "eyes" (LiDAR) can continuously and automatically adjust to the most suitable "observation mode" for the current scene, just like the human eye adapts to light and dark. This avoids a precipitous drop in perception performance or temporary blind spots that may occur due to environmental changes, and ensures continuous and reliable obstacle avoidance throughout the entire flight.

[0100] Step S112: Receive the first echo signal of the current area corresponding to the first test signal, and perform gain processing on the first echo signal based on the programmable gain circuit to obtain the first gain echo signal of the current area.

[0101] Specifically, after the lidar receiver detects the first echo signal reflected from the environment, the signal strength is affected by various factors: transmission power, target distance, material reflectivity, and the characteristics of the newly switched band. The dynamic range of the original echo signal can be very large, and direct digitization can easily lead to weak signals being overwhelmed by noise or strong signals being saturated and distorted. Therefore, adaptive amplitude conditioning is required before the signal enters the core processing unit. In this embodiment of the invention, a programmable gain circuit connected to the lidar receiver accomplishes this task. This circuit dynamically adjusts its amplification factor for the echo signal based on the currently used matching band range (band information implies the expected signal strength range) and the real-time characteristics of the signal itself. After conditioning by the programmable gain circuit, the amplitude of the output signal is stably controlled within the optimal quantization range of the subsequent analog-to-digital converter (ADC). The signal obtained at this point is called the first gain echo signal. This circuit compensates for signal strength fluctuations caused by changes in detection parameters and the environment, essentially providing an "automatic volume adjustment" function for the sensing system. This ensures that the signal input to the point cloud analysis module has consistently high quality, clearing the way for generating accurate and reliable 3D point clouds.

[0102] Furthermore, in some preferred embodiments of the present invention, the programmable gain circuit includes: a programmable bandpass filter; the parameters of the programmable bandpass filter correspond one-to-one with the matched band interval; the step of performing gain processing on the first echo signal based on the programmable gain circuit includes: determining the target parameters of the programmable bandpass filter based on the matched band interval; configuring the programmable bandpass filter based on the target parameters so that the programmable bandpass filter performs filtering processing on the first echo signal; wherein, the target parameters include the center frequency and bandwidth corresponding to the matched band interval.

[0103] Specifically, the first step in programmable gain processing is to "purify" the signal in the frequency domain. One of the core components of a programmable gain circuit is a programmable bandpass filter. The wavelength (λ) and frequency (f) of a laser are constantly correlated through the speed of light (c) (f = c / λ). Therefore, the selected matching band interval in this embodiment of the invention directly corresponds to a specific optical frequency range, and thus corresponds to the center frequency and bandwidth of the echo signal in the electronic domain after photoelectric conversion.

[0104] The corresponding target parameters are calculated based on the matched band range: center frequency (corresponding to the center wavelength of the band) and bandwidth (slightly wider than the spectral width of the emitted laser to accommodate the signal and allow for minor frequency drift). These target parameters are then sent to a programmable bandpass filter as digital commands. The filter then reconfigures the center frequency and bandwidth of its passband to the specified values. When the first echo signal (mixed with various noises) passes through the filter, only the signal components with frequencies falling within the passband (i.e., the effective echo signal with the same wavelength as the emitted laser) can pass without attenuation, while noise outside the passband (such as ambient background light noise, other electronic interference, etc.) is greatly suppressed.

[0105] Through this selective filtering in the frequency domain, the signal-to-noise ratio is improved for the first and most fundamental time before amplification. This ensures that the main signal to be amplified is a "clean" target echo, rather than noise, thus guaranteeing the quality of the point cloud data from the source.

[0106] Furthermore, in some preferred embodiments of the present invention, the programmable gain circuit further includes a variable gain amplifier; the step of performing gain processing on the first echo signal based on the programmable gain circuit further includes: amplifying the filtered first echo signal using the variable gain amplifier; wherein, the target gain factor of the variable gain amplifier is determined based on the following steps: performing first-order gain calculation based on the signal amplitude, signal-to-noise ratio, and saturation of the first echo signal, and outputting a first-order gain factor; applying the first-order gain factor to the variable gain amplifier for testing to obtain a second echo signal; determining whether the signal amplitude of the second echo signal meets a preset target peak window; if it does, determining the first-order gain factor as the target gain factor of the variable gain amplifier; if it does not, iteratively adjusting the first-order gain factor based on the error between the signal amplitude of the second echo signal and the target peak window until the output echo signal meets the target peak window, and determining the finally adjusted gain factor as the target gain factor.

[0107] Specifically, after filtering and purification, the signal amplitude needs to be precisely amplified to a level suitable for digital processing, which is accomplished by a variable gain amplifier (VGA). Its amplification factor (gain factor) is not fixed, but dynamically determined through a fast, adaptive closed-loop control algorithm.

[0108] The target gain factor of the variable gain amplifier is determined based on the following steps: First, the filtered first echo signal is quickly analyzed to evaluate its signal amplitude (peak voltage), signal-to-noise ratio (SNR), and whether it is close to saturation. Based on these initial evaluation values, an initial amplification factor, i.e., the first-order gain factor, is estimated using a preset first-order gain calculation model. Next, this gain factor is temporarily applied to the VGA, and the signal after the initial amplification (referred to as the second echo signal) is sampled.

[0109] Then, the measured peak value of the second echo signal is compared with a predefined "target peak window". This target peak window represents the voltage range of the optimal operating region of the back-end ADC. If the peak value of the second echo signal falls exactly within this window, it indicates that the first-order gain calculation is very accurate, and this first-order gain factor is locked as the final target gain factor.

[0110] If the peak value of the second echo signal falls outside the window (too small or too large), it indicates that the initial estimate was not accurate enough. At this point, the fine-tuning iteration phase begins. It calculates the error between the current peak value and the center value of the target window, and based on this error, uses algorithms such as proportional-integral-derivative (PID) control to calculate the adjustment amount for the current gain factor, thus obtaining a corrected second-order gain factor. The new gain factor is applied again, and a new echo signal is sampled and its peak value is checked. This "application-measurement-comparison-adjustment" cycle iterates rapidly several times, usually within a very short time (within a few pulse cycles) to make the peak value of the output signal stably fall within the target peak window. The gain factor used at this point is finally determined as the target gain factor.

[0111] Ultimately, the programmable bandpass filter and the VGA operating at the optimal target gain factor work together to sequentially filter, denoise, and precisely amplify the first echo signal, outputting a high-quality, amplitude-normalized first-gain echo signal. This closed-loop gain control technology enables the UAV to cope with a wide variety of echo signal intensities, always adjusting it to the optimal processing state, fundamentally solving the problem of point cloud data quality degradation caused by excessively large signal dynamic range.

[0112] Step S114: Perform point cloud analysis on the first gain echo signal of the current area to obtain the point cloud analysis dataset of the current area.

[0113] Specifically, the first gain echo signal is a high-quality standardized signal after band optimization and gain conditioning. In this embodiment of the invention, the core point cloud analysis algorithm is executed by the signal processing unit built into the lidar. The analysis process mainly includes: First, accurately measuring the time difference (Time of Flight, ToF) from the emission to the reception of each laser pulse, and combining it with the speed of light constant, calculating the precise distance between the lidar and each reflection point in the scene. Then, combining the precise angle (azimuth and pitch) of the internal scanning mirror of the lidar when emitting the pulse, and the real-time attitude (roll, pitch, yaw) and position (longitude, latitude, altitude) of the UAV in space obtained by the inertial measurement unit (IMU) and the global navigation satellite system (GNSS), through a series of complex coordinate transformations (from the lidar coordinate system to the aircraft coordinate system, and then to the global geographic coordinate system), the polar coordinates (distance, angle) of each point are converted into three-dimensional rectangular coordinates (X, Y, Z) in the geodetic coordinate system.

[0114] In addition to spatial coordinates, the analysis process also adds information such as echo intensity (corresponding to the amplitude of the first gain echo signal), precise timestamp, and the laser beam from which each point originates. The collection of all these points constitutes the point cloud analysis dataset for the current region. Due to the high quality of the input signal (high signal-to-noise ratio, appropriate amplitude), the analyzed point cloud data not only boasts high accuracy in three-dimensional geometric position, but also allows the intensity information of each point to more realistically and stably reflect the reflective characteristics of the material surface, forming a reliable three-dimensional digital model rich in detail, low in noise, and suitable for high-precision environmental perception and understanding.

[0115] Step S116: Determine the obstacle clusters in the current area based on the point cloud parsing dataset, and optimize the expected planned path based on the obstacle clusters to obtain the optimized obstacle avoidance path.

[0116] Specifically, the process begins with semantic understanding of the discrete 3D point cloud, i.e., obstacle clustering. Then, algorithms such as Euclidean clustering or DBSCAN are used to group points belonging to the same physical entity in space, forming "obstacle clusters." For example, all points belonging to the same tree are grouped into one cluster, and points belonging to the same building wall are grouped into another. Each obstacle cluster is treated as an independent object that needs to be avoided.

[0117] Next, the threat posed by these obstacles to the current flight mission is assessed. Spatial conflict detection is performed on the 3D contours of each obstacle cluster (typically represented by a simplified bounding box or convex hull) and the UAV's predetermined planned path (pre-defined by the mission planning module or generated in real-time). This is not merely a simple distance determination, but rather calculating the probability (conflict probability) of the UAV colliding with each obstacle cluster while flying along the planned path over a future period. Based on this probability assessment, all obstacle clusters with conflict probabilities exceeding a certain safety threshold (preset conflict probability) are marked as "obstacle avoidance points," i.e., high-risk areas that must be avoided.

[0118] Finally, starting from the current UAV position and ending at the mission objective point, while adhering to the strict safety constraint of "avoiding all obstacle avoidance points," and considering optimization objectives such as path length, smoothness (satisfying UAV dynamic constraints), and approximating the original planned path as closely as possible, a new flight trajectory—the optimized obstacle avoidance path—is calculated in real time. This path, while ensuring absolute safety, maximizes mission continuity and efficiency. This path is ultimately converted into control commands and sent to the UAV's flight control system for execution, thus achieving safe and intelligent autonomous obstacle avoidance within the entire closed-loop process of perception, decision-making, and control.

[0119] Furthermore, in some preferred embodiments of the present invention, the step of determining obstacle clusters in the current area based on the point cloud parsing dataset and optimizing the expected planned path based on the obstacle clusters to obtain an optimized obstacle avoidance path includes: performing obstacle clustering based on the point cloud parsing dataset to output at least one obstacle cluster; calculating the conflict probability between each obstacle cluster and the expected planned path of the UAV, and marking obstacle clusters with conflict probabilities greater than a preset conflict probability as obstacle avoidance points; and replanning the expected planned path with the constraint of avoiding all obstacle avoidance points to generate an optimized obstacle avoidance path.

[0120] Specifically, obstacle clustering is a key step in elevating low-level perception data (points) to high-level environmental cognition (objects). It enables path planning algorithms to handle meaningful object entities, rather than millions of unordered points. Conflict probability calculation is a risk assessment process that combines spatial geometric relationships (distance) with system uncertainties (sensor errors, control errors), quantifying risk with a probability value. This is more scientific and precise than a simple binary "safe / dangerous" judgment, allowing drones to differentiate their responses to obstacles of different risk levels. Setting a preset conflict probability threshold essentially defines the system's "risk tolerance," and all obstacles with risks exceeding this tolerance are considered for obstacle avoidance. The final path replanning involves solving a trajectory problem that satisfies multi-objective optimization (shortest, smoothest, closest to the original plan) within a constrained space with obstacle avoidance points as no-fly zones. Commonly used algorithms include fast random search trees, dynamic window methods (DWA), or optimization-based methods (such as quadratic programming). This ensures that obstacle avoidance decisions are both reactive (responding to real-time threats) and forward-looking (planning the entire safe path), thereby achieving robust autonomous navigation.

[0121] Furthermore, in some preferred embodiments of the present invention, the step of calculating the conflict probability between each obstacle cluster and the expected planned path of the UAV, and marking obstacle clusters with conflict probabilities greater than preset conflict probabilities as obstacle avoidance points, includes: calculating the shortest spatial distance between the outer contour of the obstacle cluster and the expected planned path; calculating the collision probability based on the shortest spatial distance, the safe radius of the UAV, the relative speed of the UAV, the positioning error of the UAV, and the control error of the UAV; and marking obstacle clusters with collision probabilities greater than preset conflict probabilities as obstacle avoidance points.

[0122] Specifically, first, the shortest Euclidean distance between the outer contour of the obstacle cluster (such as an axial bounding box or convex hull) and the line segment of the expected planned path is calculated.

[0123] However, distance alone is insufficient for reliable judgment due to various uncertainties in actual flight. Therefore, a probabilistic collision model is needed. This model treats the future position of the drone as a probability distribution (e.g., a Gaussian distribution), whose uncertainty consists of several parts: the drone's physical size, simplified as a sphere by introducing a "safety radius"; the system's state estimation error, mainly the positioning error after GPS / IMU fusion; and the control system's execution error, i.e., the deviation between the drone's actual flight trajectory and the commanded trajectory. Furthermore, the drone's relative speed determines the time it takes to "meet" an obstacle; the faster the speed, the shorter the time available for error correction, and the higher the risk.

[0124] Based on the shortest distance and all the aforementioned uncertainty parameters, the model can calculate the probability that the sphere representing the drone's position will spatially overlap with an obstacle entity within a future time window. This probability value is the "collision probability." For example, even if the current distance is acceptable, the collision probability may still be high if the positioning error is large or the drone is moving laterally at high speed.

[0125] Finally, the calculated collision probability of each obstacle cluster is compared to a pre-set, acceptable "preset collision probability" threshold. Only obstacle clusters with collision probabilities exceeding this threshold are officially marked as "obstacle avoidance points." This method allows the drone to intelligently balance safety margins and flight efficiency. For obstacles that are far away or have low uncertainty (low collision probability), the drone can maintain its original path or make slight adjustments to maintain flight smoothness; it only resolutely avoids obstacles that truly pose a high risk (high collision probability). This makes the entire obstacle avoidance behavior both safe and efficient, avoiding "startle" overreactions, and is particularly suitable for smooth and reliable autonomous flight in complex and dynamic operating environments.

[0126] This invention provides an autonomous obstacle avoidance method for unmanned aerial vehicles (UAVs) based on lidar, applied to the control system of a UAV. The transmitting end of the UAV's lidar includes a band selection circuit; the receiving end of the lidar includes a programmable gain circuit. The method includes: performing a global scan of the area currently occupied by the UAV based on the lidar to obtain a global initial echo signal; clustering the global initial echo signal and determining a first clustered echo signal based on the signal proportion of multiple clustered echo signals in the clustering results; extracting the real-time feature vector of the first clustered echo signal as the surface material optical feature of the current area; searching for a matching band interval of the current area in a preset surface material optical feature fingerprint database based on the surface material optical feature of the current area; adjusting the band selection circuit so that the lidar emits a first test signal in the matching band interval; and receiving the current band interval corresponding to the first test signal. The first echo signal of the current region is obtained by performing gain processing on the first echo signal based on a programmable gain circuit. Point cloud analysis is then performed on the first gain echo signal of the current region to obtain a point cloud analysis dataset. Based on the point cloud analysis dataset, obstacle clusters in the current region are determined, and the expected planned path is optimized based on these obstacle clusters to obtain an optimized obstacle avoidance path. By extracting the optical features of the surface material of the current region and matching the optimal band interval based on a pre-built feature fingerprint database, adaptive optimization of the transmission wavelength is achieved, thereby improving the echo signal quality from the source. By adjusting the programmable gain of the echo signal at the receiving end according to the matched band, the stability and reliability of the point cloud data are ensured. Based on the obtained high-quality point cloud, obstacle recognition accuracy is significantly improved, thus making the environmental model on which path planning relies more accurate.

[0127] Example 2

[0128] Based on the above embodiments, this invention provides an autonomous obstacle avoidance device for unmanned aerial vehicles (UAVs) based on lidar, applied to the control system of a UAV. The transmitting end of the UAV's lidar includes a band selection circuit; the receiving end of the lidar includes a programmable gain circuit. See also Figure 2 The diagram shown is a structural schematic of an autonomous obstacle avoidance device for unmanned aerial vehicles (UAVs) based on lidar, according to an embodiment of the present invention. The device includes:

[0129] The initial echo signal determination module 310 is used to perform a global scan of the area where the UAV is currently located based on the lidar to obtain the global initial echo signal.

[0130] The first cluster echo signal determination module 320 is used to cluster the global initial echo signal and determine the first cluster echo signal based on the signal ratio of multiple cluster echo signals in the clustering results.

[0131] The surface material optical feature determination module 330 is used to extract the real-time feature vector of the first cluster echo signal as the surface material optical feature of the current region.

[0132] The matching band interval determination module 340 is used to search for the matching band interval of the current region in a preset surface material optical feature fingerprint database based on the surface material optical features of the current region.

[0133] The radar signal transmitting module 350 is used to adjust the band selection circuit so that the lidar transmits the first test signal in the matching band interval.

[0134] The radar signal receiving module 360 ​​is used to receive the first echo signal of the current area corresponding to the first test signal, and to perform gain processing on the first echo signal based on the programmable gain circuit to obtain the first gain echo signal of the current area.

[0135] The point cloud analysis data processing module 370 is used to perform point cloud analysis on the first gain echo signal of the current area to obtain the point cloud analysis dataset of the current area.

[0136] The obstacle avoidance path processing module 380 is used to determine the obstacle clusters in the current area based on the point cloud parsing dataset, and to optimize the expected planned path based on the obstacle clusters to obtain the optimized obstacle avoidance path.

[0137] Furthermore, in some preferred embodiments of the present invention, the first clustered echo signal determination module 320 is used to cluster the global initial echo signal to obtain multiple clustered echo signals; calculate the percentage of echo points contained in each clustered echo signal to the total number of echo points; and determine the clustered echo signal with the largest percentage as the first clustered echo signal.

[0138] Furthermore, in some preferred embodiments of the present invention, the apparatus further includes: a surface material optical feature fingerprint database construction module, used to acquire feature vector samples of different surface material samples under different test bands; wherein, the feature vector samples include echo intensity distribution, echo signal attenuation ratio, and multiple echo characteristics; based on the band response sensitivity samples and preset optimization targets, determining the matching band intervals corresponding to each surface material sample; wherein, the optimization targets include at least one of the following: signal-to-noise ratio target, feature enhancement target, and material discrimination target; establishing a mapping relationship between different surface material samples, optimization targets, and matching band intervals to construct a surface material optical feature fingerprint database.

[0139] Furthermore, in some preferred embodiments of the present invention, the matching band interval determination module 340 is used to calculate the similarity between the optical features of the surface material in the current region and the feature vectors of each surface material sample pre-stored in the surface material optical feature fingerprint database; determine the matching feature vector based on the similarity calculation result; and search for the corresponding matching band interval from the mapping relationship based on the matching feature vector and the currently selected optimization target.

[0140] Furthermore, in some preferred embodiments of the present invention, the apparatus further includes: a matching band interval update module, configured to: redetermine the first clustered echo signal if the data change rate of the signal proportion is greater than a preset threshold; rediscover the corresponding matching band interval based on the optical characteristics of the surface material corresponding to the redetermined first clustered echo signal; and adjust the band selection circuit according to the redetermined matching band interval.

[0141] Furthermore, in some preferred embodiments of the present invention, the programmable gain circuit includes: a programmable bandpass filter; the parameters of the programmable bandpass filter correspond one-to-one with the matching band interval; a radar signal receiving module 360, used to determine the target parameters of the programmable bandpass filter based on the matching band interval; and to configure the programmable bandpass filter based on the target parameters so that the programmable bandpass filter performs filtering processing on the first echo signal; wherein, the target parameters include the center frequency and bandwidth corresponding to the matching band interval.

[0142] Furthermore, in some preferred embodiments of the present invention, the programmable gain circuit further includes a variable gain amplifier; the radar signal receiving module 360 ​​is used to amplify the filtered first echo signal using the variable gain amplifier; wherein, the target gain factor of the variable gain amplifier is determined based on the following steps: performing a first-order gain calculation based on the signal amplitude, signal-to-noise ratio, and saturation of the first echo signal, and outputting a first-order gain factor; applying the first-order gain factor to the variable gain amplifier for testing to obtain a second echo signal; determining whether the signal amplitude of the second echo signal meets a preset target peak window; if it does, determining the first-order gain factor as the target gain factor of the variable gain amplifier; if it does not meet, iteratively adjusting the first-order gain factor based on the error between the signal amplitude of the second echo signal and the target peak window until the output echo signal meets the target peak window, and determining the finally adjusted gain factor as the target gain factor.

[0143] Furthermore, in some preferred embodiments of the present invention, the obstacle avoidance path processing module 380 is used to perform obstacle clustering based on the point cloud parsing dataset and output at least one obstacle cluster; calculate the conflict probability between each obstacle cluster and the expected planned path of the UAV, mark the obstacle cluster with a conflict probability greater than a preset conflict probability as an obstacle avoidance point; and replan the expected planned path to generate an optimized obstacle avoidance path with the constraint of avoiding all obstacle avoidance points.

[0144] Furthermore, in some preferred embodiments of the present invention, the obstacle avoidance path processing module 380 is used to calculate the shortest spatial distance between the outer contour of the obstacle cluster and the expected planned path; calculate the collision probability based on the shortest spatial distance, the safe radius of the UAV, the relative speed of the UAV, the positioning error of the UAV, and the control error of the UAV; and mark the obstacle cluster with a collision probability greater than the preset conflict probability as obstacle avoidance points.

[0145] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the lidar-based UAV autonomous obstacle avoidance device described above can be referred to the corresponding process in the aforementioned embodiments of the lidar-based UAV autonomous obstacle avoidance method, and will not be repeated here.

[0146] Example 3

[0147] This invention also provides an electronic device for running a lidar-based autonomous obstacle avoidance method for unmanned aerial vehicles; see also Figure 3 The diagram shown is a structural schematic of an electronic device provided by an embodiment of the present invention. The electronic device includes a memory 400 and a processor 401. The memory 400 is used to store one or more computer instructions, which are executed by the processor 401 to realize the above-mentioned autonomous obstacle avoidance method for UAVs based on lidar.

[0148] Furthermore, Figure 3 The electronic device shown also includes a bus 402 and a communication interface 403. The processor 401, the communication interface 403 and the memory 400 are connected via the bus 402.

[0149] The memory 400 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 403 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 402 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0150] Processor 401 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 401 or by instructions in software form. Processor 401 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a readily available storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 400, and processor 401 reads information from memory 400 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0151] This invention also provides a computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are called and executed by a processor, they cause the processor to implement the aforementioned autonomous obstacle avoidance method for UAVs based on lidar. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0152] The computer program product of the UAV autonomous obstacle avoidance method, device and electronic device based on lidar provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0153] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and / or device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0154] Furthermore, in the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.

[0155] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0156] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for autonomous obstacle avoidance of unmanned aerial vehicles (UAVs) based on lidar, characterized in that, A control system for an unmanned aerial vehicle (UAV), wherein the transmitting end of the UAV's lidar includes a band selection circuit; the receiving end of the lidar includes a programmable gain circuit; the method includes: The laser radar is used to perform a global scan of the area where the UAV is currently located to obtain the global initial echo signal. The global initial echo signal is clustered, and the first cluster echo signal is determined based on the signal proportion of multiple cluster echo signals in the clustering results. Extract the real-time feature vector of the first clustered echo signal as the surface material optical feature of the current region; Based on the optical characteristics of the surface material in the current region, search for the matching band interval of the current region in the preset surface material optical characteristic fingerprint database; Adjust the band selection circuit so that the lidar transmits a first test signal in the matched band interval; Receive the first echo signal of the current area corresponding to the first test signal, and perform gain processing on the first echo signal based on the programmable gain circuit to obtain the first gain echo signal of the current area. Perform point cloud analysis on the first gain echo signal of the current region to obtain the point cloud analysis dataset of the current region; Based on the point cloud parsing dataset, the obstacle clusters in the current area are determined, and the expected planned path is optimized based on the obstacle clusters to obtain an optimized obstacle avoidance path.

2. The autonomous obstacle avoidance method for unmanned aerial vehicles based on lidar according to claim 1, characterized in that, The step of clustering the global initial echo signal and determining the first clustered echo signal based on the signal proportions of multiple clustered echo signals in the clustering results includes: The global initial echo signal is clustered to obtain multiple clustered echo signals; Calculate the percentage of echo points contained in each clustered echo signal relative to the total number of echo points; The cluster echo signal with the largest percentage is determined as the first cluster echo signal.

3. The autonomous obstacle avoidance method for unmanned aerial vehicles based on lidar according to claim 1, characterized in that, The method further includes: Obtain feature vector samples of different surface materials under different test bands; wherein, the feature vector samples include echo intensity distribution, echo signal attenuation ratio, and multiple echo characteristics; Based on the band response sensitivity samples and the preset optimization targets, the matching band intervals corresponding to each surface material sample are determined; wherein, the optimization targets include at least one of the following: signal-to-noise ratio target, feature enhancement target, and material discrimination target; Establish a mapping relationship between the different surface material samples, the optimization target, and the matching band interval to construct the surface material optical feature fingerprint library.

4. The autonomous obstacle avoidance method for unmanned aerial vehicles based on lidar according to claim 3, characterized in that, The step of searching for a matching band interval for the current region in a preset surface material optical feature fingerprint database based on the optical features of the surface material in the current region includes: The similarity between the optical features of the surface material in the current region and the feature vectors of each surface material sample pre-stored in the surface material optical feature fingerprint database is calculated. The matching feature vector is determined based on the similarity calculation results; Based on the matching feature vector and the currently selected optimization objective, the corresponding matching band interval is found from the mapping relationship.

5. The autonomous obstacle avoidance method for unmanned aerial vehicles based on lidar according to claim 1, characterized in that, The method further includes: If the rate of change of the signal proportion is greater than a preset threshold, then the first clustering echo signal is redefined. Based on the surface material optical characteristics corresponding to the redefined first cluster echo signal, the corresponding matching band interval is searched again. The band selection circuit is adjusted according to the redefined matching band range.

6. The autonomous obstacle avoidance method for unmanned aerial vehicles based on lidar according to claim 1, characterized in that, The programmable gain circuit includes: a programmable bandpass filter; the parameters of the programmable bandpass filter correspond one-to-one with the matched band intervals; the step of performing gain processing on the first echo signal based on the programmable gain circuit includes: The target parameters of the programmable bandpass filter are determined based on the matched band interval; Configure the programmable bandpass filter based on the target parameters so that the programmable bandpass filter filters the first echo signal; wherein, the target parameters include the center frequency and bandwidth corresponding to the matching band interval.

7. The autonomous obstacle avoidance method for unmanned aerial vehicles based on lidar according to claim 1, characterized in that, The programmable gain circuit further includes a variable gain amplifier; the step of performing gain processing on the first echo signal based on the programmable gain circuit further includes: The filtered first echo signal is amplified using the variable gain amplifier; wherein, the target gain factor of the variable gain amplifier is determined based on the following steps: First-order gain is calculated based on the signal amplitude, signal-to-noise ratio and saturation of the first echo signal, and a first-order gain factor is output. The first-order gain factor was applied to the variable gain amplifier for testing to obtain the second echo signal. Determine whether the signal amplitude of the second echo signal meets the preset target peak window; If the conditions are met, the first-order gain factor is determined as the target gain factor of the variable gain amplifier; If the condition is not met, the first-order gain factor is iteratively adjusted based on the error between the signal amplitude of the second echo signal and the target peak window until the output echo signal meets the target peak window, and the finally adjusted gain factor is determined as the target gain factor.

8. The autonomous obstacle avoidance method for unmanned aerial vehicles based on lidar according to claim 1, characterized in that, The steps of determining obstacle clusters in the current region based on the point cloud parsing dataset, and optimizing the expected planned path based on the obstacle clusters to obtain the optimized obstacle avoidance path include: Based on the point cloud parsing dataset, perform obstacle clustering and output at least one obstacle cluster; Calculate the conflict probability between each obstacle cluster and the expected planned path of the UAV, and mark the obstacle clusters with conflict probabilities greater than the preset conflict probability as obstacle avoidance points; The expected planned path is replanned to generate the optimized obstacle avoidance path, with the constraint of avoiding all the obstacle avoidance points.

9. The autonomous obstacle avoidance method for unmanned aerial vehicles based on lidar according to claim 8, characterized in that, The step of calculating the conflict probability between each obstacle cluster and the expected planned path of the UAV, and marking obstacle clusters with conflict probabilities greater than a preset conflict probability as obstacle avoidance points, includes: Calculate the shortest spatial distance between the outer contour of the obstacle cluster and the expected planned path; The collision probability is calculated based on the shortest spatial distance, the safe radius of the UAV, the relative speed of the UAV, the positioning error of the UAV, and the control error of the UAV. Obstacle clusters with a collision probability greater than the preset conflict probability are marked as obstacle avoidance points.

10. An autonomous obstacle avoidance device for unmanned aerial vehicles based on lidar, characterized in that, A control system for an unmanned aerial vehicle (UAV), wherein the transmitting end of the UAV's lidar includes a band selection circuit; the receiving end of the lidar includes a programmable gain circuit; the device includes: The initial echo signal determination module is used to perform a global scan of the area where the UAV is currently located based on the lidar to obtain a global initial echo signal. The first cluster echo signal determination module is used to cluster the global initial echo signal and determine the first cluster echo signal based on the signal ratio of multiple cluster echo signals in the clustering results. The surface material optical feature determination module is used to extract the real-time feature vector of the first clustered echo signal as the surface material optical feature of the current region; The matching band interval determination module is used to search for the matching band interval of the current region in a preset surface material optical feature fingerprint database based on the optical features of the surface material in the current region. A radar signal transmitting module is used to adjust the band selection circuit so that the lidar transmits a first test signal in the matching band interval; The radar signal receiving module is used to receive the first echo signal of the current area corresponding to the first test signal, and to perform gain processing on the first echo signal based on the programmable gain circuit to obtain the first gain echo signal of the current area. The point cloud analysis data processing module is used to perform point cloud analysis on the first gain echo signal of the current area to obtain the point cloud analysis dataset of the current area; The obstacle avoidance path processing module is used to determine the obstacle clusters in the current area based on the point cloud parsing dataset, and to optimize the expected planned path based on the obstacle clusters to obtain the optimized obstacle avoidance path.

Images