CN111708375A - A UAV Obstacle Avoidance System Based on LiDAR and Octagonal Air Wall Algorithm - Google Patents

Abstract

The invention is a UAV obstacle avoidance system based on laser radar and octagonal air wall algorithm, comprising a power execution module, a microprocessor module, a sensor module, an attitude angle controller, a laser radar module, Octagon air wall obstacle avoidance algorithm, in which: the power execution module provides power for the UAV; the microprocessor module receives the data sent by the sensor module and the instructions sent by the remote control, makes logical judgments on the two kinds of information, and controls the unmanned aerial vehicle. The sensor module is used to calculate the current attitude angle of the UAV; the attitude angle controller controls the UAV with the actual attitude angle by receiving the signal from the remote control; the laser radar module obtains the distance from the obstacle to the UAV and relative angle; the octagonal air wall obstacle avoidance algorithm processes the distance data and angle data collected by the lidar to provide accurate obstacle avoidance information and realize the precise obstacle avoidance of the UAV. The invention solves the problem of outliers in the data collected by the laser radar.

Description

A UAV Obstacle Avoidance System Based on LiDAR and Octagonal Air Wall Algorithm

technical field

The invention relates to the field of UAV obstacle avoidance and navigation, in particular to an UAV obstacle avoidance system based on laser radar and an improved octagonal air wall obstacle avoidance algorithm.

Background technique

There are currently three practical UAV obstacle avoidance schemes: obstacle avoidance schemes based on ultrasonic sensors; obstacle avoidance schemes based on binocular vision sensors; obstacle avoidance schemes based on lidar sensors.

(1) Obstacle avoidance scheme based on ultrasonic sensor:

Ultrasonic sensors are the most widely used sensors, such as parking sensors. The ultrasonic sensor has a certain measurement constraint angle (about 10°～70°), and the measurement range is 4～10m. Ultrasonic sensors are inexpensive and have a simple measurement principle. However, the ultrasonic waves emitted by ultrasonic sensors are mechanical waves, which are prone to attenuation and interference, resulting in low measurement accuracy; and the data measured by ultrasonic sensors is small, which is not conducive to smooth obstacle avoidance by UAVs.

(2) Obstacle avoidance scheme based on binocular vision sensor:

The binocular vision sensor can not only measure the distance of obstacles, but also obtain rich image information and conduct 3D modeling of obstacles. The binocular vision sensor has the advantages of wide measurement range and high precision. Since the binocular vision sensor collects image information, the calculation and data transmission volume of the data processing module is also large, resulting in high power consumption; the binocular vision sensor is greatly affected by light and cannot be used in dark scenes.

(3) Obstacle avoidance scheme based on lidar sensor:

The lidar sensor can not only collect the distance of the obstacle, but also the relative angle of the obstacle. It has the advantages of high precision, a measurement range of 360°, and good performance in low light environments; the disadvantage is that it is easily interfered by strong light, but this shortcoming can be compensated by algorithms.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a UAV obstacle avoidance system based on laser radar and octagonal air wall algorithm, so as to solve the problem that the current UAV obstacle avoidance system has low precision and cannot achieve accurate UAV avoidance. handicap defect.

The present invention solves its technical problem and adopts following technical scheme:

The UAV obstacle avoidance system based on the laser radar and the octagonal air wall algorithm provided by the present invention specifically includes: a power execution module, a microprocessor module, a sensor module, an attitude angle controller, a laser radar connected by electrical signals Module, octagonal air wall obstacle avoidance algorithm, in which: the power execution module provides power for the UAV; the microprocessor module receives the data sent by the sensor module and the instructions sent by the remote control, makes logical judgments on the two kinds of information, and controls the UAV navigation and flight; the sensor module is used to calculate the current attitude angle of the UAV; the attitude angle controller controls the UAV with the actual attitude angle by receiving the remote control signal; the lidar module obtains obstacles to the UAV distance and relative angle; the octagonal air wall obstacle avoidance algorithm processes the distance data and angle data collected by the lidar to provide accurate obstacle avoidance information and realize the precise obstacle avoidance of the UAV.

The power execution module is composed of an electronic governor, a propeller and a brushless DC motor; the propeller is installed on the brushless DC motor, and the rotation of the motor drives the rotation of the propeller to provide lift for the drone; the electronic governor is installed on the microcomputer. Between the processor and the DC brushless motor, since the size of the PWM signal output by the microprocessor module is not enough to drive the DC brushless motor to rotate, it is necessary to add an electronic speed governor to amplify the power, so that the microprocessor can output a sufficiently large output. PWM signal to drive the DC brushless motor.

The microprocessor chip used in the microprocessor module is STM32F427VIT6, and its rich peripheral interfaces can communicate with sensors and data storage chips using different communication methods.

The sensor module includes a three-axis accelerometer, a three-axis magnetometer and a three-axis gyroscope, wherein: the three-axis accelerometer is installed on the PIXHAWK flight control board, and is used to measure the three-axis acceleration when the drone is flying; The magnetometer is installed on the PIXHAWK flight control board to measure the three-axis magnetic field strength when the UAV is flying; the three-axis gyroscope is installed on the PIXHAWK flight control board to measure the angular velocity around the three-axis when the UAV is flying.

The invention adopts the six-axis sensor ICM-20608-G and the magnetometer HMC5983. The six-axis sensor ICM-20608-G integrates two sensors, a three-axis accelerometer and a three-axis gyroscope. The pitch angle and roll angle of the man-machine; the yaw angle of the UAV is calculated by the data collected by the magnetometer HMC5983, so as to obtain the attitude angle information of the current UAV when it is flying.

The attitude angle controller adopts a double closed-loop PID controller, the outer loop is designed as an angle loop, the input of the angle loop is the deviation between the desired attitude angle and the actual attitude angle, and the output is the desired angular velocity; the inner loop is designed is the angular velocity loop, the input of the angular velocity loop is the deviation of the expected angular velocity and the actual angular velocity, and the output is the expected torque, so as to control the attitude of the UAV during flight.

In the design of the outer ring of the present invention, the desired attitude angle is the attitude angle input by the remote controller, and the actual attitude angle is the attitude angle calculated by the inertial unit.

In the inner loop design of the present invention, the desired angular velocity is the output of the angular loop, and the actual angular velocity is the measurement value of the gyroscope.

The described lidar module adopts omnidirectional lidar SF40, which collects the distance and relative angle from the obstacle to the UAV, and provides initial data support for the subsequent UAV obstacle avoidance.

The described octagonal air wall obstacle avoidance algorithm is: collecting and merging raw distance measurements through the lidar module, including storing and using the nearest distance and angle in 8 sectors; the width of each sector is 45 degrees , sector 0 points to the front, sector 1 points to the right, and so on; from these distances and angles, a fence, a two-dimensional vector array, is built around the drone, and the fence points fall between zones. In a straight line, keep a certain distance; when no obstacles are detected in all 8 sectors, the available empty sectors will be filled with the distance from the adjacent sector, forming a "cup" fence to protect the unmanned The machine does not hit the target.

Compared with the prior art, the present invention has the following main beneficial effects:

Under the interference of strong light, the lidar sensor will output erroneous distance data, which affects the UAV's accurate obstacle avoidance. To this end, the present invention proposes a later algorithmic processing for this shortcoming of the lidar: the original data collected by the lidar are filtered by using a median filter, a low-pass filter and a sliding window average filter in sequence. Experiments show that after filtering by median filter, low-pass filter and sliding window average filter, outliers in the distance data are filtered out, making the distance data more accurate. Enables UAVs to avoid obstacles more accurately.

Description of drawings

Figure 1 shows the overall scheme of the UAV obstacle avoidance system.

Figure 2 is a physical map of the SF40 omnidirectional lidar.

Figure 3 shows the octagonal air wall obstacle avoidance algorithm.

Figure 4 shows the UAV obstacle avoidance method when no obstacle is detected.

Figure 5 shows the data collected by the octagonal air wall obstacle avoidance algorithm before optimization. Among them: Figures 5-1 to 5-8 are schematic diagrams of specific data waveforms collected by the octagonal air wall obstacle avoidance algorithm before optimization.

Figure 6 shows the original data filtering process.

Figure 7 shows the data collected by the optimized octagonal air wall obstacle avoidance algorithm. Among them: Figures 7-1 to 7-8 are schematic diagrams of specific data waveforms collected by the optimized octagonal air wall obstacle avoidance algorithm.

Figure 8 is a flow chart of obstacle avoidance program design.

Figure 9 is the change diagram of the obstacle distance facing the UAV direction, the input value of the pitch channel of the remote control, and the actual pitch angle. Among them: Figure 9-1 is the obstacle distance facing the UAV, Figure 9-2 is the input value of the pitch channel of the remote control, and Figure 9-3 is the change diagram of the actual pitch angle.

In Figure 2: 1. Laser module; 2. Brushless motor module.

Detailed ways

The invention provides a UAV obstacle avoidance system based on laser radar and an improved octagonal air wall algorithm. The system selects the SF40 omnidirectional laser radar sensor to obtain the distance and angle information of obstacles in the current flight environment, and then uses the eight The edge air wall obstacle avoidance algorithm is used for real-time obstacle avoidance, and finally combined with the quadrotor UAV flight control system to complete the autonomous obstacle avoidance flight of the quadrotor aircraft. In order to overcome the defect that lidar is easily affected by strong light interference, three filters, median filter, low-pass filter, and sliding window average filter, are added successively to filter out the interference points generated under strong light, so that the drone can operate at the same time. Accurate obstacle avoidance in strong light.

Below in conjunction with embodiment and accompanying drawing, it is further illustrated, but does not limit the present invention.

The UAV obstacle avoidance system based on the laser radar and the improved octagonal air wall obstacle avoidance algorithm provided by the present invention is improved on the basis of the PIXHAWK open source flight control, and is suitable for autonomous UAVs in the actual flight process. Obstacle avoidance, especially suitable for quadrotor UAV to avoid obstacles accurately.

The UAV obstacle avoidance system based on the laser radar and the improved octagonal air wall obstacle avoidance algorithm provided by the present invention includes a power execution module, a microprocessor module, a sensor module, an attitude angle controller, and a laser connected by electrical signals. The radar module and the octagonal air wall obstacle avoidance algorithm, in which: the power execution module provides power for the UAV; the microprocessor module receives the data sent by the sensor module and the instructions sent by the remote controller, and makes logical judgments on the two kinds of information. Control the navigation and flight of the drone; the sensor module includes a three-axis accelerometer, a three-axis magnetometer and a three-axis gyroscope, which are used to calculate the current attitude angle of the drone; the attitude angle controller receives the signal from the remote control to actual The attitude angle controls the UAV; the lidar module obtains the distance and relative angle from the obstacle to the UAV; the octagonal air wall obstacle avoidance algorithm processes the distance data and angle data collected by the lidar to provide accurate obstacle avoidance information , to achieve accurate obstacle avoidance of UAVs.

The present invention has two sub-obstacle avoidance systems, as shown in FIG. 1 .

The first sub-obstacle avoidance system is a data processing system, which is composed of a laser radar, a data processing module and a communication port. Its main function is: first obtain the distance and angle between the UAV and the obstacle through lidar, then process the collected data through the data processing module, and finally send the processed data to the flight control system through the communication port.

The second sub-obstacle avoidance system is the flight control system, which consists of the PIXHAWK flight control board and the remote control. Its main function is: read the distance and angle information sent by the data processing system, and at the same time receive the control signal sent by the remote control, and then combine the data sent by the data processing system and the remote control signal to make judgments to determine whether the drone is performing obstacle avoidance. The attitude is still flying according to the remote control signal.

The laser radar SF40 used in the present invention is an existing sensor. SF40 is a two-dimensional lidar with low power consumption and a detection range of 360 degrees. The effective measurement distance of this lidar is 0-100m, and the measurement angle resolution is less than 1°. The resulting 2D point cloud data can be used for map, coordinate and environment modeling. SF40 uses a detection frequency of 1Hz, 2.25Hz or 4.5Hz, and the three detection frequencies correspond to distance resolutions of 0.03m, 0.06m and 0.12m.

The structure of the lidar SF40 is shown in Figure 2. A laser module 1 is provided, so that the lidar SF40 can transmit and receive laser light; a brushless motor module 2 is provided, which drives the laser module 1 to rotate, so that the lidar SF40 can collect distance information within a 360-degree range.

The power execution module is composed of an electronic governor, a propeller and a DC brushless motor. The brushless DC motor not only has the advantages of large starting torque and wide speed regulation range of the DC brush motor, but also eliminates the mechanical commutation structure and uses an electronic converter, which can control the speed of the motor more accurately. Since the size of the PWM modulation signal output by the microprocessor module is not enough to drive the brushless DC motor to rotate, it is necessary to add an electronic speed governor to amplify the power of the PWM's internal FET and the microprocessor, so that the microprocessor can output a sufficiently large output. PWM modulation signal to drive the DC brushless motor. The propeller is installed on the DC brushless motor, and the rotation of the motor drives the propeller to rotate, providing lift for the quadrotor UAV. The brushless DC motor selected Motor6010-KV270 motor from Corbett Company, and the electronic governor selected SkyWallker-40A.

The microprocessor chip of the microprocessor module is STM32F427VIT6. STM32F427VIT6 has an ARM Cortex-M4 core with a main frequency of up to 168MHz, a floating-point operation unit and a full set of executable Digital Signal Processor (DSP) instruction sets, so the data processing capability of the processor is greatly improved; STM32F427VIT6 has 2MB The on-chip Flash memory unit, a 256kB Static Random Access Memory (SRAM), a 4kB SRAM, and a PseudoStatic Random Access Memory (PSRAM) can store all The code and data required; STM32F427VIT6 has a wealth of peripheral interfaces, which can communicate with sensors and data storage chips using different communication methods.

The sensor module selects the six-axis sensor ICM-20608-G and the magnetometer HMC5983. The six-axis sensor ICM-20608-G integrates two sensors, a three-axis accelerometer and a three-axis gyroscope. The data collected by the three-axis accelerometer is used to calculate the pitch angle and roll angle of the UAV; the data collected by the magnetometer HMC5983 The data calculates the yaw angle of the UAV to obtain the current attitude angle information of the UAV.

The attitude angle controller selects a double closed-loop PID controller. The outer loop of the controller is designed as an angle loop, the input of the angle loop is the deviation between the desired attitude angle (the attitude angle input by the remote control) and the actual attitude angle (the attitude angle calculated by the inertial unit), and the output is the desired angular velocity ; The inner loop of the controller is designed as an angular velocity loop, the input of the angular velocity loop is the deviation between the expected angular velocity (the output of the angle loop) and the actual angular velocity (the measured value of the gyroscope), and the output is the expected torque, so as to control the unmanned quadrotor. machine attitude. This double closed-loop cascade PID controller is beneficial to improve the stability and accuracy of the system, and can control the precise and stable flight of the quadrotor UAV. The power execution module, the sensor module, the microprocessor module and the attitude angle controller module constitute the flight control system of the UAV.

The lidar module selects omnidirectional lidar SF40. The omnidirectional lidar SF40 collects the distance and relative angle from the obstacle to the UAV to provide initial data support for the subsequent UAV obstacle avoidance.

The octagonal air wall obstacle avoidance algorithm in the present invention is that the omnidirectional lidar sensor SF40 collects raw distance measurements, which are combined so that only the closest distances and angles within the 8 sectors are stored and used. Each sector is 45 degrees wide, with sector 0 pointing forward, sector 1 pointing right, and so on. From these distances and angles, a fence (a 2D vector array) is built around the drone. The fence points fall on the straight line between the districts and keep a certain distance, as shown in Figure 3. When no obstacle is detected for all 8 sectors, empty sectors (if available) will be filled with distances from adjacent sectors. This conveniently creates a "cup" fence that is more likely to protect the drone from hitting the target, as shown in Figure 4.

Set the obstacle avoidance distance to 5m, set the maximum detection distance of the omnidirectional lidar SF40 to 20m, and let the lidar collect data. The lidar collects 8 sectors, and draws a curve with the number of data as the abscissa and the distance as the ordinate. Figure 5-1 to Figure 5-8 respectively show the distance from obstacles in sectors 0-7 collected by lidar SF40 to the UAV. It can be seen from Figure 5 that the maximum distance in the data collected by the lidar is 20m, but the minimum distance in the data is far less than 5m. This is because the wave speed of the omnidirectional lidar is close to the frequency of sunlight, which is easy to be interfered by sunlight. Thus, erroneous data is output. Incorrect data will affect the UAV's obstacle avoidance flight. Therefore, it is necessary to optimize the data collected by the lidar and filter out the wrong data, so that the UAV can avoid obstacles accurately.

The obstacle avoidance algorithm for the octagonal air wall is: adding a median filter, a low-pass filter and a sliding window average filter successively to filter the data of the eight sectors, which solves the problem of the existence of data in the eight sectors. Outliers problem.

The specific optimization method for the octagonal air wall obstacle avoidance algorithm in the present invention is as follows: the wave speed of the omnidirectional laser radar is close to the frequency of sunlight, which is easy to be interfered by sunlight, thereby outputting wrong data. Incorrect data will affect the UAV's obstacle avoidance flight. Therefore, it is necessary to optimize the data collected by the lidar and filter out the wrong data, so that the UAV can avoid obstacles accurately. The present invention successively adopts median filter, low-pass filter and sliding window average filter for data processing. The process of filtering the original data is shown in Figure 6.

The distance data collected by the optimized octagonal air wall obstacle avoidance algorithm is shown in Figure 7. Among them, Figure 7-1 to Figure 7-8 respectively represent the distance from the obstacles in the 0-7 sectors collected by the optimized lidar SF40 to The distance of the drone. As can be seen from the comparison of Fig. 5 and Fig. 7, after the distance data of 8 sectors is processed by median filter, low-pass filter and sliding window average filter, the data whose distance is far less than 5m is filtered out, which illustrates the present invention. The adopted optimization algorithm can effectively remove noise data, so as to output accurate distance data, and provide data support for the UAV to achieve accurate obstacle avoidance.

The software flow design of the obstacle avoidance system of the rotor UAV in the present invention is shown in FIG. 8 . After the UAV flight control system is powered on and running, it reads the distance information from the UAV to the obstacle and the input remote control signal cyclically. When the distance information is greater than the set safe distance, the attitude calculation is performed on the input remote control signal, and the drone is controlled to fly normally according to the remote control signal. When the distance information is less than the set safe distance, the attitude calculation is performed on the input remote control signal: if the remote control signal requires the drone to stay away from obstacles, the drone will fly normally according to the remote control signal; if the remote control signal requires the drone to approach If there is an obstacle, the attitude of the UAV is calculated according to the preset safe distance, and the UAV is controlled to stay away from the obstacle. When the drone returns to a safe distance, it continues to calculate the attitude according to the input signal of the remote controller to control the flight of the drone. This program is constantly looping, constantly judging.

As shown in Figure 9, in which: Figure 9-1 is the obstacle distance facing the UAV, Figure 9-2 is the input value of the pitch channel of the remote control, and Figure 9-3 is the change diagram of the actual pitch angle. From Figures 9-1, 9-2 and 9-3, it can be obtained that the actual pitch angle changes with the distance of the obstacle in front of the UAV and the change of the remote control input. The forward acceleration of the drone is adjusted by the change of the pitch angle, so that the drone can adjust the speed and achieve the effect of avoiding obstacles.

The invention analyzes the reason for the existence of outliers in the distance data collected by the laser radar, that is, the laser radar wave speed is close to the frequency of sunlight, which is easy to be interfered by sunlight, thereby outputting wrong data; and a method for filtering the collected original distance data is proposed. , that is, adding median filter, low-pass filter and sliding window average filter successively to filter the original distance data, which solves the problem of outliers in the data collected by lidar.

Claims (10)

1. The utility model provides an unmanned aerial vehicle keeps away barrier system based on laser radar and octagon air wall algorithm, characterized by includes that the power execution module, microprocessor module, sensor module, attitude angle controller, laser radar module, the octagon air wall that link to each other with the signal of telecommunication keep away the barrier algorithm, wherein: the power execution module provides power for the unmanned aerial vehicle; the microprocessor module receives the data sent by the sensor module and the instruction sent by the remote controller, makes logic judgment on the two kinds of information and controls the unmanned aerial vehicle to fly in a navigation mode; the sensor module is used for resolving the current attitude angle of the unmanned aerial vehicle; the attitude angle controller controls the unmanned aerial vehicle at an actual attitude angle by receiving a remote controller signal; the laser radar module acquires the distance and the relative angle from the obstacle to the unmanned aerial vehicle; the distance data and the angle data collected by the laser radar are processed by the octagonal air wall obstacle avoidance algorithm, accurate obstacle avoidance information is provided, and accurate obstacle avoidance of the unmanned aerial vehicle is achieved.

2. The unmanned aerial vehicle obstacle avoidance system of claim 1, characterized by: the power execution module consists of an electronic speed regulator, a propeller and a direct current brushless motor; the propeller is arranged on the direct current brushless motor, and the motor rotates to drive the propeller to rotate so as to provide lift force for the unmanned aerial vehicle; the electronic speed regulator is arranged between the microprocessor and the DC brushless motor, so that the microprocessor can output a PWM signal with enough magnitude to drive the DC brushless motor.

3. The unmanned aerial vehicle obstacle avoidance system of claim 1, characterized by: the microprocessor module adopts a microprocessor chip STM32F427VIT6, and the abundant peripheral interfaces of the microprocessor module can communicate with sensors and data storage chips adopting different communication modes.

4. The unmanned aerial vehicle obstacle avoidance system of claim 1, wherein the sensor modules comprise a three-axis accelerometer, a three-axis magnetometer, and a three-axis gyroscope, all mounted on the pixawk flight control panel, wherein: the three-axis accelerometer is used for measuring the acceleration of three axes when the unmanned aerial vehicle flies; the three-axis magnetometer is used for measuring the magnetic field intensity of three axes when the unmanned aerial vehicle flies; the three-axis gyroscope is used for measuring the angular velocity around three axes when the unmanned aerial vehicle flies.

5. The unmanned aerial vehicle obstacle avoidance system of claim 4, wherein six-axis sensors ICM-20608-G and a magnetometer HMC5983 are adopted, the six-axis sensors ICM-20608-G integrate two sensors, namely a three-axis accelerometer and a three-axis gyroscope, and the pitch angle and the roll angle of the unmanned aerial vehicle are calculated through data collected by the three-axis accelerometer; the yaw angle of the unmanned aerial vehicle is solved through data collected by the magnetometer HMC5983, and therefore attitude angle information of the current unmanned aerial vehicle during flying is obtained.

6. The unmanned aerial vehicle obstacle avoidance system of claim 1, characterized by: the attitude angle controller adopts a double closed-loop PID controller, the outer ring of the attitude angle controller is designed into an angle ring, the input of the angle ring is the deviation amount of an expected attitude angle and an actual attitude angle, and the output is an expected angular velocity; the inner ring is designed into an angular velocity ring, the input of the angular velocity ring is the deviation amount of the expected angular velocity and the actual angular velocity, and the output is expected torque, so that the flying attitude of the unmanned aerial vehicle is controlled.

7. The unmanned aerial vehicle obstacle avoidance system of claim 6, wherein: in the design of the outer ring, the expected attitude angle is the attitude angle input by the remote controller, and the actual attitude angle is the attitude angle calculated by the inertial unit.

8. The unmanned aerial vehicle obstacle avoidance system of claim 6, wherein: in the inner ring design, the desired angular velocity is the output of the angle ring, and the actual angular velocity is the measurement of the gyroscope.

9. The unmanned aerial vehicle obstacle avoidance system of claim 1, characterized by: the laser radar module adopts omnidirectional laser radar SF40, and it gathers the distance and the relative angle of barrier to unmanned aerial vehicle, keeps away the barrier for follow-up unmanned aerial vehicle and provides initial data support.

10. The unmanned aerial vehicle obstacle avoidance system of claim 1, wherein the octagonal air wall obstacle avoidance algorithm is: collecting and combining raw range measurements by a lidar module, including storing and using the nearest range and angle measurements in 8 sectors; the width of each sector is 45 degrees, the 0 sector points to the front, the 1 sector points to the right, and so on; from these distances and angles, a fence, i.e. a two-dimensional vector array, is established around the drone, with the fence points falling on the straight line between the zones, keeping a certain distance; when no obstacle is detected by all 8 sectors, the available empty sectors will be filled with the distance from the adjacent sectors, forming a "cup" fence to protect the drone from hitting the target.

Images