WO2022082843A1

WO2022082843A1 - Multi-sensor integrated unmanned vehicle detection and obstacle avoidance system and obstacle avoidance method

Info

Publication number: WO2022082843A1
Application number: PCT/CN2020/124842
Authority: WO
Inventors: 李华京; 陈禹伸
Original assignee: 垒途智能教科技术研究院江苏有限公司; 英智医疗科技南京有限公司; 贝利尔科技发展南京有限公司
Priority date: 2020-10-19
Filing date: 2020-10-29
Publication date: 2022-04-28
Also published as: CN112180941A

Abstract

Provided are a multi-sensor integrated unmanned vehicle detection and obstacle avoidance system and obstacle avoidance method; the obstacle avoidance system comprises a plurality of single-line lidars, dual CCD cameras, microwave radar, front and rear blind-spot ultrasonic sensor sets, and a triple-core controller based on ARM+FPGA+NUC; the single-line lidar signals are processed by an NUC microcomputer; the binocular vision graphic data from the CCD cameras is processed by the ARM+FPGA controller together, and blind-spot detection and obstacle avoidance, human-computer interface, path planning, online output, and other functions are left to the STM32F767 to perform alone; the ARM+FPGA controller precisely controls a DC brushless servo motor by means of decoding the output control signal, driving the unmanned vehicle. Thus the unmanned vehicle is capable of detecting obstacles in complex environments from a greater distance in any weather and quickly performing effective obstacle avoidance, thereby improving the safety and stability of the unmanned vehicle when traveling at high speed.

Description

A multi-sensor fusion unmanned vehicle detection and obstacle avoidance system and obstacle avoidance method

technical field

The invention belongs to the technical field of unmanned driving, and in particular relates to a multi-sensor fusion unmanned vehicle detection and obstacle avoidance system and an obstacle avoidance method.

Background technique

With the rapid economic development, automobiles have become an increasingly important part of people's lives. The negligence of drivers can lead to many accidents. Therefore, car manufacturers focus on designing systems that can ensure car safety. Safety is one of the main factors driving the demand for driverless cars; Traffic jams make driving not so good, and artificial intelligence-based unmanned vehicles can completely solve problems such as traffic jams instead of human drivers; in addition, poor air conditions are also a "catalyst" for the promotion of driverless cars.

A driverless car is a smart car that senses the road environment through the on-board sensing system, automatically plans the driving route and controls the vehicle to reach the predetermined target. It uses on-board sensors to perceive the surrounding environment of the vehicle, and controls the steering and speed of the vehicle according to the road, vehicle position and obstacle information obtained by the perception, so that the vehicle can drive on the road safely and reliably. The driverless car integrates many technologies such as automatic control, architecture, artificial intelligence, and visual computing. It is the product of the highly developed computer science, pattern recognition and intelligent control technology. It is also an important indicator of a country's scientific research strength and industrial level. , has broad application prospects in the fields of national defense and national economy.

At present, the development of unmanned vehicles is still in its infancy, and countries have successively started research on intelligent unmanned vehicles. No matter what level of intelligent driving, the first step is to perceive, that is, to perceive the complex road conditions and environment around the vehicle. Only on this basis can the corresponding path planning and driving behavior decisions be made. The choice of perception sensor is the unmanned vehicle. A prerequisite for successful obstacle avoidance. Commonly used ranging sensors include: ultrasonic ranging sensors, infrared ranging sensors, CCD vision systems, millimeter-wave radar, microwave radar, and lidar, etc.

Lidar is actually a radar that works in the optical band (special band). Lidar is active detection and does not depend on external lighting conditions or the radiation characteristics of the target itself. It only needs to emit its own laser beam and emit laser light through detection. The echo signal of the beam is used to obtain target information. The laser wavelength is short, it can emit a laser beam with a very small divergence angle, and the multi-path effect is small, and it can detect low-altitude/ultra-low-altitude targets. Single-line lidar is a kind of lidar. Since there is only one transmission and one reception, the structure is relatively simple and the use is relatively convenient; the scanning period of single-line lidar is short, the scanning speed of the forward direction environment is fast, and the angle resolution is high , the radar itself is small in size, relatively light in weight, relatively low in power consumption, high in reliability, and relatively low in cost; single-line lidar has a relatively wide detection range and can provide a large amount of environmental scanning point distance information, which can be used for control decisions. Greater convenience and the above advantages make single-line lidar a preferred choice for unmanned vehicles to perceive unknown environments.

The general structure of a simple unmanned vehicle is shown in Figure 1, and the principle of the detection and obstacle avoidance system is shown in Figure 2. The unmanned vehicle is detected by the lidar sensor detection system and sent to the PC, and then the PC is coded and processed to send control commands to the lower computer based on the single-chip microcomputer, and the single-chip control module sends control commands to the DC brushless motor after communication decoding. The controller drives multiple DC brushless motors to move; the single-chip control system adjusts the speed of the motors according to changes in the surrounding environment, and then controls the position of the unmanned vehicle in the actual environment to realize the operation of the unmanned vehicle in the actual working conditions. For walking and obstacle avoidance, the existing simple unmanned vehicle control systems are all implemented by a single single-chip microcomputer controlling a single single-line lidar sensor or a multi-line lidar sensor to achieve the above functions. Existing unmanned vehicles have many problems in long-term operation, mainly including:

(1) Since the unmanned vehicle is disturbed by the unstable factors of the surrounding environment, the anti-interference ability of the controller based on the single-chip microcomputer is poor, and abnormalities often occur, causing the unmanned vehicle to lose control;

(2) The existing unmanned vehicles all use low-level DSP and ARM series chips, and the maximum operating frequency is only about 100 MHz, which cannot meet the fast calculation of complex data of unmanned vehicles;

(3) Affected by the performance of the PC of the unmanned vehicle, the data collected by the sensors of the unmanned vehicle cannot be quickly calculated and stored;

(4) The data obtained by the single-line lidar is 2D data, and the information such as the height of the target cannot be distinguished. Some small objects will be ignored and eventually become obstacles. The navigation of the single-line lidar sensor has become a bottleneck in the vehicle field;

(5) A single single-line lidar cannot obtain road information, and needs to cooperate with other sensors to read and discriminate the ground information;

(6) Although the multi-line laser radar can realize 2.5D or 3D data, judge the height of obstacles, and process the information on the ground, etc., the price is relatively expensive. The price of a 64-beam laser radar is as high as 700,000 yuan. Widespread use;

(7) A single single-line lidar cannot detect information such as corners, road cliffs, etc., and needs to be used with other sensors to read the surrounding obstacle signals or positioning sensor signs;

(8) The existing unmanned vehicles basically only consider forward detection and obstacle avoidance, and do not consider the information of obstacles in the rear. achieve accelerated avoidance;

(9) Based on a single single-line LiDAR unmanned vehicle, there is a detection blind spot at the moment of startup. Once an obstacle is in the blind spot, it is easy to cause traffic accidents;

(10) A single-line lidar-based unmanned vehicle will also have a detection blind spot during the actual driving process, and once an obstacle enters the movement blind spot during the movement, a traffic accident will also occur;

(11) The unmanned vehicle based on the single-line lidar has a slow acquisition speed of the road image ahead, which affects the rapid travel of the unmanned vehicle;

(12) In long-distance driving, the unmanned vehicle based on single-line lidar has poor recognition of the surrounding environment and cannot achieve precise positioning;

(13) In regular traffic, there are various traffic signs on the ground of the driving path of the unmanned vehicle, but the single-line lidar cannot be recognized, and the auxiliary navigation when the unmanned vehicle is traveling quickly is lost;

(14) In regular traffic, there are traffic lights and other signs in the air of the driving path of the unmanned vehicle, but the single-line lidar cannot be recognized, which weakens the safety of the unmanned vehicle when it travels fast;

(15) Affected by the price and performance of lidars, the detection range of lidars with relatively high cost performance is generally less than 100 meters, which is not conducive to the judgment of unmanned vehicles to quickly travel obstacles.

The principle and structure of the visual sensor are similar to the human sensory organization, and the visual sensor has the advantages of small size, low cost, convenient installation, good concealment, wide detection range and large amount of information. Adding a camera to the unmanned vehicle environment detection system can sense the surrounding environment in real time, collect data, and identify, detect and track static and dynamic objects. The controller senses possible dangers, effectively increasing the comfort and safety of car driving; the disadvantages of lidar and vision sensors are:

(1) Affected by the defects of monocular vision itself, the sensor detection system of the unmanned vehicle is improved compared with the single-line lidar, but this distance is not conducive to the high-speed driving of the unmanned vehicle;

(2) CCD-based monocular vision obstacle identification requires a feature library with a large data capacity. Once an object has no feature library data to match with it, it will cause obstacles to be undistinguished and thus cannot accurately estimate the target's characteristics. The distance is not conducive to the high-speed driving of unmanned vehicles;

(3) Whether it is a single-line lidar, a multi-line lidar or a camera, it is very sensitive to the weather with rain, fog, dust, and smoke. Weather, lidar and video signal performance will be greatly reduced, which will have a greater impact on the safety of unmanned vehicles;

(4) Whether it is a single-line laser radar, a multi-line laser radar or a camera, it is very sensitive to the weather with strong light. The strong sunlight can sometimes greatly reduce the performance of the laser radar and the camera, and sometimes there is no signal output. The safety of the vehicle is greatly affected.

Microwaves are radio waves with very short wavelengths. The directionality of microwaves is very good, and the speed is equal to the speed of light. When the microwaves encounter obstacles, they are immediately reflected back and can be received by the radar meter. The microwave radar measures the distance of obstacles according to the round-trip time of electromagnetic waves. Compared with optical guidance such as infrared and laser, microwave has strong ability to penetrate fog, smoke and dust, and has the characteristics of all-weather (except heavy rain) all day.

SUMMARY OF THE INVENTION

In view of the deficiencies in the prior art, the present invention provides a multi-sensor fusion unmanned vehicle detection and obstacle avoidance system and obstacle avoidance method, so that the unmanned vehicle can find obstacles in a complex environment from a distance all-weather and quickly realize effective Avoid obstacles, thereby improving the safety of unmanned vehicles when driving at high speed.

The present invention achieves the above technical purpose through the following technical means.

A multi-sensor fusion method for unmanned vehicle detection and obstacle avoidance, specifically:

The on-board computer NUC retrieves the driving path and navigation map information of the unmanned vehicle through the control station, and the ARM+FPGA controller determines that the unmanned vehicle starts to accelerate automatically when the blind spot ultrasonic sensor group determines that there are no obstacles in the blind spot.

At the moment when the unmanned vehicle starts, it enters the working condition selection mode according to the weather conditions: if the weather is good, the microwave radar, single-line lidar sensor and CCD camera all work, and the CCD camera and microwave radar transmit long-distance obstacle information to the ARM+FPGA controller , the single-line lidar sensor transmits short-range obstacle information to the NUC. After processing, the obstacle information is used as the feedback distance signal for the autonomous navigation of the unmanned vehicle; if the weather is bad, only the microwave radar works, and the unmanned vehicle is based on the feedback distance signal. Start to slow down and navigate autonomously;

After the unmanned vehicle enters the movement route, if the weather is good, the ARM+FPGA will adjust the position and posture of the unmanned vehicle before normal driving according to the received road marking points, drive normally according to the vehicle map information, and approach the forward direction of the long-distance obstacle. And implement obstacle avoidance; if the weather is bad, the unmanned vehicle will avoid obstacles at medium and long distances.

Further, when the unmanned vehicle approaches a long-distance obstacle, the microwave radar at the front and the top of the unmanned vehicle cooperates with the single-line laser radar to avoid obstacles. Specifically, the microwave radar MR1 and the single-line laser radar L1 cooperate to detect the front. For the undulations of the road, MR1 first conducts medium and long-distance detection, and the single-line lidar L1 further confirms the depth and width of the undulations; the microwave radars MR1 and MR2 cooperate with the single-line lidars L1 and L3 to determine whether there are obstacles in front of them: microwave radar MR2 first conducts medium and long-distance detection, and after finding suspected obstacles, it is confirmed by microwave radar MR1. The suspected obstacles are roughly determined, and then single-line laser radar L1 and L3 are used for precise position confirmation; microwave radar MR1, MR2 and single-line laser radar L3 , L2 cooperate to determine whether there is an obstacle in the front left: microwave radar MR2 first conducts medium and long-distance detection, after finding a suspected obstacle, microwave radar MR1 conducts a second confirmation, and the suspected obstacle is roughly determined, and then single-line laser radar L3, L2. Accurate location confirmation; microwave radar MR1, MR2 cooperate with single-line lidar L3, L4 to determine whether there is an obstacle in the front right: microwave radar MR2 first performs medium and long-distance detection, and microwave radar MR1 performs secondary confirmation after finding a suspected obstacle. After the obstacles are roughly determined, the single-line lidars L3 and L4 are used for precise location confirmation.

Further, when the single-line lidar L3 detects the obstacle information, it enters the single-line lidar precise positioning and navigation mode:

If the single-line lidar L3 and L1 detect that there are undulations on the road ahead, if the height and width of the undulations exceed the requirements for the unmanned vehicle to cross, the unmanned vehicle will perform forward avoidance protection; Within the range, it will drive at the set normal speed;

If the single-line lidar L1 and L3 detect that there is an obstacle in the moving path ahead, the unmanned vehicle will perform emergency obstacle avoidance to the left or right; if there is no obstacle, the unmanned vehicle will accelerate to the set normal speed. ;

If the single-line lidar L2 and L3 detect an obstacle in the left front motion path, the unmanned vehicle will perform an emergency obstacle avoidance to the right; if there is no obstacle, the unmanned vehicle will accelerate to the set normal speed;

If the single-line lidar L4 and L3 detect an obstacle in the right front motion path, the unmanned vehicle will perform an emergency obstacle avoidance to the left; if there is no obstacle, the unmanned vehicle will accelerate to the set normal speed.

Further, after the unmanned vehicle enters the movement route, the single-line laser radar and microwave radar behind the unmanned vehicle always detect the environment behind it, and if it is judged that there is an obstacle in the rear approaching the unmanned vehicle, the rear obstacle avoidance protection is performed.

Further, after the unmanned vehicle enters the moving route, the front blind area ultrasonic sensor group and the rear blind area ultrasonic sensor group always detect the environment of the blind area, and if it is judged that a temporary obstacle is approaching the blind area of the unmanned vehicle, the blind area obstacle avoidance protection is performed.

Further, when the unmanned vehicle is running normally, the CCD camera reads various navigation signs on both sides of the moving direction, and after processing by the ARM+FPGA controller, it is used as the navigation signs for the high-speed unmanned vehicle running.

Further, when the unmanned vehicle is running normally, the CCD camera reads the sign of the site where the unmanned vehicle arrives, so as to realize the automatic walking, position tracking and scheduling of the unmanned vehicle.

A multi-sensor fusion detection and obstacle avoidance system for unmanned vehicles, including multiple single-line laser radars, dual CCD cameras, microwave radar, front blind area ultrasonic sensor group, rear blind area ultrasonic sensor group and three-core control based on ARM+FPGA+NUC Multiple single-line lidars communicate with NUC, dual CCD cameras, microwave radar, front blind area ultrasonic sensor group, rear blind area ultrasonic sensor group all communicate with ARM+FPGA controller, and NUC communicates with ARM+FPGA controller.

In the above technical solution, the plurality of single-line laser radars include a single-line laser radar L1 arranged on the roof of the unmanned vehicle and having an included angle with the horizontal plane of α, where α is 5 to 15°.

In the above technical solution, the single-line laser radar further includes a single-line laser radar group arranged in front of the unmanned vehicle and a single-line laser radar group arranged behind the unmanned vehicle, and microwave radars are arranged between the single-line laser radar groups.

The beneficial effects that the present invention has are:

(1) The data fusion of multiple single-line lidars of unmanned vehicles is processed by the NUC, which makes the control relatively simple, greatly improves the operation speed, solves the bottleneck of slow single ARM soft operation, shortens the development cycle, and the program can be Strong transplant ability.

(2) The present invention completely realizes the single-board control of the unmanned vehicle, saves the space occupied by the control board, and also realizes the effective detection and obstacle avoidance of multiple independent areas of the unmanned vehicle, which is beneficial to improve the stability of the unmanned vehicle system performance and dynamic performance.

(3) Since the controller in the present invention uses NUC to process the data and algorithms of multiple single-line lidar sensors, and fully considers the surrounding interference sources, the ARM is relieved from the heavy workload and effectively prevents the main motion control. The program's "running flight" greatly enhances the anti-interference ability of unmanned vehicles.

(4) The synchronous acquisition system inside the FPGA of the present invention ensures the synchronization of the data acquisition of the dual CCD cameras, and ensures the accuracy of the subsequent distance calculation.

(5) Since the controller in the present invention uses FPGA to process a large amount of binocular vision image data, and fully considers the surrounding interference sources, the ARM is freed from the heavy image processing work, which not only improves the operation speed, but also effectively The "runaway" of the motion control main program is prevented, and the anti-interference ability of the unmanned vehicle is greatly enhanced.

(6) The image data collection of the CCD camera in the present invention is farther than the detection distance of the economical and practical single-line laser radar, so that the obstacle detection range of the unmanned vehicle is wider, and the existence of the microwave radar fills the detection distance of the CCD camera and the single-line laser radar. A blank area between is conducive to the tracking of obstacles and the determination of the distance, which is conducive to the acceleration and deceleration of the unmanned vehicle, and improves the dynamic performance of the unmanned vehicle.

(7) In the present invention, binocular vision has no restriction on the recognition rate, and the parallax principle is used to directly measure objects for all obstacles, and the measurement accuracy is higher than that of monocular vision, so that the distance from the obstacle can be more accurately estimated and realized in advance. Obstacle avoidance warning.

(8) The CCD camera in the present invention can effectively detect the obstacles protruding from the ground around the running direction of the high-speed unmanned vehicle, which can not only improve the accuracy of obstacle avoidance, but also provide accurate positioning for the unmanned vehicle navigation. .

(9) The CCD camera in the present invention can effectively distinguish road signs such as lane detection lines, straight driving and turning in regular traffic, and the unmanned vehicle can rely on these signs to correct its own position and attitude, which improves the free running time of the unmanned vehicle. The stability and accuracy of autonomous navigation.

(10) The CCD camera in the present invention can effectively distinguish traffic prompts such as green light, yellow light and red light in regular traffic, and the unmanned vehicle can adjust its own speed according to these information to meet the needs of fast driving, fast parking, etc. Safety and stability of unmanned vehicles when they travel freely.

(11) In the present invention, the single-line laser radar L1 has a certain angle with the ground. This angle can help the single-line laser radar L1 to further accurately locate the undulations of the moving road surface found by the CCD camera, and prevent the deep pit caused by road damage from affecting the unmanned vehicle. normal driving.

(12) In the present invention, the single-line laser radar L1 has a certain angle with the ground. This angle can help the single-line laser radar L1 to further accurately locate the CCD camera to find small obstacles temporarily left on the moving road, and notify the unmanned vehicle control system twice to realize Emergency avoidance ensures the normal driving of the unmanned vehicle.

(13) In the present invention, the fusion of multiple single-line laser radars and ultrasonic sensors in the front can accurately locate the location of the obstacles found by the CCD camera, and notify the unmanned vehicle control system twice to achieve avoidance, which is beneficial to improve the speed of unmanned vehicles. and security.

(14) In the present invention, a plurality of single-line laser radars and microwave radars are integrated in the front. Since the directions of the single-line laser radar and the microwave radar are intersected, the cylindrical objects on both sides found by the CCD camera can be accurately detected, which provides certain conditions for the forward positioning of the unmanned vehicle. s help.

(15) In the present invention, multiple single-line laser radars and microwave radars are integrated in the front. Since the directions of the single-line laser radar and the microwave radar are intersected, the free areas on both sides found by the CCD camera can be accurately detected, which is the best way for the unmanned vehicle to turn and avoid. provide some assistance.

(16) In the present invention, the fusion of multiple single-line laser radars and microwave radars in the rear can effectively detect the distance between the unmanned vehicle and the moving obstacles behind. When encountering an emergency, the unmanned vehicle can speed up and escape with the help of the controller. Dangerous areas, play the role of protecting the unmanned vehicle body.

(17) The front blind spot detection and obstacle avoidance system in the present invention can effectively eliminate the close blind spot that occurs when the unmanned vehicle just starts to accelerate forward, and improves the safety and reliability of the unmanned vehicle when it starts to accelerate forward; It can effectively eliminate the short-range blind spots that appear in real time when the unmanned vehicle is driving normally, and further improve the safety and reliability of the unmanned vehicle.

(18) The rear blind spot detection and obstacle avoidance system in the present invention can effectively eliminate the short-range blind spot that occurs when the unmanned vehicle just starts reversing, thereby improving the safety and reliability of the unmanned vehicle when reversing; and can effectively eliminate the unmanned vehicle The short-range blind spot that appears in real time when the car is reversing further improves the safety and reliability of the unmanned vehicle.

(19) In a weather condition with a lot of rain, fog, smoke or dust, the microwave radar in the present invention is activated to detect the forward environment at long and medium distances, and the microwave radar is used for navigation under the condition that the laser radar is interfered, which is beneficial to Improve the safety of unmanned vehicles in harsh environments.

(20) For the unmanned vehicle in the present invention, in order to meet the large-scale multi-site operation, a site sensor with a certain degree of redundancy is added, which is not only conducive to the positioning of the unmanned vehicle, but also helps the terminal to detect the unmanned vehicle. tracking.

(21) The binocular vision in the present invention does not require the ARM controller to compare a huge sample feature library with the collected images, which directly solves the identification failure caused by the lack of the data feature library, and improves the performance of the high-speed unmanned vehicle. safety.

(22) The CCD camera in the present invention can transmit on-site images to the control station through a wireless device when the unmanned vehicle encounters an emergency, and the control station prejudges and makes a plan that requires emergency treatment;

(23) During the movement process, the role of the battery in this system is fully considered. Based on the ARM+FPGA+NUC three-core controller, the running state of the unmanned vehicle is monitored and calculated at all times, avoiding the generation of large currents. Therefore, the impact of high current on the battery is fundamentally solved, and the occurrence of excessive aging of the battery caused by high current discharge is avoided.

Description of drawings

Fig. 1 is a two-dimensional structure diagram of an ordinary simple unmanned vehicle;

Figure 2 is a schematic diagram of an ordinary unmanned vehicle detection and obstacle avoidance system;

Fig. 3 is the schematic diagram of image processing of the connection between ARM and FPGA of the present invention;

Figure 4 is a two-dimensional structural diagram of a multi-sensor fusion unmanned vehicle;

Figure 5 is a two-dimensional structural diagram of the arrangement of the front multi-element radar group;

Fig. 6 is a two-dimensional structure diagram based on binocular vision CCD black and white camera arrangement;

Figure 7 is a two-dimensional structural diagram of the arrangement of ultrasonic sensor groups in the front blind area;

Figure 8 is a two-dimensional structural diagram of the arrangement of the rear multi-radar group and the rear blind area ultrasonic sensor group;

Figure 9 is a schematic diagram of a multi-sensor fusion unmanned vehicle detection and obstacle avoidance system;

Figure 10 is a schematic diagram of the operation of the multi-sensor fusion unmanned vehicle;

Figure 11 is the acceleration and deceleration curve diagram of the unmanned vehicle operation.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto.

SICK's lidar adopts the mature laser-time-of-flight principle and multiple echo technology, non-contact detection, and can set various graphic protection areas according to the needs of the scene, and can simply modify the graphics at any time according to the needs of the scene. The sensor has reliable anti-interference performance through internal filtering and multiple echo technology. LMS151 and LMS122 are new high-performance lidars from SICK, which are respectively aimed at short-range detection. The LMS151 series is aimed at objects with a 10% reflectivity, and the distance can reach 50 meters, and the LMS122 detection distance can reach 20 meters. In view of the above characteristics, the present invention uses a LMS1XXX series based lidar group to form an unmanned vehicle short-range front and rear obstacle detection and protection system: the present invention adopts a position slightly higher than the roof of the vehicle, and the included angle with the horizontal plane is α (5 ~ 15°), the LMS151-10100 single-line lidar L1 located at the center of the front of the roof, diagonally downwards (Figure 4, 5), with a set of LMS151-10100 single-line about 40cm above the ground and parallel to the horizontal plane The lidar group FLT (usually 3, L2, L3, L4, see Figures 4 and 5) constitutes an accurate forward short-range detection and obstacle avoidance system, in which L2 and L4 are located at the front left and right of the front of the vehicle, respectively. They have an included angle of approximately 30° in the center direction away from the movement direction, which can effectively detect obstacles on the left and right sides of the unmanned vehicle, respectively. L3 is located at the center of L2 and L4, and its center direction is consistent with the movement direction; The present invention adopts a group of LMS122-10100 single-line laser radar group BLTs (generally 2 pieces, L5 and L6 respectively, see Figures 4 and 8) that are about 40cm-60cm above the ground and parallel to the horizontal plane to form the rear detection and detection system of the unmanned vehicle. Protection System.

The principle of the binocular CCD camera acquisition system is similar to the human eye. The binocular vision based on the CCD camera is to add a CCD camera to the monocular vision. The binocular vision can perceive the distance of the object. The farther the object is, the smaller the parallax. On the contrary, the greater the parallax; the binocular vision directly measures all obstacles, and the measurement accuracy is higher than that of the monocular vision, and the distance is calculated directly by the parallax, without the need for a huge data sample library. Based on the above advantages, the present invention installs two (CCD1 and CCD2) CCD cameras on the front windshield of the unmanned vehicle to form binocular vision (see Figures 4 and 6) for long-distance environmental detection and obstacle avoidance.

Germany's Infineon Technologies AG specializes in the production of microwave radar for vehicles: the vehicle radar system emits radio waves, which are reflected back by vehicles or other objects ahead. Infineon's radar chip is responsible for sending and receiving these high-frequency signals, and passing them to the radar electronic control unit (ECU), which measures the distance between the car and other moving objects and their speed, for people and unmanned people. Driving provides a distance criterion; Infineon microwave radars mainly include 77GHz and 24GHz. 77GHz is the standard frequency range for radar applications such as adaptive cruise control and collision warning. Even in low visibility conditions, the 77GHz radar chip can Let the unmanned vehicle "recognize" obstacles and other road usage within a distance of 250 meters; the 24GHz radar chip can also "identify" obstacles and other road usage within a distance of 100 meters. This microwave radar is made of silicon germanium technology and The new product operating in the 24GHz ISM band (24.0GHz to 24.25GHz) is equipped with an on-chip radar transceiver with the industry's highest integration level and a receive-only auxiliary chip, enabling system design flexibility to achieve low cost and high performance design goals. The three devices of the new series are BGT24MTR11 (single transmitter and single receiver channel), BGT24MTR12 (single transmitter and double receiver channel) and BGTMR2 (dual receiver); for the reasons of cost performance, the present invention adopts BGT24MTR11 for medium and long distance detection, as shown in Figure 5 As shown in the figure, the setting position and method of the microwave radar MR1 are the same as those of the single-line laser radar L1, and its included angle with the horizontal plane is β, and β<α; Between radar L5 and single-line lidar L6 (Figure 8).

Due to the combination of sensors, the unmanned vehicle generally has a blind spot in the forward motion area when it starts to drive forward. In order to prevent collision during starting, the present invention adds a set of ultrasonic sensors US1, US2, US3 to the bottom of the unmanned vehicle. , US4, US5 composed of front blind spot detection and obstacle avoidance system (see Figure 7, that is, the front blind zone ultrasonic sensor group FBZT in Figure 4). At the moment when the unmanned vehicle starts to drive forward, the front blind spot detection system works. If there is no obstacle in the safe area when the unmanned vehicle starts to accelerate and drive forward, the unmanned vehicle will switch to the multi-sensor fusion navigation state. Due to the combination of sensors, the unmanned vehicle generally has a blind spot in the rear movement area when it starts to reverse. The rear blind spot detection and obstacle avoidance system composed of , US9, and US10 (see Figure 8, that is, the rear blind spot ultrasonic sensor group BBZT in Figure 4); at the moment when the unmanned vehicle starts to reverse, the rear blind spot detection system works. When the human vehicle starts to accelerate and reverse, there is no obstacle in the safe area, and the unmanned vehicle will enter the multi-sensor fusion navigation state.

The new STM32F7 MCU series produced by STM is the world's first mass-produced microcontroller with a 32-bit ARM Cortex-M7 processor. The products are equipped with a Cortex-M7 core with floating-point arithmetic units and DSP expansion functions. , operating speed up to 216MHz; with AXI and multi-AHB bus matrix for core, peripheral and memory interconnection, using 6-stage superscalar pipeline and Floating Point Unit (FPU); two general-purpose DMA controllers and one DMA dedicated to graphics accelerator; peripheral speed is independent of CPU speed (dual clock support), so that system clock changes do not affect peripheral work; compared to the previous STM32 series, it has more peripherals; the above excellent energy efficiency is due to ST's market-leading 90nm manufacturing process, unique flash memory access time reduction, advanced clock and power optimization techniques, and 100µA typical current in stop mode where all registers and SRAM contents are retained At the same time, STM32F7 has excellent instruction and pin compatibility: Cortex-M7 is backward compatible with Cortex-M4 instruction set, STM32F7 series is pin compatible with STM32F4 series; The advantages of Cortex-M4) are used to the extreme, and the efficiency is nearly twice that of DSP. The above characteristics make STM32F7 very suitable to replace the STM32F4 series chips for data processing of multi-sensor fusion of unmanned vehicles.

Image processing can be roughly divided into low-level processing and high-level processing. Low-level processing has a large amount of data, simple algorithms, and large parallelism; high-level processing has complex algorithms and small data volumes. In the low-level image processing stage, using software processing is a very time-consuming process, but using hardware processing, a large amount of data can be processed in parallel, which can greatly improve the processing speed. The FPGA itself is just a standard cell array and does not have the functions of a general integrated circuit, but users can recombine and connect its interior through specific placement and routing tools according to their own design needs, and design their own in the shortest time. Application-specific integrated circuits, because FPGA adopts the design idea of software to realize the design of hardware circuit, which makes the system based on FPGA design have good reusability and modification. This new design idea has been gradually applied to high-performance image Processing and rapid development.

Combining the advantages of ARM and FPGA, the binocular vision synchronous acquisition and image data segmentation processing of the present invention are handed over to the FPGA, while the object binocular visual distance calculation is handed over to the ARM. The data processing connection of the two is shown in Figure 3.

In order to overcome the shortcomings of poor stability, poor rapidity and poor cost performance of the existing unmanned vehicles, the invention abandons the single-line laser radar or multi-line laser radar working modes adopted by domestic unmanned vehicles, and proposes a microcomputer based on the seventh generation NUC. +ARM (the latest embedded STM32F767)+FPGA's new three-core control mode. In order to reduce the overall hardware cost of unmanned vehicles and improve the detection distance of unmanned vehicles, multiple single-line lidars + dual CCD cameras + ultrasonic sensor fusion technology are used to detect and avoid obstacles. The control board takes STM32F767 as the processing core, receives in real time the multi-sensor digital fusion signal of the host computer based on NUC7 (seventh generation NUC microcomputer) and the image signal collected by the CCD camera based on FPGA, and responds to various interrupts in real time to realize and control the main station data communication and storage of real-time signals.

In order to improve the computing speed and ensure the rapidity, stability and reliability of the unmanned vehicle control system, the present invention introduces FPGA and Intel's seventh-generation NUC microcomputer into the STM32F767-based controller to form a three-stage system based on ARM+FPGA+NUC. Nuclear controller, this controller integrates multiple single-line lidar detection and obstacle avoidance system controller systems, and fully considers the role of batteries in this system to realize the detection and obstacle avoidance of unmanned vehicles in various areas. The processing of multiple single-line lidar signals with the largest workload in the unmanned vehicle control system is handed over to the NUC microcomputer for processing, giving full play to the NUC microcomputer's fast data processing speed, and the binocular visual graphics data processing of the CCD camera is handed over to ARM. Co-processing with FPGA to give full play to their advantages in different stages of image processing, while blind spot detection and obstacle avoidance, human-machine interface, path planning, online output and other functions are handed over to STM32F767 to complete independently, thus realizing ARM, FPGA, NUC micro The division of labor between the computers, and real-time communication between the three for data exchange and calling.

For the three-core controller based on ARM+NUC+FPGA designed by the present invention, when the power is turned on, the ARM, FPGA and NUC first complete initialization, and then the vehicle-mounted computer NUC retrieves the driving path and map of the unmanned vehicle through the unmanned vehicle control station information, and then the blind spot ultrasonic sensor group, CCD-based binocular vision, microwave radar and multiple single-line lidars start to work. The ARM+FPGA controller determines that there are no obstacles entering the working area and then turns on the unmanned vehicle walking mode, and calculates the dual It collects the image data and microwave radar data of the CCD camera, and communicates with the NUC at the same time. The NUC receives and decodes multiple single-line lidar feedback signals in real time, and then communicates with the ARM+FPGA controller and transmits control signals to the ARM+FPGA controller. The ARM+FPGA controller precisely controls the DC brushless servo motor by decoding the output control signal. The DC brushless servo motor drives the unmanned vehicle to drive after the mechanical device transforms the power, and feeds back signals such as displacement, speed and acceleration to the ARM controller in real time.

Referring to Figure 9, the unmanned vehicle control system is divided into two parts: the upper computer system based on the on-board computer NUC and the ARM+FPGA lower computer system based on STM32F767. Among them, the NUC host computer system based on the on-board computer completes the functions of path and map input, multi-sensor data fusion and online output; based on the ARM+FPGA host computer control system, the unmanned vehicle system servo control, CCD binocular vision data processing and Microwave radar ranging, I/O control and other functions, among which the multi-axis DC brushless servo system control with the largest workload, CCD-based binocular vision data processing and microwave radar ranging are jointly processed by ARM+FPGA, giving full play to ARM+ The advantages of FPGA's respective data processing, thus realizing the division of labor between NUC and ARM+FPGA, at the same time, communication between the three can be carried out, and data exchange and call can be carried out in real time.

The specific implementation process is as follows:

1) Before the unmanned vehicle receives the motion command, it generally waits in the waiting area for the start command issued by the control station. If the voltage is low, the unmanned vehicle will automatically dock with the charging device for charging.

2) Once the unmanned vehicle receives the departure task during the waiting period, the on-board computer NUC retrieves the driving path and navigation map information of the unmanned vehicle through the control station, and then the ARM+FPGA controller turns on the front blind spot ultrasonic sensors US1~US5 to detect the front blind spot. During scanning, if there is an obstacle entering the blind spot, the ARM+FPGA controller will issue an alarm and wait for the obstacle to be cleared; if there is no obstacle entering the blind spot, the unmanned vehicle will start to accelerate automatically.

3) At the moment when the unmanned vehicle starts to start, enter the working condition selection mode according to the weather conditions: if the weather is good, turn on the microwave radar MR1~MR3, the single-line lidar sensors L1~L6 and the CCD camera. The CCD camera and MR1~MR3 start to ARM+ The FPGA controller transmits long-distance obstacle information, and at the same time, multiple single-line lidars begin to transmit short-range obstacle information to the NUC. The ARM+FPGA controller and NUC begin to decode the obstacle information and convert it into the distance between the obstacle and the unmanned vehicle. Signals and communicate with each other, the unmanned vehicle starts autonomous navigation with the help of these feedback distance signals, and starts to drive along the specified route; if the weather is bad, the ARM+FPGA controller will prohibit the binocular CCD camera and the single-line lidar sensor L1~L6 from working, only Turn on the microwave radars MR1~MR3, and MR1~MR3 begin to transmit the information of mid- and long-distance obstacles to the ARM+FPGA-based controller. At the same time, the ARM+FPGA controller and the NUC communicate with each other, and the unmanned vehicle starts to decelerate autonomously with the help of MR1~MR3 feedback distance signals. Navigate and start driving along the prescribed route.

4) After the unmanned vehicle enters the movement route, if the weather is good, the first task of the CCD camera is to find the road marking point based on the existing road map information: this marking point may be the end of the road or the turning point , or some docking sites, the CCD camera communicates with the FPGA after finding these marking points, the FPGA decodes the image of the CCD camera and then communicates with the ARM, the ARM continues to process the marking point data information according to the internal algorithm, and then converts it into DC brushless The PWM control signal of the servo motor drives the unmanned vehicle to perform positioning and attitude adjustment before normal driving; after the unmanned vehicle combines the CCD camera image acquisition to complete the positioning and attitude adjustment, it will drive normally according to the on-board map information, and the CCD camera will drive normally during the driving process. The real-time image is transmitted to the FPGA through the FPGA synchronous acquisition system, and the FPGA decodes and processes the data and transmits the data to the ARM. The ARM controller uses the internal algorithm to convert the real-time decoded image data into the distance between the unmanned vehicle and the obstacle, and then the ARM+FPGA controller starts Fine-tune the control of the brushless DC servo motor, so that the unmanned vehicle starts to navigate autonomously and approach and avoid obstacles in the forward direction away from obstacles. If the weather is bad, the CCD camera will be prohibited from working, and the unmanned vehicle will abandon the long-distance obstacle avoidance mode, and can only rely on the microwave radar MR1 to MR3 to first achieve medium and long distance obstacle avoidance.

When the unmanned vehicle approaches a long-distance obstacle, the mid- and long-distance detection microwave radars MR1 and MR2 cooperate with the front short-range detection single-line lidar L1 and L2, L3, and L4 to detect the environment ahead, and the unmanned vehicle starts to implement obstacle avoidance. L1 and MR1 can work independently. Due to their certain inclination angle, MR1 and L1 can cooperate well to detect the ups and downs of the road ahead. First, MR1 detects the medium and long distances, and then L1 performs further accurate confirmation. MR1 and L1 are very easy Find the depth and width of the undulations; MR1, MR2, L1 and L3 cooperate to mainly detect the presence of obstacles in front of them: MR2 first detects the medium and long distances, and MR1 conducts secondary confirmation after the suspected obstacles are found. The suspected obstacles are roughly After confirmation, there are L1 and L3 for accurate position confirmation; MR1, MR2, L3 and L2 cooperate to detect the existence of obstacles in the left front: firstly, MR2 detects the medium and long distance, and after finding the suspected obstacles, MR1 conducts secondary confirmation, The suspected obstacle is roughly determined, and then L3 and L2 confirm the precise position; MR1, MR2, L3, and L4 detect the existence of the right front obstacle: first, MR2 detects the medium and long distance, and after finding the suspected obstacle, MR1 carries out the second step. After confirmation, the suspected obstacle is roughly determined, and then L3 and L4 confirm the precise location.

In the normal driving process, MR2 detects the medium and long distances in the driving direction at all times, and the detection signal is input to the ARM controller for decoding to obtain the approximate distance of the obstacle. Then, when the obstacle enters the detection range of MR1, the detection signals of MR2 and MR1 At the same time, it will be sent to the ARM controller for decoding, and the distance information of obstacles will be further obtained. If there is no obstacle in the direction of movement, the unmanned vehicle continues to drive at the original speed; if there is indeed a suspected obstacle, the unmanned vehicle decelerates to a low speed according to the speed control mode in Figure 11 and enters the medium and long distance detection and obstacle avoidance mode. +FPGA communicates with NUC at all times. If the sensor detects that the weather is very bad, the multiple single-line lidars will be severely interfered at this time, and the ARM+FPGA controller will continue to receive the microwave radar MR2 and MR1 signals and discard the multiple single-line lidar L1~L6 feedback signals transmitted by the NUC At the same time, the unmanned vehicle turns on the low-speed driving mode, continues to use MR2 and MR1 as forward navigation sensors to approach obstacles and relies on microwave radar signals to avoid obstacles; if the weather is good, the microwave signals and multiple single-line lidar signals are good at this time, ARM +FPGA controller will continue to receive microwave radar MR2 and MR1 signals, and the NUC will receive multiple single-line lidar L1~L6 feedback signals, and the unmanned vehicle will enter the microwave radar signal and single-line laser radar signal transfer area. Once the single-line laser radar L3 detects Obstacle information. At this time, the collected signal of the microwave radar will be used as an auxiliary signal, and the controller will enter the single-line laser radar precise positioning and navigation mode:

If L3 and L1 detect that there is a certain height of undulating pits in the forward motion path, if the height and width exceed the requirements for unmanned vehicles to cross, they will send an interrupt request to the STM32F767 in the ARM+FPGA controller and at the same time put the undulating pits. The data is transmitted to the NUC and ARM+FPGA for processing, and the STM32F767 will give priority to the interrupt and enter the front avoidance protection subroutine; if the height and width of the undulating pit are within the tolerance range of the unmanned vehicle, the unmanned vehicle will follow the set normal speed. to drive;

If L1 and L3 detect an obstacle in the forward motion path, they will send an interrupt request to the STM32F767 in the ARM+FPGA controller and transmit the obstacle data to the NUC for processing. The STM32F767 will prioritize the interrupt and enter the emergency forward obstacle avoidance Protection subroutine: STM32F767 enters the left or right emergency obstacle avoidance yield according to the data of NUC communication; if no obstacle enters the operating range, the unmanned vehicle will accelerate to the set normal speed according to the speed control mode in Figure 11 .

If L2 and L3 detect an obstacle in the left front motion path, they will send an interrupt request to the STM32F767 in the ARM+FPGA controller and transmit the obstacle data to the NUC for processing. The STM32F767 will prioritize the interrupt and enter the emergency left front avoidance. Obstacle protection subroutine: STM32F767 enters the right emergency obstacle avoidance yield according to the data of NUC communication; if no obstacle enters the operating range, the unmanned vehicle will accelerate to the set normal speed according to the speed control mode in Figure 11;

If L4 and L3 detect an obstacle in the right front motion path, they will send an interrupt request to the STM32F767 in the ARM+FPGA controller and transmit the obstacle data to the NUC for processing. The STM32F767 will prioritize the interrupt and enter the emergency right forward avoidance. Obstacle protection subroutine: STM32F767 enters the emergency obstacle avoidance to the left according to the data of NUC communication; if no obstacle enters the operating range, the unmanned vehicle will accelerate to the set normal speed according to the speed control mode in Figure 11.

5) After the unmanned vehicle enters the moving route, the single-line lidar L5, L6 and microwave radar MR3 detect the environment behind it at all times. When MR3, L5 or L6 judges that there is an obstacle in the rear approaching the unmanned vehicle, it will send the ARM+FPGA controller to the controller. The STM32F767 in the STM32F767 sends an interrupt request and transmits the obstacle data to the NUC for processing. The STM32F767 will give priority to the interrupt, and then enter the rear obstacle avoidance protection subroutine and issue an alarm; if there is no obstacle behind entering the protection range, the unmanned vehicle will follow the Drive at the set normal speed.

6) After the unmanned vehicle enters the moving route, the front blind spot ultrasonic sensors US1～US5 and the rear blind spot ultrasonic sensors US6～US10 always detect the environment of the blind spot. If US1～US5 or US6～US10 judges that there is a temporary obstacle to the unmanned vehicle blind spot When approaching, an interrupt request will be sent to the STM32F767 and the obstacle data will be transmitted to the NUC for processing. The STM32F767 will give priority to the interrupt, and the unmanned vehicle will then enter the blind spot obstacle avoidance protection subroutine and issue an alarm; if there is no obstacle in the blind spot, it enters the protection range , the unmanned vehicle will drive at the set normal speed.

7) Under the condition that the unmanned vehicle enters the track and the normal running speed meets the requirements, its single-line lidar L1~L6, ultrasonic sensor US1~US10, binocular CCD camera and microwave radar MR1~MR3 will collect peripheral signals in real time, and give feedback. The signal is sent to the NUC or ARM+FPGA controller. First, the NUC performs the laser radar data fusion processing, the FPGA performs the binocular CCD image data processing, the ARM performs the microwave radar data processing and responds to various interruption protection, and then the three communicate with each other. The STM32F767 generates a control signal to the brushless DC servo motor based on the sensor fusion decoded signal. The controller adjusts the motion of the servo motor to change the motion speed and direction of the unmanned vehicle, so that the unmanned vehicle can easily follow the on-board input path.

8) When the unmanned vehicle enters the track and runs normally, the binocular CCD camera reads various navigation signs on the ground in some key sign areas on both sides of the moving direction in real time, and then directly transmits the collected images to the FPGA, which is decoded and processed by the FPGA. Transmitted to ARM, STM32F767 will process the decoded data of these images as one of the navigation signs for tasks such as fast forwarding, parking, starting and turning of high-speed unmanned vehicles, and perform secondary attitude adjustment.

9) Because the unmanned vehicle is not a one-stop service mode in most cases, and there are many places to reach, in order to realize the site function of the unmanned vehicle, the present invention adds a ground sign to the site position, when the unmanned vehicle is about to arrive. When the station is on, the ARM+FPGA controller will read the station logo through the binocular CCD camera. When the station is read, it will automatically accumulate. In order to realize the automatic walking cycle function of the unmanned vehicle, the unmanned vehicle will reach the last station. Auto-zero and recount from station 1.

10) When the unmanned vehicle enters the docking station, the ARM controller stores and generates the entry information record table, and then sends it to the control station through the wireless device, which is beneficial to the control station's tracking of the position of the unmanned vehicle and the scheduling of the unmanned vehicle.

11) In order to meet the actual functional needs of unmanned vehicles in special situations such as scenic spots, the present invention adds a stop selection function: in the initial stage of unmanned vehicle operation, the control station can freely set the stops that unmanned vehicles need to go to, Then the unmanned vehicle can complete this setting independently by relying on its own sensors. If it encounters an emergency during operation, the control station needs to change the running path or stop site. The control master station communicates with the unmanned vehicle ARM+FPGA+NUC through a wireless device. The core controller communicates and sends the change of walking information to the on-board computer NUC through wireless. The unmanned vehicle on-board computer NUC will receive and automatically update the route and stop site information and communicate with the ARM+FPGA controller. The driver drives the unmanned vehicle according to the new requirements to complete the task.

12) When the unmanned vehicle travels along a fixed path, various sound and light alarm systems on the system will work, and it is easy to remind the surrounding pedestrians of the existence of unmanned vehicles. When the unmanned vehicle loses communication with the main station, ARM+FPGA controls The device will send an automatic stop signal to directly lock the motion servo motor of the unmanned vehicle on the spot, so that it is not easy to collide with other unmanned vehicles. At this time, since the control master station cannot collect the transmission information of the unmanned vehicle, it will A docking station information to fast track and troubleshoot problems.

The embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above-mentioned embodiments, and any obvious improvement, replacement or All modifications belong to the protection scope of the present invention.

Claims

A multi-sensor fusion method for detecting and avoiding obstacles for an unmanned vehicle, which is characterized in that:

The on-board computer NUC retrieves the driving path and navigation map information of the unmanned vehicle through the control station, and the ARM+FPGA controller determines that the unmanned vehicle starts to accelerate automatically when the blind spot ultrasonic sensor group determines that there are no obstacles in the blind spot.

The moment the unmanned vehicle starts to start, it enters the working condition selection mode according to the weather conditions: if the weather is good, the microwave radar, single-line lidar sensor and CCD camera all work, and the CCD camera and microwave radar transmit long-distance obstacle information to the ARM+FPGA controller , the single-line lidar sensor transmits short-range obstacle information to the NUC. After the obstacle information is processed, it is used as the feedback distance signal for the autonomous navigation of the unmanned vehicle; if the weather is bad, only the microwave radar works, and the unmanned vehicle is based on the feedback distance signal. Start to slow down and navigate autonomously;

After the unmanned vehicle enters the movement route, if the weather is good, the ARM+FPGA will adjust the position and posture of the unmanned vehicle before normal driving according to the received road marking points, drive normally according to the vehicle map information, and approach the forward direction of the long-distance obstacle. And implement obstacle avoidance; if the weather is bad, the unmanned vehicle will avoid obstacles at medium and long distances.
The method for detecting and avoiding obstacles for an unmanned vehicle based on multi-sensor fusion according to claim 1, wherein when the unmanned vehicle approaches a long-distance obstacle, microwave radar and single-line laser light located in front of and at the top of the unmanned vehicle Radar cooperates to avoid obstacles, specifically: microwave radar MR1 and single-line lidar L1 cooperate to detect the undulations of the road ahead, MR1 first performs medium and long-distance detection, and single-line lidar L1 further confirms accurately to determine the depth and width of the undulation; microwave radar MR1 , MR2 cooperates with single-line lidar L1, L3 to determine whether there is an obstacle in front of it: microwave radar MR2 first conducts medium and long-distance detection, and after finding a suspected obstacle, microwave radar MR1 conducts a second confirmation, and the suspected obstacle is roughly determined. Single-line lidar L1 and L3 are used for precise location confirmation; microwave radars MR1 and MR2 cooperate with single-line lidars L3 and L2 to determine whether there is an obstacle in the front left: microwave radar MR2 first performs mid- and long-distance detection, and after finding a suspected obstacle, the microwave radar MR1 is used for secondary confirmation, and the suspected obstacle is roughly determined, and then the single-line lidar L3 and L2 are used for precise position confirmation; the microwave radar MR1, MR2 and the single-line lidar L3 and L4 cooperate to determine whether there is an obstacle in the front right: microwave radar MR2 The medium and long-distance detection is carried out first, and after the suspected obstacle is found, the microwave radar MR1 will carry out the secondary confirmation.
The multi-sensor fusion method for detecting and avoiding obstacles for an unmanned vehicle according to claim 2, wherein when the single-line laser radar L3 detects the obstacle information, it enters the single-line laser radar precise positioning and navigation mode:

If the single-line lidar L3 and L1 detect that there are undulations on the road ahead, if the height and width of the undulations exceed the requirements for the unmanned vehicle to cross, the unmanned vehicle will perform forward avoidance protection; Within the range, it will drive at the set normal speed;

If the single-line lidar L1 and L3 detect that there is an obstacle in the moving path ahead, the unmanned vehicle will perform emergency obstacle avoidance to the left or right; if there is no obstacle, the unmanned vehicle will accelerate to the set normal speed. ;

If the single-line lidar L2 and L3 detect an obstacle in the left front motion path, the unmanned vehicle will make an emergency obstacle avoidance to the right; if there is no obstacle, the unmanned vehicle will accelerate to the set normal speed;

If the single-line lidar L4 and L3 detect an obstacle in the right front motion path, the unmanned vehicle will make an emergency obstacle avoidance to the left; if there is no obstacle, the unmanned vehicle will accelerate to the set normal speed.
The method for detecting and avoiding obstacles for an unmanned vehicle based on multi-sensor fusion according to claim 1, wherein after the unmanned vehicle enters the moving route, the single-line laser radar and the microwave radar behind the unmanned vehicle always detect the environment behind, If it is judged that there is an obstacle in the rear approaching the unmanned vehicle, the rear obstacle avoidance protection is performed.
The multi-sensor fusion method for detecting and avoiding obstacles for an unmanned vehicle according to claim 1, wherein after the unmanned vehicle enters the moving route, the front blind spot ultrasonic sensor group and the rear blind spot ultrasonic sensor group always detect the environment of the blind spot, If it is judged that a temporary obstacle is approaching the blind spot of the unmanned vehicle, the blind spot obstacle avoidance protection is performed.
The multi-sensor fusion method for detecting and avoiding obstacles for an unmanned vehicle according to claim 1, wherein the CCD camera reads various navigation signs on both sides of the moving direction when the unmanned vehicle is running normally, and the ARM+ After being processed by the FPGA controller, it is used as a navigation sign for high-speed unmanned vehicles.
The multi-sensor fusion method for detecting and avoiding obstacles for an unmanned vehicle according to claim 1, wherein when the unmanned vehicle is running normally, the CCD camera reads the sign of the site where the unmanned vehicle arrives, so as to realize the unmanned vehicle. Automatic walking, location tracking and scheduling of vehicles.
An obstacle avoidance system for an unmanned vehicle detection and obstacle avoidance method based on multi-sensor fusion according to any one of claims 1 to 7, characterized in that it includes a plurality of single-line laser radars, dual CCD cameras, microwave radars, front blind spots Ultrasonic sensor group, rear blind area ultrasonic sensor group and three-core controller based on ARM+FPGA+NUC, multiple single-line lidars communicate with NUC, dual CCD cameras, microwave radar, front blind area ultrasonic sensor group, rear blind area ultrasonic sensor group Both communicate with the ARM+FPGA controller, and the NUC communicates with the ARM+FPGA controller.
The multi-sensor fusion unmanned vehicle detection and obstacle avoidance system according to claim 8, wherein the plurality of single-line laser radars comprise single-line laser radars arranged on the roof of the unmanned vehicle and having an angle α with the horizontal plane L1, the α is 5-15°.
The multi-sensor fusion unmanned vehicle detection and obstacle avoidance system according to claim 9, wherein the single-line laser radar further comprises a single-line laser radar group arranged in front of the unmanned vehicle and a single-line laser radar set arranged behind the unmanned vehicle A lidar group, and microwave radars are arranged between the single-line lidar groups.