Unmanned vehicle autonomous following method and device based on multi-source perception in complex cross-country environment

Abstract

According to the unmanned vehicle autonomous following device and method based on multi-source perception, sensors such as a laser radar and a millimeter wave radar are utilized, detection and tracking of a guided vehicle are achieved through multi-source information fusion, then a feedback controller based on a fuzzy strategy is designed, speed planning, relative distance regulation and control and relative speed deviation control are conducted on an unmanned vehicle, and the unmanned vehicle autonomous following method based on multi-source perception is achieved. And the robustness and safety in a special scene are improved by using an emergency barrier stopping and response hysteresis strategy, and finally steering control of the unmanned vehicle is realized by adopting a transverse controller based on a kinematic model, so that longitudinal distance keeping and transverse deviation control of the unmanned vehicle and the guide vehicle are realized. The method can effectively solve the problem of autonomous following driving of the unmanned vehicle in typical scenes such as muddy, bumpy, curved roads and satellite denial in a complex cross-country environment.

Description

Technical Field

[0001] The present invention relates to the field of multi-agent group collaboration technology, and in particular to a method for formation driving or following driving of multiple ground unmanned platforms in an off-road environment. Background Art

[0002] The technology of multi-unmanned vehicle platooning or following driving is one of the key technologies in the field of multi-agent group collaborative technology. For complex off-road environments, it uses perception detection, motion planning and collaborative control, and adjusts the lateral and longitudinal control of each unmanned vehicle in real time to maintain the relatively stable relative position and motion state of each platoon vehicle while meeting the mission requirements and adapting to the constraints of the surrounding environment. With the realization of the multi-unmanned platform platooning function, the cost of vehicle fuel, human resources, etc. is greatly saved, and the transportation efficiency is effectively improved.

[0003] The use of the guide-follow system is an important way to achieve the following driving of unmanned vehicle formations. At present, there are two main ways to implement the guide-follow system: one is based on vehicle-to-vehicle communication (V2V) technology to achieve information sharing between the pilot vehicle and the unmanned vehicle. The unmanned vehicle records or fits the driving trajectory of the pilot vehicle according to the position, posture, speed and other information of the pilot vehicle, and realizes the lateral tracking of the vehicle through trajectory tracking and radar-based obstacle avoidance behavior, and realizes the longitudinal tracking of the vehicle through linear feedforward and feedback controllers. The vehicle following is realized by lateral and longitudinal coupled tracking. However, the vehicle following technology based on vehicle-to-vehicle communication (V2V) is overly dependent on the vehicle positioning and communication system, and has poor adaptability to complex off-road denial environments; the other is to use visual detection, radar detection or multi-source sensors through fusion detection and other methods to identify and track the guide vehicle, but the following technology based on perception detection is limited by the effective detection distance of the sensor, especially for a single sensor in typical complex off-road environments such as turning and shielding. There are many limitations, which makes the pilot vehicle target easy to lose, resulting in the failure of multi-vehicle formations. Summary of the invention

[0004] The present invention is aimed at the problem of special vehicle formation following in off-road environments. In order to solve the problems of poor environmental adaptability of unmanned vehicle following technology based on vehicle-to-vehicle communication (V2V) in electromagnetic interference sections or satellite denial conditions, and the easy loss of the leading vehicle target in following technology based on single sensor perception detection, an unmanned vehicle autonomous following method based on multi-source perception suitable for complex off-road environments is provided. This method focuses on the idea of ​​detection-tracking, and uses the fusion of millimeter-wave radar and laser radar to improve the detection accuracy of the guided vehicle; for the relative speed and position noise of the workshop, a feedback controller based on fuzzy strategy and an emergency braking algorithm are designed to achieve longitudinal control; and the lateral tracking and maintenance of the guide vehicle is achieved by designing a lateral controller based on a kinematic model.

[0005] The present disclosure provides an autonomous following method for an unmanned vehicle based on multi-source perception in a complex off-road environment, comprising the following steps:

[0006] S1, using the laser radar and millimeter wave radar installed in front of the unmanned vehicle to detect the guide vehicle as the tracking target, automatically determines the sensor with high confidence as the main detection device;

[0007] Filter the main detection sensor data to obtain the target vehicle position estimate P r and the relative speed estimation V between the unmanned vehicle and the guide vehicle r ;

[0008] S2, using feedback control based on fuzzy strategy, plans the longitudinal speed of the unmanned vehicle, and performs distance error regulation and relative speed error control to obtain the expected acceleration a des , to maintain a safe following distance L between the unmanned vehicle and the lead vehicle r ;

[0009] S3, setting emergency braking strategy to deal with emergencies during autonomous following;

[0010] S4, using the steering angle control coupled with the speed and angle deviation, implements lateral control to constrain the vehicle's steering behavior.

[0011] Furthermore, the step S1 specifically includes:

[0012] S11, placing the guide vehicle in front of the unmanned vehicle and within the detection range of the laser radar and millimeter wave radar;

[0013] S12, using the target position of the guide vehicle detected by the millimeter wave radar as the initial position of the laser radar algorithm recognition and tracking. During the recognition process, the target position in the previous frame of data in the laser point cloud is used to select the region of interest, and then cluster analysis is performed on the region, and the cluster center is used as the target position of the guide vehicle. The relative speed of the target vehicle is calculated according to the position change of the cluster center between two adjacent frames;

[0014] S13, making a confidence judgment on the millimeter wave radar detection results and the laser radar detection results:

[0015] (1) Check the θ of the millimeter wave radar target Target: If the guide vehicle target is within 10 degrees to the left and right of the unmanned vehicle, the guide vehicle target is reliable based on the tracking characteristics of the millimeter-wave radar;

[0016] Then, the Euler distance between the two detection targets from the millimeter-wave radar and the lidar is calculated. When the distance is greater than the set threshold, the lidar detection result is determined to be unreliable, and the millimeter-wave radar detection result is output. Otherwise, the lidar detection result is considered reliable, and the lidar clustering result is output due to the high detection accuracy of the lidar.

[0017] (2) If the millimeter-wave radar is unreliable, the laser radar detection result is directly output until the unreliability continues until the millimeter-wave radar detection result falls into the set target area θ target The ID of the vehicle can be used to track the target vehicle in front.

[0018] S14, using an extended Kalman filter, using the confidence judgment result as the measurement value of the Kalman filter algorithm to generate the best relative speed estimate V of the guide vehicle target r and position estimate P r .

[0019] Furthermore, the step S2 specifically includes:

[0020] S21, the relative distance and speed of the two vehicles are continuously changing, and the expected distance L between the unmanned vehicle and the guide vehicle is set des and the speed of the driverless car V b The proportional relationship:

[0021] L des =m·V b

[0022] Where m is L des and V b The ratio coefficient between

[0023] According to the uniform motion mode:

[0024]

[0025] Among them, V p is the speed of the preceding vehicle;

[0026] S22, design an acceleration gain controller based on relative distance and velocity feedback to obtain the desired acceleration A des :

[0027]

[0028] K is the acceleration gain;

[0029] The state equation of the system is obtained:

[0030]

[0031] In the state equation, X = [Vb L r ] T is the state variable, u=V p is the input variable;

[0032] Let Y = V b is the output variable, the closed-loop transfer function G(s)

[0033]

[0034] Let the output variable Y = L r , then the state function is

[0035]

[0036] Y=[0 1]·X+0·u

[0037] The closed-loop transfer function is:

[0038]

[0039] Among them, take m>0 and K>0;

[0040] S23, design a fuzzy table to replace the speed V of the unmanned vehicle b The expected distance L des The linear relationship between them, where the slope of the fuzzy table is greater than zero, to ensure that V b and L r stability; at the same time, the slope of the fuzzy curve is 2 <V b <10, gradually increase to adapt to the hysteresis characteristics of the unmanned platform response;

[0041] For example: When V b = 0.0, set L des =7.0.

[0042] Furthermore, the specific method of step S3 includes:

[0043] V b 、V r and L r Extract obvious mutation features to assist in judging the emergency braking state of the guided vehicle:

[0044] Set the collision time T c

[0045] T c =L r / abs(V r )

[0046] If V r >0, then T cis the collision time of two vehicles assuming uniform motion;

[0047] At this time, if V b = 0 means the unmanned vehicle is in a braking stop state, or T c If it is greater than the collision time threshold, the state is judged as normal driving;

[0048] If the current state is normal driving, the acceleration is set to the output acceleration calculated by the longitudinal control algorithm in step S2;

[0049] If V b is greater than the threshold, i.e. high speed state, and T c If it is less than the collision time threshold, the state is judged as emergency braking;

[0050] In case of emergency braking, the deceleration is directly set to the maximum.

[0051] Furthermore, the step S4 specifically includes:

[0052] S41, estimate P according to the target vehicle position r The lateral error X contained in r and longitudinal error Y r , calculate the angle deviation θ err :

[0053]

[0054] S42, use angle gain K ang Calculate the required steering angle θ des :

[0055] θ de s=K ang ×θ err

[0056] S43, according to the vehicle kinematics model:

[0057]

[0058] The lateral velocity of the vehicle is proportional to the actual velocity, so the vehicle velocity is used to attenuate K ang :

[0059]

[0060] Among them, K max is the maximum angular gain, V max is the maximum speed, which should be greater than the maximum achievable speed of the unmanned vehicle.

[0061] The unmanned vehicle autonomous following device based on multi-source perception in complex off-road environments based on the above method mainly includes:

[0062] The target vehicle detection and identification module includes: a millimeter wave radar and a laser radar arranged at the front of the unmanned vehicle, which are used to detect and identify the guide vehicle as the target and perform steady-state tracking;

[0063] Longitudinal control module, used for longitudinal speed planning of the unmanned vehicle, to achieve longitudinal distance maintenance with the target vehicle, relative speed error control and safety assurance;

[0064] The lateral control module is used to constrain the steering behavior of the unmanned vehicle.

[0065] Furthermore, in the target vehicle detection and identification module, the initial detection distance of the millimeter wave radar to the target vehicle is not greater than 20 meters, and the detection field of view is not less than ±10°; the laser radar adopts a three-dimensional laser radar.

[0066] Furthermore, the target vehicle detection and identification module judges the confidence of the laser radar and the millimeter wave radar, and determines the sensor with high confidence as the main detection device; filters the main detection sensor data to obtain the target vehicle position estimate and the relative speed estimate between the unmanned vehicle and the guide vehicle.

[0067] Furthermore, the longitudinal expansion module adopts feedback control based on fuzzy strategy to plan the speed of the unmanned vehicle, and performs distance error regulation and relative speed error control to generate the expected acceleration a des , in order to maintain the safe following distance L set between the unmanned vehicle and the leading vehicle r At the same time, an emergency braking strategy is set to deal with emergencies during autonomous following.

[0068] Furthermore, the lateral control module generates a steering control amount by coupling the angle deviation with the current vehicle speed.

[0069] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The target vehicle identification and detection module in the present invention is mainly based on multi-source sensor information such as laser radar and millimeter wave radar, and can realize the identification and tracking of the front vehicle target without relying on global positioning system (GNSS) data and vehicle-to-vehicle communication (V2V), thereby effectively overcoming the problem of autonomous following under complex environmental electromagnetic interference or satellite denial conditions; (2) Special off-road scenes are complex and changeable, and sensors are easily damaged. The present invention performs detection and identification based on multi-source sensor information, which can effectively avoid the problem of tracking failure caused by damage to a single sensor; (3) The use of variable slope fuzzy tracking can enable the unmanned vehicle to have better following performance when driving at low speeds, which is very helpful for target detection when following driving turns, and is effective in off-road environments with forest cover; (4) The response hysteresis judgment is set in the emergency braking strategy, which can effectively deal with detection noise and enable the guide vehicle to accelerate again without affecting the vehicle following performance; (5) The steering control amount is generated by coupling the angle deviation with the current vehicle speed, thereby enhancing the adaptability and anti-interference ability of the unmanned vehicle to the off-road environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The above and other objects, features and advantages of the present disclosure will become more apparent through a more detailed description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, wherein like reference numerals generally represent like components in exemplary embodiments of the present disclosure.

[0071] Figure 1 is a flow chart of an unmanned vehicle following method according to an exemplary embodiment of the present disclosure;

[0072] Figure 2 is an exemplary autonomous vehicle following schematic diagram;

[0073] Figure 3 is the characteristic root of the longitudinal control closed-loop system when m=3;

[0074] Figure 4 is the relationship between the expected speed and distance;

[0075] Figure 5 This is the emergency braking strategy flow chart. DETAILED DESCRIPTION

[0076] The preferred embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0077] The autonomous following behavior of unmanned vehicles targeted by the present disclosure includes at least two basic elements: the front vehicle is the leading vehicle, and the rear vehicle is equipped with various types of sensors as an autonomous following vehicle. The autonomous following method and device of unmanned vehicles based on multi-source perception in complex off-road environments provided by the present disclosure are mainly applied to the rear vehicle.

[0078] In an exemplary embodiment according to the present disclosure, as shown in the attached Figure 1 As shown in the figure, according to the execution sequence, the unmanned vehicle autonomous following device is divided into three modules: target vehicle detection and identification, longitudinal control and lateral control. The unmanned vehicle autonomous following function is realized through information sharing and continuous work of the three modules, among which:

[0079] The target vehicle detection and identification module is used to detect and identify the target vehicle and perform steady-state tracking;

[0080] The longitudinal control module is used to realize the speed planning of the unmanned vehicle, as well as the longitudinal distance maintenance with the target vehicle, relative speed error control and safety assurance;

[0081] The lateral control module is used to enable the unmanned vehicle to avoid environmental obstacles and maintain the lateral tracking accuracy of the target vehicle;

[0082] The following method mainly includes the following working steps:

[0083] a) The vehicle is detected and guided by LiDAR and millimeter-wave radar, and then the confidence level of each sensor is used to determine the detection reliability of the target vehicle, and the sensor with the highest confidence level is determined as the main detection device.

[0084] b) Use the extended Kalman filter (EKF) to filter the main detection sensor data to generate the target vehicle position P r (P r By the lateral error X r and longitudinal error Y r composition) and the relative speed V between the unmanned vehicle and the guide vehicle r and other information, and the relative speed V r and position P r Output to the follower control module.

[0085] c) Design a feedback controller based on fuzzy strategy to plan the speed of the unmanned vehicle, and perform distance error regulation and relative speed error control to generate the expected acceleration a des , maintain the safe following distance L set by the unmanned vehicle and the guide vehicle during high-speed autonomous maneuvers r .

[0086] d) Design and use emergency obstacle stopping and hysteresis judgment strategies to stop environmental obstacles and target vehicles during autonomous following driving, ensuring the safety of autonomous following driving.

[0087] e) Design a lateral controller to constrain the vehicle's steering behavior and generate the desired steering angle θ des , to avoid environmental obstacles and enhance the ability to resist interference in off-road environments.

[0088] The following method is further described in detail as follows:

[0089] Step S1, the specific execution steps of the target vehicle detection and identification module are as follows:

[0090] S101, in the initialization phase, the guide vehicle is set in a sector 20 meters in front of the unmanned vehicle and 10 degrees to the left and right. At this time, the millimeter-wave radar installed in the front of the unmanned vehicle can easily detect the guide vehicle target. The millimeter-wave radar can directly output the position and speed information of 64 targets. When the obstacle target remains within the field of view, the obstacle ID remains unchanged from 0 to 63, and the millimeter-wave radar can stably identify and track the guide vehicle target in the obstacle. Based on the characteristics of the millimeter-wave radar such as sensitivity to metal and insensitivity to stationary objects, the guide vehicle target can be easily detected.

[0091]

[0092] S102, the target position of the guide vehicle detected by the millimeter-wave radar is used as the initial position for laser radar algorithm recognition and tracking. During the recognition process, the target position in the previous frame of data in the laser point cloud is used to select the area of ​​interest, and then the area is clustered and analyzed. The cluster center is used as the target position of the guide vehicle, and the relative speed of the target vehicle is calculated based on the position change of the cluster center between the two frames.

[0093] S103, make a confidence judgment on the millimeter wave radar detection results and the laser radar detection results. Check the θ target Target:

[0094] If the guide vehicle target is within 10 degrees to the left and right of the unmanned vehicle, the guide vehicle target is determined to be reliable based on the tracking characteristics of the millimeter-wave radar. Then the Euler distance between the two detection targets from the millimeter-wave radar and the lidar is calculated. When the distance is greater than the set threshold, the lidar detection result is determined to be unreliable, and the millimeter-wave radar detection result is output; otherwise, the lidar detection result is considered reliable, and due to the high detection accuracy of the lidar, the lidar clustering result is output.

[0095] If the millimeter-wave radar is unreliable, the laser radar detection result is directly output until the unreliability continues until the millimeter-wave radar detection result falls into the set target area θ tar g et range, and its ID can be used to track the vehicle ahead.

[0096] S104, using the confidence judgment result as the measurement value of the Kalman filter algorithm to generate the best relative speed estimate V of the guided vehicle target r and position estimate P r .

[0097] Step S2: Design a longitudinal controller with fuzzy strategy to maintain the required safe following distance L between the unmanned vehicle and the leading vehicle. r And ensure the stable high-speed driving of the unmanned vehicle. At the same time, the distance error and relative speed between the two vehicles are adjusted and controlled. Finally, an emergency braking strategy is set to deal with emergencies in the autonomous following process in typical scenarios such as mud and sand in off-road environments.

[0098] The specific steps of the longitudinal control module are as follows:

[0099] S201, the relative distance and speed of the two vehicles are continuously changing, and the expected distance between the unmanned vehicle and the guide vehicle and the speed of the unmanned vehicle V are set. b The proportional relationship:

[0100] L des =m·V b (2)

[0101] Where m is L des and V b According to the uniform motion mode:

[0102]

[0103] Among them, V p is the speed of the car in front.

[0104] S202, design an acceleration gain controller based on relative distance and velocity feedback to obtain the desired acceleration:

[0105]

[0106] K is the acceleration gain. The state equation of the system can be obtained:

[0107]

[0108] In the state equation, X = [V b L r ] T is the state variable, u=V p is the input variable; Y = V b is the output variable, and then the closed-loop transfer function G(s) is obtained

[0109]

[0110] Verification: When m>0 and K>0, the characteristic roots of the closed-loop system are all located in the left half of the complex plane, so the system is stable. When m=3, the root locus of the system is as follows Figure 3 (a) shown.

[0111] S203, set output variable Y=L r , then the state function is

[0112]

[0113] The closed-loop transfer function is:

[0114]

[0115] Verification: When m>0 and K>0, the characteristic roots of the closed-loop system are all located in the left half of the complex plane, so the system is stable. When m=3, the root locus of the system is as follows Figure 3 (b) as shown.

[0116] In summary, the speed V of the vehicle in this system b and the relative distance L r Is stable

[0117] S204, considering the kinematic constraints of the unmanned vehicle's braking performance, the guarantee of safe distance, and the flexibility of distance control, a fuzzy table is designed to replace the speed V b The expected distance L des The linear relationship between them.

[0118] According to the analysis of steps S203 and S204, when m>0, it is equivalent to the slope of the fuzzy table being greater than zero, V b and L r will be stable, and the relative speed V r will also be zero. According to the fuzzy rule table 1, the slope is greater than zero, so the system is stable.

[0119] Table 1

[0120] <![CDATA[V b (m / s)]]> 0.0 2.0 4.0 6.0 8.0 10.0 <![CDATA[L des (m)]]> XX XX XX XX XX XX

[0121] In order to ensure sufficient safety distance, when V b = 0.0, set L des =7.0, V b With L des The relationship curve is as follows Figure 4 Considering that the response of the unmanned platform has a hysteresis characteristic, the slope of the fuzzy curve is 2 <V b <10 and gradually increase. <V bThe slope of the interval <6 is low, which can make the unmanned vehicle have better following performance when driving at low speed. This is very helpful for target detection when following turns, and is effective in off-road environments with forest cover.

[0122] S209, set the emergency braking strategy to deal with emergencies during autonomous following. In off-road environments, places with low friction coefficients such as mud and sand will greatly affect the braking performance. It is difficult to ensure the safety of following the vehicle in emergency situations using only the longitudinal speed control algorithm, so the emergency braking strategy needs to be used at the same time.

[0123] 1) By V b 、V r and L r Extract obvious mutation features to assist in detecting the emergency braking state of the guided vehicle. Obviously, the relative speed V r is an obvious feature. When the leading vehicle brakes suddenly, V r The amplitude will increase sharply. In addition, when using the target detection method without V2V communication, due to the bumps in the off-road environment, there will be large random noise in the relative speed data, and the Kalman filter algorithm in the detection module is difficult to weaken this noise. r and V r There are some common noise sources, such as the impact of uneven roads on the guide vehicle and the unmanned vehicle, the delay noise of the detection module, etc., so the typical scene noise can be reduced by type. Considering the motion model, the collision time T can be obtained c

[0124] T c =L r / abs(V r ) (9)

[0125] Rapid braking of the lead vehicle will result in an increase in relative speed and a decrease in relative distance, so the change in impact time is more dramatic than the change in relative speed. R >0,T c is the collision time of two vehicles assuming uniform motion.

[0126] 2) Setting hysteresis judgment can effectively deal with detection noise and can ensure that the vehicle following performance will not be affected when the lead vehicle accelerates again. Figure 5 As shown:

[0127] If the current state is emergency braking, set the output acceleration equal to the maximum deceleration. b =0 (the unmanned vehicle stops by braking) or T c If the time-to-collision threshold is greater than the leading vehicle (the leading vehicle is faster and safe enough to follow), the state is set to normal driving;

[0128] If the current state is normal driving, the output acceleration calculated by the longitudinal control algorithm is set;

[0129] If V b Greater than the threshold (high speed state) and T c If the time to collision is less than the threshold (high safety risk), the state is set to emergency braking.

[0130] 3) In case of emergency braking, the deceleration is directly set to the maximum, and the cooperation of the platform anti-lock braking system is required to achieve the fastest braking effect.

[0131] Step 3: Design lateral control, generate steering control value by coupling angle deviation with current vehicle speed, and enhance the adaptability and anti-interference ability of the unmanned vehicle to off-road environment.

[0132] S301, such as Figure 2 As shown, the angle error θ err You can use X r and Y r calculate:

[0133]

[0134] S302, then use the angle gain K ang Calculate the required angle θ des :

[0135] θ des =K×θ err (11)

[0136] S303, according to the vehicle kinematics model:

[0137]

[0138] S304: The lateral speed of the vehicle is proportional to the actual speed. Therefore, the vehicle speed is used to attenuate K ang :

[0139]

[0140] K max is the maximum angular gain, V max is the maximum speed, which should be greater than the maximum achievable speed of the unmanned vehicle. Using an angle controller that couples speed and angle error, the unmanned vehicle can follow the lead vehicle in the longitudinal direction.

[0141] Based on the above steps, through the comprehensive use of the target vehicle detection and identification module, the longitudinal control module and the lateral control module and the refinement of the steps, the unmanned vehicle equipped with multi-source sensors can achieve autonomous following of the guide vehicle.

[0142] The above technical scheme is only an exemplary embodiment of the present invention. For those skilled in the art, it is easy to make various types of improvements or modifications based on the application methods and principles disclosed in the present invention, and it is not limited to the method described in the above specific embodiment of the present invention. Therefore, the method described above is only preferred and does not have a restrictive meaning.

Claims

1. An autonomous following device for an unmanned vehicle based on multi-source perception in a complex off-road environment, characterized in that: include: The target vehicle detection and identification module includes: a millimeter wave radar and a laser radar arranged at the front of the unmanned vehicle, which are used to detect and identify the guide vehicle as the target and perform steady-state tracking; Longitudinal control module, used for longitudinal speed planning of the unmanned vehicle, to achieve longitudinal distance maintenance with the target vehicle, relative speed error control and safety assurance; The lateral control module is used to constrain the steering behavior of the unmanned vehicle.

2. The device according to claim 1, characterized in that In the target vehicle detection and identification module, the initial detection distance of the millimeter wave radar to the target vehicle is not less than 20 meters, and the detection field of view is not less than ±10°; the laser radar adopts a three-dimensional laser radar.

3. The device according to claim 1 or 2, characterized in that: The target vehicle detection and identification module determines the confidence of the laser radar and the millimeter wave radar, and determines the sensor with high confidence as the main detection device; The main detection sensor data is filtered to obtain the target vehicle position estimate and the relative speed estimate between the unmanned vehicle and the guide vehicle.

4. The device according to claim 1 or 2, characterized in that: The longitudinal expansion module adopts feedback control based on fuzzy strategy to plan the speed of the unmanned vehicle, and performs distance error regulation and relative speed error control to generate the expected acceleration a des , in order to maintain the safe following distance L set between the unmanned vehicle and the leading vehicle r At the same time, an emergency braking strategy is set to deal with emergencies during autonomous following.

5. The device according to claim 1 or 2, characterized in that: The lateral control module generates a steering control amount by coupling the angle deviation with the current vehicle speed.

6. An autonomous following method for an unmanned vehicle based on multi-source perception in a complex off-road environment, comprising the following steps: S1, using the laser radar and millimeter wave radar installed in front of the unmanned vehicle to detect the guide vehicle as the tracking target, automatically determines the sensor with high confidence as the main detection device; Filter the main detection sensor data to obtain the target vehicle position estimate P r and the relative speed estimation V between the unmanned vehicle and the guide vehicle r ; S2, using feedback control based on fuzzy strategy, plans the longitudinal speed of the unmanned vehicle, and performs distance error regulation and relative speed error control to obtain the expected acceleration a des , to maintain a safe following distance L between the unmanned vehicle and the lead vehicle r ; S3, setting emergency braking strategy to deal with emergencies during autonomous following; S4, using the steering angle control coupled with the speed and angle deviation, implements lateral control to constrain the vehicle's steering behavior.

7. The method according to claim 6, characterized in that The step S1 specifically includes: S11, placing the guide vehicle in front of the unmanned vehicle and within the detection range of the laser radar and millimeter wave radar; S12, using the target position of the guide vehicle detected by the millimeter wave radar as the initial position of the laser radar algorithm recognition and tracking. During the recognition process, the target position in the previous frame of data in the laser point cloud is used to select the region of interest, and then cluster analysis is performed on the region, and the cluster center is used as the target position of the guide vehicle. The relative speed of the target vehicle is calculated according to the position change of the cluster center between two adjacent frames; S13, making a confidence judgment on the millimeter wave radar detection results and the laser radar detection results: (1) Check the θ of the millimeter wave radar target Target: If the guide vehicle target is within 10 degrees to the left and right of the unmanned vehicle, the guide vehicle target is reliable based on the tracking characteristics of the millimeter-wave radar; Then, the Euler distance between the two detection targets from the millimeter-wave radar and the lidar is calculated. When the distance is greater than the set threshold, the lidar detection result is determined to be unreliable, and the millimeter-wave radar detection result is output. Otherwise, the lidar detection result is considered reliable, and the lidar clustering result is output due to the high detection accuracy of the lidar. (2) If the millimeter-wave radar is unreliable, the laser radar detection result is directly output until the unreliability continues until the millimeter-wave radar detection result falls into the set target area θ target The ID of the vehicle can be used to track the target vehicle in front. S14, using an extended Kalman filter, using the confidence judgment result as the measurement value of the Kalman filter algorithm to generate the best relative speed estimate V of the guide vehicle target r and position estimate P r .

8. The method according to claim 6 or 7, characterized in that: The step S2 specifically includes: S21, the relative distance and speed of the two vehicles are continuously changing, and the expected distance L between the unmanned vehicle and the guide vehicle is set des and the speed of the driverless car V b The proportional relationship: L des =m·V b Where m is L des and V b The ratio coefficient between According to the uniform motion mode: Among them, V p is the speed of the preceding vehicle; S22, design an acceleration gain controller based on relative distance and velocity feedback to obtain the desired acceleration A des : K is the acceleration gain; The state equation of the system is obtained: In the state equation, X = [V b L r ] T is the state variable, u=V p is the input variable; Let Y = V b is the output variable, the closed-loop transfer function G(s) Let the output variable Y = L r , then the state function is Y=[0 1]·X+0·u The closed-loop transfer function is: Among them, take m>0 and K>0; S23, design a fuzzy table to replace the speed V of the unmanned vehicle b The expected distance L des The linear relationship between them, where the slope of the fuzzy table is greater than zero, to ensure that V b and L r stability; at the same time, the slope of the fuzzy curve is 2 <V b <10, gradually increase to adapt to the hysteresis characteristics of the unmanned platform response; For example: When V b = 0.0, set L des =7.

0.

9. The method according to claim 8, characterized in that The specific method of step S3 includes: V b 、V r and L r Extract obvious mutation features to assist in judging the emergency braking state of the guided vehicle: Set the collision time T c T c =L r / abs(V r ) If V r >0, then T c is the collision time of two vehicles assuming uniform motion; At this time, if V b = 0 means the unmanned vehicle is in a braking stop state, or T c If it is greater than the collision time threshold, the state is judged as normal driving; If the current state is normal driving, the acceleration is set to the output acceleration calculated by the longitudinal control algorithm in step S2; If V b is greater than the threshold, i.e. high speed state, and T c If it is less than the collision time threshold, the state is judged as emergency braking; In case of emergency braking, the deceleration is directly set to the maximum.

10. The method according to claim 6, characterized in that The step S4 specifically includes: S41, estimate P according to the target vehicle position r The lateral error X contained in r and longitudinal error Y r , calculate the angle deviation θ err : S42, use angle gain K ang Calculate the required steering angle θ des : i des =K ang ×θ err ; S43, according to the vehicle kinematics model: The lateral velocity of the vehicle is proportional to the actual velocity, so the vehicle velocity is used to attenuate K ang : Among them, K msc is the maximum angular gain, V msc is the maximum speed, which should be greater than the maximum achievable speed of the unmanned vehicle.

Images