Unmanned aerial vehicle height determination and obstacle avoidance method and system based on multi-channel array laser ranging

Abstract

This invention belongs to the field of data processing technology, specifically relating to a method and system for altitude holding and obstacle avoidance of unmanned aerial vehicles (UAVs) based on multi-channel array laser ranging. The method includes: acquiring multi-channel laser ranging values, attitude angles, and installation offset angles; calculating the cosine of the gravity angle from the attitude angles and installation offset angles, and projecting the vertical depth and horizontal projection of the ranging values; obtaining the gradient deviation by neighborhood difference of the vertical depth sequence, and statistically generating adaptive boundaries to classify terrain points and obstacle points; determining the altitude using the average depth of terrain points, and avoiding obstacles by taking the minimum horizontal projection of obstacle points; when the proportion of terrain points is below a threshold, calculating the control altitude by extrapolating the vertical velocity of inertial navigation and historical altitudes, and then fusing the extrapolated and calculated altitudes. This invention can adaptively distinguish between terrain and obstacles, maintain the stability of altitude holding continuity even when data is missing, and improve flight safety and obstacle avoidance reliability.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to a method and system for altitude holding and obstacle avoidance of unmanned aerial vehicles (UAVs) based on multi-channel array laser ranging. Background Technology

[0002] Unmanned aerial vehicles (UAVs) utilize multi-channel array laser rangefinders to acquire slant range information of the surrounding environment. During high-maneuver flight, the continuous changes in the pitch and roll angles of the fuselage cause the laser beams of each channel to deflect non-fixedly from the vertical direction, resulting in a non-linear coupling between the original range value and the vertical altitude.

[0003] To eliminate this coupling, existing technologies often employ three-dimensional rigid body transformations based on quaternions or rotation matrices for full attitude registration. On embedded microcontrollers with limited computing resources, intensive matrix multiplication and trigonometric function operations can easily cause processing cycles to exceed limits, leading to data interruptions in the constant-height control loop and making it difficult to meet real-time requirements.

[0004] In distinguishing between obstacles and terrain, existing solutions often use a preset fixed absolute height threshold for binary judgment. When flying over hills, slopes and other undulating terrain, the change in the overall height of the terrain makes the absolute depth of continuous ground points approach the threshold, causing a large number of effective surface points to be mistakenly classified as obstacles, resulting in the loss of fixed height reference data.

[0005] Furthermore, the violent maneuvers of drones can cause some laser beams to be directed towards areas without echoes, resulting in a significant sparseness of effective measurement channels. In such scenarios, existing technologies fail to compensate for and fuse the temporal motion state, calculating the altitude solely based on the current sparse frame. This leads to drastic jumps in the output results, making it impossible to provide a continuous and reliable altitude reference.

[0006] The aforementioned problems restrict the flight safety and mission reliability of UAVs in complex environments. Therefore, there is an urgent need for a technical solution that can achieve attitude decoupling under low computing power conditions, adaptively distinguish between terrain and obstacles, and maintain altitude control continuity through time-domain state inference when data is severely lacking. Summary of the Invention

[0007] To address the technical challenges of achieving attitude decoupling, adaptively distinguishing between terrain and obstacles under low computing power conditions, and maintaining the continuity of altitude control through temporal state deduction when data is severely lacking, this invention provides solutions in the following aspects.

[0008] In a first aspect, the present invention provides a method for altitude control and obstacle avoidance of a UAV based on multi-channel array laser ranging, employing the following technical solution: The method for altitude control and obstacle avoidance of a UAV based on multi-channel array laser ranging includes: acquiring the ranging values ​​and real-time attitude angles of the multi-channel laser ranging, as well as the installation offset angle of each channel; calculating the cosine of the gravity angle between each beam based on the attitude angle and the installation offset angle, and projecting the ranging values ​​to obtain the vertical depth and horizontal projection; obtaining the local depth gradient deviation by neighborhood difference of the vertical depth sequence; generating an adaptive boundary based on the deviation to divide terrain points and obstacle points; obtaining the altitude control height from the average vertical depth of the terrain points, and obtaining the obstacle avoidance distance from the minimum horizontal projection of the obstacle points; calculating the effective proportion of terrain points; if it is lower than the threshold, estimating the altitude based on the inertial navigation vertical velocity and the altitude control height of the previous cycle, and calculating the weight from the proportion, fusing the estimated altitude with the current terrain point calculated altitude to obtain the control altitude.

[0009] Preferably, the cosine of the gravity angle between each beam is obtained based on the attitude angle and the installation offset angle, including: for each effective ranging channel, the real-time pitch angle is superimposed with the calibrated pitch offset angle of the corresponding channel, and the real-time roll angle is superimposed with the calibrated roll offset angle of the corresponding channel to obtain the equivalent gravity angle factor; the cosine value of the equivalent gravity angle factor is calculated as the cosine of the gravity angle between each beam.

[0010] Preferably, the method of obtaining vertical depth and horizontal projection by projecting the ranging value includes: multiplying the original ranging value of the current channel by the corresponding cosine value of the gravity angle to obtain the vertical depth; multiplying the original ranging value of the current channel by the absolute value of the sine value of the equivalent gravity angle factor to obtain the horizontal projection; and traversing all effective channels to form a vertical depth sequence and a horizontal projection sequence respectively.

[0011] Preferably, obtaining the local depth gradient deviation by neighborhood difference of the vertical depth sequence includes: selecting the current ranging channel and the immediately preceding ranging channel according to the spatial arrangement order of the sensor array; calculating the absolute value of the difference between the vertical depth of the current ranging channel and the vertical depth of the immediately preceding ranging channel as the local depth gradient deviation.

[0012] Preferably, the adaptive boundary is generated based on the deviation statistics to divide the terrain points and obstacle points, including: calculating the average value of the total local depth gradient deviation as the basic roughness benchmark; superimposing the basic roughness benchmark with the sensor's intrinsic ranging noise tolerance to obtain the adaptive boundary; for each ranging channel, if its local depth gradient deviation is less than the adaptive boundary, the corresponding ranging point is determined as a terrain point; otherwise, it is determined as an obstacle point.

[0013] Preferably, the fixed height is obtained from the average vertical depth of the terrain points, and the obstacle avoidance distance is obtained from the minimum horizontal projection of the obstacle points. This includes: averaging the vertical depths corresponding to the terrain points to obtain the fixed height; and extracting the minimum value from the horizontal projection of the obstacle points as the obstacle avoidance distance.

[0014] Preferably, the effective proportion of terrain points is calculated, including: calculating the total number of terrain points in the current period; and dividing the total number of terrain points by the total number of channels in the sensor array to obtain the effective proportion of terrain points.

[0015] Preferably, if the effective proportion of terrain points is below a threshold, the altitude is estimated based on the inertial navigation vertical velocity and the previous cycle altitude, and a weight is calculated from this proportion. The estimated altitude is then fused with the calculated altitude of the current terrain points to obtain the control altitude. This includes: when the effective proportion of terrain points is below the threshold, obtaining a prior estimate of the vertical velocity from the flight controller's inertial navigation unit as the inertial navigation vertical velocity; adding the results obtained by multiplying the previous cycle altitude and the inertial navigation vertical velocity by the time interval of the control cycle to obtain the estimated altitude; calculating the time-series decay confidence coefficient based on the effective proportion of terrain points, the threshold, and a preset confidence decay constant; and using the time-series decay confidence coefficient, performing a weighted average of the calculated altitude of the current terrain points and the estimated altitude to obtain the control altitude.

[0016] Secondly, the present invention provides a drone altitude-keeping and obstacle avoidance system based on multi-channel array laser ranging, which adopts the following technical solution: the drone altitude-keeping and obstacle avoidance system based on multi-channel array laser ranging includes: a processor and a memory, the memory storing computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned drone altitude-keeping and obstacle avoidance method based on multi-channel array laser ranging is implemented.

[0017] The embodiments of the present invention have at least the following beneficial effects: The present invention has the following beneficial effects: 1. By constructing a non-orthogonal dimension-reduced projection model based on the absolute gravity angle factor, the equivalent gravity angle cosine is obtained by superimposing the factory-installed offset angle of each channel with the real-time pitch and roll angles. Only one scalar multiplication is needed to decompose the original slant range into vertical depth and horizontal projection. The three-dimensional rotation matrix is ​​replaced by table lookup and multiplication accumulation, which can reduce the nonlinear coupling between attitude and ranging in high-maneuver flight on a low computing power platform and avoid misjudging the ground as a vertical obstacle due to the forward tilt of the sensor.

[0018] 2. By using adaptive topography segmentation based on local depth gradient, the absolute gradient deviation of adjacent channels in the vertical depth sequence is calculated, and the mean deviation is superimposed with the sensor intrinsic noise to form a dynamic continuity boundary. This boundary can be adaptively adjusted according to the overall undulation of the terrain, fully preserving gently changing ground points such as hills and slopes as elevation references, and only separating abrupt change points as obstacles, effectively avoiding the erroneous removal of elevation data by fixed thresholds under undulating terrain. At the same time, the minimum value of the horizontal projection corresponding to the obstacle point is extracted as the obstacle avoidance distance, realizing terrain perception and safe avoidance control.

[0019] 3. When the effective proportion of shape points falls below a preset threshold, the system actively calls upon the prior inertial navigation Z-axis velocity and the historical altitude of the previous cycle to perform kinematic integration and extrapolation, obtaining an extrapolated altitude independent of the current laser observation. The system also calculates a dynamic trust coefficient based on the real-time effective coverage and trust decay constant, and performs soft weighted fusion of the sparse frame solution altitude and extrapolated altitude to reduce control quantity jumps and glitch oscillations caused by data loss. This ensures that the altitude signal is continuous and smooth during the blinding period, and automatically and gradually returns to the normal weights when the observation is restored, thereby enhancing the robustness and flight safety of the altitude hold system. Attached Figure Description

[0020] Figure 1 The schematic diagram illustrates the steps of the UAV altitude hold and obstacle avoidance method based on multi-channel array laser ranging in this invention. Detailed Implementation

[0021] 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, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] S1: Obtain the raw ranging values ​​of multiple channels, real-time attitude angles, and installation offset angles of each channel.

[0024] The multi-channel array laser ranging module carried by the drone includes Each independent transmit / receive channel, typical value The value is 16, which can be adjusted by the implementer according to the specific implementation scenario.

[0025] The flight control system's processing and control module sends read commands to the ranging module at fixed intervals via the communication bus to directly obtain the original ranging value sequence for the current period.

[0026] Each channel returns a raw echo distance value in meters (m), and invalid sampling signals such as no echo or blind spots are marked for state rejection, so that the processing module can directly identify the valid measurement channels. The state rejection marking of invalid sampling signals such as no echo or blind spots is a well-known technical means and will not be described in detail here.

[0027] The processing and control module reads real-time attitude angles from the inertial measurement unit of the flight controller via an internal link. These real-time attitude angles specifically include the real-time pitch angle. and real-time roll angle The unit is degrees (°), which is continuously calculated, updated, and output by the inertial measurement unit.

[0028] While acquiring the real-time attitude angles, the processing and control module also needs to read the set of installation offset angles for each channel corresponding to the UAV model.

[0029] This set of installation offset angles includes each individual channel. The corresponding calibration pitch offset angle and calibrated roll offset angle The unit is degrees (°).

[0030] These offset angles precisely describe the inherent field-of-view divergence angle of each beam within the machine coordinate system, and are hardware constants that are fixed after factory calibration.

[0031] After the above parallel reading operations, the processing and control module obtains the basic data for sensing within the current control cycle, including... The original ranging value sequence of each channel, and the real-time pitch angle With real-time roll angle And the calibrated pitch offset angle of each channel. and calibrated roll offset angle .

[0032] S2: Calculate the cosine of the equivalent gravity angle and project it to obtain the vertical depth and horizontal projection.

[0033] Because during high-maneuver flight, the real-time pitch and roll angles of the drone cause the multi-channel laser beam to deviate from the vertical direction, affecting the original ranging value. A nonlinear attitude coupling occurs between the pitch angle and the actual vertical height. Decoupling this coupling using a full 3D rigid body transformation on an embedded microcontroller would result in excessive computational overhead. Therefore, this step eliminates this coupling through scalar projection, leveraging the invariance of the absolute gravity direction. The input is the acquired real-time pitch angle. Real-time roll angle The calibrated pitch offset angle of each channel and calibrated roll offset angle and the original ranging values ​​of each effective channel. The angle unit is degrees (°), and the distance unit is meters (m).

[0034] For each effective ranging channel Real-time pitch angle The calibrated pitch offset angle of the corresponding channel Add them together, and simultaneously add the real-time roll angle. The calibrated roll offset angle of the corresponding channel Adding them together yields the equivalent gravitational angle factor. This factor combines the instantaneous attitude of the fuselage and the factory installation deviation of the channel, and represents the actual angle between the current laser beam and the direction of gravity normal, in degrees (°).

[0035] calculate When determining the cosine value, based on engineering conditions where both the pitch and roll stacking angles are within a small angle range, where the small angle is taken as an empirical value of 15° (a hyperparameter that can be adjusted by the implementer according to the specific implementation scenario), an approximate method of multiplying the pitch cosine and roll cosine is used to obtain the cosine ratio required for vertical projection with extremely low computational load. The specific formula is as follows: In the formula, The cosine value of the composite pitch angle. The product of the two values ​​is the cosine of the roll angle. The calculation process only requires two cosine lookup tables or quick approximation calculations and one multiplication.

[0036] The original ranging value of the current channel The cosine value of the angle between the angle and the gravity is required. Multiply to obtain the vertical depth of the channel. The calculation formula is: in, The unit is meters. For dimensionless values, the product The unit is meters, which represents the actual ground clearance of the ranging point along the vertical direction to the UAV ranging reference plane, after eliminating depth distortion caused by attitude angle.

[0037] At the same time, the absolute value of the sine of the equivalent gravitational angle factor is obtained based on the trigonometric relationship. Compare it with the original distance measurement value Multiply to obtain the horizontal projection The calculation formula is: In the formula, It can be obtained by looking up a table or by taking the square root of a previously calculated cosine value using the square relationship; it is a dimensionless value. The unit is meters, and its physical meaning is the projection distance of the laser beam on the horizontal plane, which maps the degree of approach of obstacles in front of the drone.

[0038] Following the physical channel numbering order of the multi-channel array laser ranging hardware, traverse all ranging channels that are marked as valid at the bottom layer. For each channel, the above-mentioned equivalent gravity angle factor synthesis, cosine calculation, vertical projection, and horizontal projection operations are repeated to determine the vertical depth of each channel. Arrange the vertical depth sequence sequentially, and project the horizontal depth of each channel. By sequentially assembling the horizontal projection sequence, we obtain the pure vertical depth set and horizontal projection set of the decoupled attitudes output in this step.

[0039] S3: Obtain the local depth gradient deviation through neighborhood difference.

[0040] After obtaining the vertical depth sequence of the decoupled attitude, neighborhood difference calculation is performed on the sequence to extract the local depth change information at each channel position. Since the vertical depth of adjacent laser beams on continuous terrain gradually changes smoothly along the sensor arrangement direction, and the vertical depth changes significantly when encountering the edge of an obstacle, the degree of local change at that position can be quantified by calculating the absolute value of the difference between the vertical depths of adjacent channels. The resulting value is the local depth gradient deviation.

[0041] For vertical depth sequences ,in For the number of effective channels and The unit is meters (m), and the channel index is determined according to the spatial arrangement order of the sensor array. from Starting from the first, increment sequentially, calculate the channel. Local depth gradient deviation The calculation formula is: In the formula Indicates channel Vertical depth Indicates channel The vertical depths, both taken from the vertical depth sequence output in step S2, are calculated by taking their absolute values. It only reflects the degree of depth difference without distinguishing the direction of terrain uplift or subsidence. For the first channel in a channel sequence, since it does not have an adjacent preceding channel, its local depth gradient deviation is set to... Right now This corresponds to the assumption that there are no abrupt changes at the physical boundary.

[0042] After traversing all valid ranging channels, the length is obtained as Local depth gradient deviation sequence In this sequence, the deviation values ​​corresponding to continuous ground regions approach the following values. The deviation values ​​corresponding to terrain fractures or obstacle edges are relatively large, reaching several meters in magnitude. This step completes the conversion from the vertical depth sequence to the local depth gradient deviation sequence.

[0043] S4: Generate adaptive boundaries based on deviation statistics and divide terrain points and obstacle points.

[0044] Adaptive boundaries are generated based on deviation statistics to distinguish between terrain points and obstacle points. Since the real terrain surface exhibits spatially gradual changes, the local depth gradient deviation corresponding to continuous ground ranging points... The values ​​are concentrated in small ranges, while those corresponding to obstacles or terrain faults are... The numerical values ​​are significantly large. Therefore, the processing and control module performs central tendency statistics on the output deviation sequence to obtain a basic roughness benchmark characterizing the inherent undulation of the terrain. The statistical method can use the average value; this implementation uses the average value as the basic roughness benchmark and calculates all... The average value is denoted as The unit is meters. The average is a well-known technical method and will not be elaborated further.

[0045] Obtaining the basic roughness baseline Furthermore, because laser ranging hardware, when measuring the same plane in a stationary state, will also produce slight random deviations between adjacent channels due to the sensor's intrinsic electrical noise and quantization error, it is necessary to consider the following in engineering practice: Compared with the sensor's intrinsic ranging noise margin Superimposed to form an adaptive limit that can correctly distinguish between terrain continuity and obstacle steps. , Based on the specifications of the selected multi-channel array laser ranging module, the preset parameters are as follows: Meters, a hyperparameter that can be adjusted by the implementer based on the noise level of the specific sensor model, are measured in meters. The formula for calculating the adaptive limit is: In the formula The unit is meters, and this limit reflects the upper limit of depth drop tolerance that can still be considered a continuous surface under the current terrain roughness.

[0046] The processing and control module handles the index. from to For each valid ranging channel, a determination is made, and the local depth gradient deviation of the current channel is calculated. With adaptive limits If a comparison is made, If the distance measurement point corresponding to the channel is found to be in the terrain point set, then it is added to the obstacle point set; otherwise, it is added to the obstacle point set. (For the first channel index...) ,That It has been set in the previous step. The value is less than 1 meter, so it is included in the terrain point set. After traversal, the current set of terrain points and obstacle points are output, thus completing the division of terrain and obstacles.

[0047] S5: Obtain the fixed height from the average vertical depth of the terrain points, and obtain the obstacle avoidance distance from the minimum horizontal projection of the obstacle points.

[0048] After dividing the terrain points and obstacle points, the processing and control module obtains the terrain point set and the obstacle point set. This is because the vertical depth of each ranging point in the terrain point set... In continuous ground areas, the depths fluctuate slightly around the true ground height. Averaging these depths can suppress transient noise and reflect the current average terrain height. Therefore, the processing and control module calculates the vertical depth of all ranging channels within the terrain point set. Perform mean calculation to determine the fixed height. The calculation formula is: In the formula, The total number of topographic points, dimensionless. For terrain point channel index set, For the first The vertical depth of the passage, measured in meters (m). The current cycle's altitude is set in meters (m).

[0049] Meanwhile, the ranging points in the obstacle point set correspond to forward protrusions, and the horizontal projections of these points are obtained in step S2. The approach distance to the obstacle in the horizontal direction is directly quantified; the smaller the value, the closer the obstacle. Therefore, the processing and control module traverses the set of obstacle points. If the value is not empty, the minimum value is taken as the obstacle avoidance distance. ;otherwise Meters indicate no near-field obstacles. Among them, For obstacle point channel index set, For the first The horizontal projection of the passageway, in meters (m). The obstacle avoidance distance is measured in meters (m), which represents the horizontal distance to the nearest obstacle in the current field of view.

[0050] S6: Inertial extrapolation weighted fusion control height based on effective proportion.

[0051] After obtaining the fixed altitude for the current period, the processing and control module performs a quantitative assessment of the terrain point coverage used to calculate the fixed altitude.

[0052] During the period of intense attitude maneuvering of the UAV, some laser beams are pointed to the sky or the area without echo, which significantly reduces the number of ranging channels that are included in the terrain point set in the current period. If the average vertical depth of a few terrain points is directly used as the control altitude, it will introduce significant random fluctuations. Therefore, this step first calculates the effective proportion of terrain points in the current period.

[0053] The processing and control module iterates through the terrain point set, counts the total number of terrain points it contains, and denotes it as... , dimensionless.

[0054] Will Divide by the total number of channels in the sensor array The effective percentage of terrain points was obtained. The calculation formula is: In the formula This refers to the total number of independent transmit and receive channels in a multi-channel array laser ranging module, typically 16, and is dimensionless. The effective percentage of terrain points in the current period represents the coverage ratio of effective ground information with quantifiable height in the ranging field of view, and is dimensionless.

[0055] After obtaining the effective percentage of terrain points in the current period. Then, the processing and control module compares it with the preset lower limit threshold for the effective terrain sampling ratio. Compare them.

[0056] Based on the specifications of the array hardware, the empirical value is 18%, which is dimensionless and is a hyperparameter that can be adjusted by the implementer according to the degree of sensor redundancy.

[0057] when Greater than or equal to When this occurs, it indicates that there are a sufficient number of continuous ground reference points within the current field of view, and the fixed altitude calculated in step S5 is obtained. It can reliably reflect the actual ground clearance, and the processing and control module directly... As the control height of the current cycle The output is in meters (m).

[0058] when Less than When this occurs, it indicates that the effective shape points are severely sparse, and relying solely on the current average vertical depth to calculate the height will introduce a significant risk of jumps. At this point, the processing and control module activates the time-domain state compensation and fusion mechanism.

[0059] The processing and control module sends a data request to the flight controller's inertial navigation unit via an internal high-speed link to obtain the prior estimate of the vertical velocity output by the inertial navigation unit's Kalman filter in the current cycle. As the vertical velocity for inertial navigation, the unit is meters per second (m / s). It should be noted that... The value is defined as positive when it is upward.

[0060] At the same time, the historical fixed altitude value stored and output in the previous control cycle is read from memory. The unit is meters (m).

[0061] By utilizing the kinematic integral relationship and the temporal accumulation of height and vertical velocity from the previous cycle, the estimated height can be deduced without relying on current laser observations. The calculation formula is: In the formula, This is the fixed control cycle time interval for the flight control unit, empirically set to 0.01 seconds (s). The kinematic height, measured in meters (m), provides a priori estimate of the current period's height above the ground using the inertial motion of the organism.

[0062] Obtain the height of the simulation Subsequently, the processing and control module calculates the time-series decay trust coefficient based on the current effective proportion of points, the preset lower threshold, and the trust decay constant. This is used to control the fusion weight of the inferred height and the current terrain point solution height in the final output.

[0063] The calculation formula is: In the formula, the minimum value of 0.1 is used to avoid relying entirely on inertial deduction. This represents the effective percentage of terrain points in the current period, dimensionless. To effectively form the lower limit threshold of the sampling ratio, it is dimensionless; 0.8 is the trust decay constant, which is empirically taken as 0.8, dimensionless, and is a hyperparameter that can be adjusted by the implementer according to the requirements of fusion smoothness.

[0064] The value is limited to Within the interval, it represents the degree of confidence in the height calculated based on actual laser measurements, given the current sparsity. The closer it is to 0, the more it depends on the projected height; the closer it is to 1, the more it depends on the calculated height of the current terrain point.

[0065] Calculate the height of the current terrain point That is, the fixed height obtained in step S5 Highly related to kinematic deduction Perform weighted fusion to calculate the control height for the current cycle. The formula is: In the formula, That is The unit is meters (m). To extrapolate the height, the unit is meters (m). The final output control height, in meters (m).

[0066] This weighted fusion method prioritizes inertial height estimation when terrain points are sparse, ensuring output continuity and stability. However, as the proportion of effective terrain points increases, the confidence coefficient... The height is increased synchronously, and the control height is smoothly regressed to the laser measured height result. The entire processing is triggered by the control ratio and completes the fusion convergence.

[0067] This invention also discloses a UAV altitude-keeping and obstacle avoidance system based on multi-channel array laser ranging, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the UAV altitude-keeping and obstacle avoidance method based on multi-channel array laser ranging of this invention. The system also includes other components well-known to those skilled in the art, such as a communication bus and communication interface; their configurations and functions are known in the art and will not be described further here.

[0068] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method for UAV altitude setting and obstacle avoidance based on multi-channel array laser ranging, characterized in that, include: Acquire the ranging values ​​and real-time attitude angles of multi-channel laser ranging, as well as the installation offset angle of each channel; The cosine of the gravity angle between each beam is calculated based on the real-time attitude angle and the installation offset angle, and the vertical depth and horizontal projection are obtained by projecting the ranging value onto it. The local depth gradient deviation is obtained by subtracting the neighborhood of the vertical depth sequence; an adaptive boundary is obtained based on the local depth gradient deviation to distinguish between terrain points and obstacle points. The fixed height is obtained from the average vertical depth of the terrain points, and the obstacle avoidance distance is obtained from the minimum horizontal projection of the obstacle points. The effective percentage of terrain points is statistically analyzed; if it is lower than the threshold, the height is estimated based on the vertical velocity of inertial navigation and the height of the previous cycle, and the weight is calculated from the percentage. The estimated height is then combined with the calculated height of the current terrain points to obtain the control height.

2. The UAV altitude setting and obstacle avoidance method based on multi-channel array laser ranging according to claim 1, characterized in that, The acquisition of the ranging values ​​and real-time attitude angles of the multi-channel laser ranging, as well as the installation offset angles of each channel, includes: The original ranging value sequence is obtained through multi-channel array laser ranging hardware; The real-time attitude angles are obtained through the inertial measurement unit of the flight controller, including the real-time pitch angle and the real-time roll angle. The installation offset angle of each channel is obtained through data acquisition equipment. The installation offset angle of each channel includes the calibration pitch offset angle and the calibration roll offset angle of each channel.

3. The UAV altitude setting and obstacle avoidance method based on multi-channel array laser ranging according to claim 2, characterized in that, The step of calculating the cosine of the gravity angle between each beam based on the real-time attitude angle and the installation offset angle includes: For each effective ranging channel, the real-time pitch angle is superimposed with the calibrated pitch offset angle of the corresponding channel, and the real-time roll angle is superimposed with the calibrated roll offset angle of the corresponding channel to obtain the equivalent gravity angle factor. Calculate the cosine value of the equivalent gravity angle factor, and use it as the cosine of the gravity angle of each beam.

4. The UAV altitude setting and obstacle avoidance method based on multi-channel array laser ranging according to claim 3, characterized in that, The step of projecting the distance measurement value to obtain the vertical depth and horizontal projection includes: Multiply the original distance measurement value of the current channel by the corresponding cosine value of the gravity angle to obtain the vertical depth; Multiply the original distance measurement value of the current channel by the absolute value of the sine of the equivalent gravity angle factor to obtain the horizontal projection; Traverse all valid channels to form the vertical depth sequence and horizontal projection sequence, respectively.

5. The UAV altitude holding and obstacle avoidance method based on multi-channel array laser ranging according to claim 1, characterized in that, The local depth gradient deviation obtained by subtracting the neighborhood of the vertical depth sequence includes: According to the spatial arrangement order of the sensor array, select the current ranging channel and the immediately preceding ranging channel; The absolute value of the difference between the vertical depth of the current ranging channel and the vertical depth of the immediately preceding ranging channel is calculated as the local depth gradient deviation.

6. The UAV altitude holding and obstacle avoidance method based on multi-channel array laser ranging according to claim 1, characterized in that, The adaptive boundary obtained based on the local depth gradient deviation to delineate terrain points and obstacle points includes: Calculate the average value of the local depth gradient deviation of the entire quantity, and use it as the base roughness benchmark; The adaptive limit is obtained by superimposing the basic roughness benchmark with the sensor's intrinsic ranging noise tolerance. For each ranging channel, if its local depth gradient deviation is less than the adaptive limit, the corresponding ranging point is determined as a terrain point; otherwise, it is determined as an obstacle point.

7. The UAV altitude holding and obstacle avoidance method based on multi-channel array laser ranging according to claim 1, characterized in that, The process of obtaining a fixed height from the average vertical depth of terrain points and obtaining the obstacle avoidance distance from the minimum horizontal projection of obstacle points includes: The average vertical depth corresponding to the terrain point is used to obtain the fixed height. The minimum value is extracted from the horizontal projection corresponding to the obstacle point and used as the obstacle avoidance distance.

8. The UAV altitude holding and obstacle avoidance method based on multi-channel array laser ranging according to claim 1, characterized in that, The effective percentage of the statistical terrain points includes: Count the total number of terrain points in the current period; The effective percentage of the terrain points is obtained by dividing the total number of terrain points by the total number of channels in the sensor array.

9. The UAV altitude holding and obstacle avoidance method based on multi-channel array laser ranging according to claim 1, characterized in that, If the altitude is below a threshold, the altitude is estimated based on the inertial navigation vertical velocity and the previous cycle's fixed altitude, and a weight is calculated from this proportion. The estimated altitude is then fused with the altitude calculated from the current terrain points to obtain the control altitude, including: When the effective proportion of the terrain points is lower than the threshold, the prior estimate of the vertical velocity is obtained from the inertial navigation unit of the flight controller and used as the inertial navigation vertical velocity. The calculated altitude is obtained by adding the calculated altitude obtained by multiplying the fixed altitude of the previous cycle by the vertical velocity of the inertial navigation by the time interval of the control cycle. The time-series decay trust coefficient is calculated based on the effective proportion of the terrain points, the threshold, the preset trust decay constant, and the preset minimum trust coefficient, and the trust coefficient is not lower than the minimum trust coefficient. The control height is obtained by weighting the calculated height of the current terrain point and the projected height using the time-decay confidence coefficient.

10. A UAV altitude holding and obstacle avoidance system based on multi-channel array laser ranging, characterized in that, include: The processor and memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the UAV altitude-keeping and obstacle avoidance method based on multi-channel array laser ranging according to any one of claims 1-9.

Images