Intersection passing method, device and equipment for automatic driving vehicle, and computer program product

Abstract

The invention discloses an intersection passing method, device and equipment for an automatic driving vehicle, and a computer program product, and the method comprises the steps: obtaining the multi-sensor sensing data and positioning data of the automatic driving vehicle, and receiving the dynamic signal lamp data of an intersection region where the automatic driving vehicle is located from a road end; determining a current signal lamp scene and corresponding signal lamp data according to the multi-sensor sensing data and the dynamic signal lamp data; according to the positioning data, the current signal lamp scene and the signal lamp data, determining a passing strategy of the autonomous vehicle in the intersection area; and according to the passing strategy, determining current control parameters of the autonomous vehicle by using a preset decision control algorithm so as to control the passing of the autonomous vehicle in the intersection area. According to the invention, intelligent traffic control of the autonomous vehicle at the intersection area is realized, a reasonable intersection traffic strategy can be formulated according to the real-time road condition and the state of the signal lamp, and the safety and efficiency of the autonomous vehicle passing at the intersection are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent transportation, and in particular to a method and device for intersection passing of an autonomous vehicle, an apparatus, and a computer program product. BACKGROUND

[0002] With the rapid development of autonomous driving technology, the safety and efficiency of vehicle passing at intersections have become core challenges in this field. Intersections, as key nodes of the transportation network, have complex and variable traffic conditions, involving various traffic participants (such as motor vehicles, non-motor vehicles, and pedestrians) and dynamically changing traffic signals (such as signal light countdown), which puts high demands on the signal light recognition accuracy, dynamic decision-making real-time performance, and emergency response capability of autonomous driving systems.

[0003] However, existing autonomous driving technology has problems such as insufficient signal light recognition accuracy, lack of cooperative control, poor decision-making real-time performance, and weak fault tolerance mechanism in intersection passing scenarios, which cannot meet the strict requirements for safety and efficiency in actual traffic environments.

[0004] Therefore, it is of great practical significance to design an intersection passing decision scheme for autonomous vehicles that can solve at least one of the above problems. SUMMARY

[0005] Embodiments of the present application provide a method and device for intersection passing of an autonomous vehicle, an apparatus, and a computer program product to improve the efficiency and safety of intersection passing of autonomous vehicles.

[0006] Embodiments of the present application adopt the following technical solutions:

[0007] In a first aspect, a method for intersection passing of an autonomous vehicle is provided, which includes:

[0008] obtaining multi-sensor perception data and positioning data of the autonomous vehicle, and receiving dynamic signal light data of an intersection area where the autonomous vehicle is located from a road end;

[0009] determining a current signal light scenario and corresponding signal light data of the autonomous vehicle based on the multi-sensor perception data and the dynamic signal light data;

[0010] determining a passing strategy of the autonomous vehicle in the intersection area based on the positioning data and the current signal light scenario and corresponding signal light data of the autonomous vehicle;

[0011] According to the passing strategy of the automatic driving vehicle in the intersection area, a preset decision control algorithm is used to determine a current control parameter of the automatic driving vehicle, and the automatic driving vehicle is controlled to pass through the intersection area according to the current control parameter.

[0012] Optionally, the receiving of the dynamic traffic light data of the intersection area where the automatic driving vehicle is located from the road end comprises:

[0013] Receiving the dynamic traffic light data of the intersection area where the automatic driving vehicle is located from the road end based on an optimized V2I communication protocol;

[0014] The optimized V2I communication protocol is defined as dynamically adjusting the transmission frequency and transmission field of the signal light data based on the distance mode of the automatic driving vehicle to the intersection area.

[0015] Optionally, the multi-sensor perception data comprises visual perception data, and the determining of the current traffic light scene and corresponding traffic light data of the automatic driving vehicle according to the multi-sensor perception data and the dynamic traffic light data comprises:

[0016] Performing traffic light recognition according to the visual perception data to obtain a first visual-based traffic light recognition result;

[0017] Comparing the first visual-based traffic light recognition result with the dynamic traffic light data;

[0018] If the first visual-based traffic light recognition result is inconsistent with the dynamic traffic light data, auxiliary verification is performed using the visual perception data, and the current traffic light scene and corresponding traffic light data of the automatic driving vehicle are determined according to the auxiliary verification result;

[0019] Otherwise, the current traffic light scene and corresponding traffic light data of the automatic driving vehicle are directly determined according to the first visual-based traffic light recognition result or the dynamic traffic light data.

[0020] Optionally, the auxiliary verification using the visual perception data and the determination of the current traffic light scene and corresponding traffic light data of the automatic driving vehicle according to the auxiliary verification result comprise:

[0021] Detecting the motion state of surrounding vehicles according to the visual perception data;

[0022] Determining a second visual-based traffic light recognition result according to the motion state of the surrounding vehicles;

[0023] Determining the current traffic light scene and corresponding traffic light data of the automatic driving vehicle according to the second visual-based traffic light recognition result.

[0024] Optionally, the determining the passing strategy of the autonomous vehicle at the intersection region according to the positioning data and the current signal light scenario and corresponding signal light data of the autonomous vehicle comprises:

[0025] calculating a distance of the autonomous vehicle to a stop line according to the positioning data and stop line data of the intersection region;

[0026] if the current signal light scenario is a green light scenario, calculating a driving speed safety range of the autonomous vehicle in the current signal light scenario according to the distance of the autonomous vehicle to the stop line and the signal light data, and determining the passing strategy of the autonomous vehicle in the green light scenario according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line and a safety margin of the green light scenario;

[0027] if the current signal light scenario is a red light scenario or a yellow light scenario, determining the passing strategy of the autonomous vehicle in the red light scenario or the yellow light scenario according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line and a safety margin of the red light scenario or the yellow light scenario.

[0028] Optionally, the determining the passing strategy of the autonomous vehicle at the intersection region according to the positioning data and the current signal light scenario and corresponding signal light data of the autonomous vehicle comprises:

[0029] determining whether deceleration braking is needed according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line and a safety margin of the current signal light scenario;

[0030] if yes, establishing a virtual obstacle for braking, calculating a braking distance and determining a deceleration passing strategy according to the braking distance and the distance of the autonomous vehicle to the stop line;

[0031] otherwise, determining that the passing strategy of the autonomous vehicle in the current signal light scenario is a normal passing strategy.

[0032] Optionally, the intersection passing method of the autonomous vehicle further comprises:

[0033] detecting whether the autonomous vehicle has a fault;

[0034] in the case that the autonomous vehicle has a fault, performing dynamic decision and control on the autonomous vehicle according to the fault type by using a pre-defined multi-level fault-tolerant mechanism and an emergency braking strategy.

[0035] In a second aspect, the embodiments of the present application further provide an intersection passing device of an autonomous vehicle, the intersection passing device of the autonomous vehicle comprising:

[0036] an acquisition unit configured to acquire multi-sensor perception data and positioning data of the autonomous vehicle, and receive dynamic traffic light data of an intersection region where the autonomous vehicle is located from a road end;

[0037] a first determination unit configured to determine a current traffic light scenario and corresponding traffic light data of the autonomous vehicle according to the multi-sensor perception data and the dynamic traffic light data;

[0038] a second determination unit configured to determine a passing strategy of the autonomous vehicle in the intersection region according to the positioning data and the current traffic light scenario and corresponding traffic light data of the autonomous vehicle;

[0039] a decision control unit configured to determine a current control parameter of the autonomous vehicle by using a preset decision control algorithm according to the passing strategy of the autonomous vehicle in the intersection region, and control passing of the autonomous vehicle in the intersection region according to the current control parameter.

[0040] In a third aspect, the embodiments of the present application further provide an apparatus comprising:

[0041] a processor; and a memory arranged to store computer executable instructions that, when executed, cause the processor to perform any of the aforementioned intersection passing methods of the autonomous vehicle.

[0042] In a fourth aspect, the embodiments of the present application further provide a computer program product comprising computer programs / instructions that, when executed by a processor, implement any of the aforementioned intersection passing methods of the autonomous vehicle.

[0043] The at least one technical scheme adopted by the embodiment of the present application can achieve the following beneficial effects: the intersection passing method of the autonomous vehicle in the embodiment of the present application first acquires multi-sensor perception data and positioning data of the autonomous vehicle, and receives dynamic signal lamp data of an intersection area where the autonomous vehicle is located from a road end; then determines a current signal lamp scene and corresponding signal lamp data of the autonomous vehicle according to the multi-sensor perception data and the dynamic signal lamp data; thereafter, determines a passing strategy of the autonomous vehicle in the intersection area according to the positioning data and the current signal lamp scene and corresponding signal lamp data of the autonomous vehicle; finally, determines a current control parameter of the autonomous vehicle by using a preset decision control algorithm according to the passing strategy of the autonomous vehicle in the intersection area, and controls the passing of the autonomous vehicle in the intersection area according to the current control parameter. The intersection passing method of the autonomous vehicle in the embodiment of the present application realizes intelligent passing control of the autonomous vehicle in the intersection area by comprehensively acquiring multi-sensor perception data, positioning data and road end dynamic signal lamp data, and performing scene recognition, passing strategy determination and control parameter calculation based on these data, can accurately judge the scene where the vehicle is located according to real-time road conditions and signal lamp states, and formulate a reasonable passing strategy, thereby reducing the risk of traffic accidents caused by human judgment errors or untimely reactions, and effectively improving the safety and efficiency of the autonomous vehicle in the intersection passing. BRIEF DESCRIPTION OF DRAWINGS

[0044] The accompanying drawings, which are incorporated herein and form a part of the specification, illustrate the embodiments of the present application and, together with the description, further serve to explain the principles of the present application and to enable this present application to be put into practice. In the drawings:

[0045] Figure 1 FIG. 1 is a flowchart of an intersection passing method of an autonomous vehicle in an embodiment of the present application;

[0046] Figure 2 FIG. 2 is a structural diagram of an intersection passing device of an autonomous vehicle in an embodiment of the present application;

[0047] Figure 3 FIG. 3 is a structural diagram of an equipment in an embodiment of the present application. DETAILED DESCRIPTION

[0048] To make the objectives, technical schemes and advantages of the present application clearer, the technical schemes of the present application will be described below in conjunction with specific embodiments of the present application and corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0049] The technical solutions provided by the embodiments of the present application are described in detail below with reference to the drawings.

[0050] The prior art has many deficiencies in dealing with the core challenges faced by the field of autonomous driving, as follows:

[0051] (1) Insufficient signal light recognition accuracy

[0052] Traditional autonomous driving systems mainly rely on a single camera for visual recognition to obtain signal light information. However, the imaging quality of the camera is easily disturbed by external environmental factors. In scenes with drastic changes in light, such as backlight or night light interference, the images captured by the camera may appear overexposed or underexposed, causing the signal light features to be blurred or even lost. In adverse weather conditions (such as rain, snow, and fog), raindrops, snowflakes, or fog particles can block the camera lens, reducing image clarity and affecting accurate recognition of signal lights. In addition, dynamic occlusion problems are also prominent. Large vehicles, greenery, and other vegetation may temporarily block the signal lights during driving, making it impossible for the camera to continuously obtain complete signal light information. Public data shows that the misrecognition rate of existing signal light recognition algorithms based on a single camera is as high as 15%-20%, which seriously affects the safety and decision-making accuracy of autonomous vehicles at intersections.

[0053] (2) Lack of coordination control

[0054] Existing autonomous driving systems lack effective multi-sensor data fusion mechanisms. In intersection passing scenarios, in addition to accurately recognizing signal light information, it is also necessary to timely detect pedestrians, non-motor vehicles, and other lateral traffic participants to avoid collision accidents. However, a single sensor cannot comprehensively and accurately perceive surrounding environmental information. Laser radar can provide high-precision three-dimensional point cloud data, but it is not effective for identifying transparent or reflective objects. Millimeter wave radar can detect the distance and speed of objects, but has low resolution and is difficult to accurately identify object shapes. Although cameras can provide rich image information, their reliability is limited in complex environments. Due to the lack of multi-sensor (laser radar, millimeter wave radar, camera) data coordination fusion, the system cannot fully utilize the advantages of each sensor, resulting in insufficient detection capability for sudden disturbances such as pedestrians and non-motor vehicles. According to statistics, 80% of intersection accidents are caused by the system's failure to timely detect non-motor vehicles crossing laterally, highlighting the serious deficiencies of existing technologies in coordination control.

[0055] (2) Poor decision-making real-time performance

[0056] During intersection passing, the state of the signal light can change at any time, especially in the case of sudden mutation of the countdown, which puts high requirements on the real-time decision-making of the automatic driving system. Some existing automatic driving systems use reinforcement learning algorithms (such as Q-learning) for decision-making, but such algorithms rely on a large amount of training data to learn the optimal strategy. In actual application, due to the complexity and diversity of intersection scenarios, it is difficult to collect sufficient comprehensive and representative training data. Moreover, during online decision-making, reinforcement learning algorithms require a large amount of calculation and state evaluation, resulting in high decision-making delay (> 200ms). When the countdown of the signal light suddenly changes, the high-delay decision-making system cannot respond in time to adjust the vehicle's driving state, thereby increasing the risk of vehicle passing through the intersection.

[0057] (3) Weak fault tolerance mechanism

[0058] During the operation of the automatic driving system, various abnormal situations may occur, such as communication interruption or sensor failure. The existing technology has a weak fault tolerance mechanism when facing these abnormal situations. When communication interruption or sensor failure occurs, the existing degradation strategy can only switch to a single perception mode, and the priority rules are not clearly defined. For example, when vehicle-to-infrastructure (V2I) communication and camera fail at the same time, the system may completely lose the ability to recognize the signal light and cannot obtain critical traffic information, thereby causing the vehicle to lose behavior at the intersection, seriously threatening traffic safety.

[0059] Based on this, the embodiments of the present application provide an intersection passing method for an automatic driving vehicle, as shown in Figure 1 The flowchart of the intersection passing method for the automatic driving vehicle provided by the embodiments of the present application is shown in the figure, and the intersection passing method for the automatic driving vehicle at least includes the following steps S110 to S140:

[0060] Step S110, acquiring multi-sensor perception data and positioning data of the automatic driving vehicle, and receiving dynamic signal light data of the intersection area where the automatic driving vehicle is located from the road end.

[0061] The automatic driving vehicle is equipped with various sensors, such as cameras, millimeter wave radars, laser radars, etc. These sensors continuously collect information about the environment around the vehicle, including but not limited to the positions, speeds, pedestrian dynamics, and obstacle conditions of other vehicles, forming multi-sensor perception data. Through the fusion positioning system on the automatic driving vehicle, the accurate position, driving direction, and speed of the automatic driving vehicle in the map can be obtained.

[0062] In the vehicle-road cooperation scenario, the automatic driving vehicle also establishes a communication connection with the infrastructure at the road end (such as a signal lamp device, a roadside device, etc.), and can receive dynamic signal lamp data of the intersection area where the automatic driving vehicle is located from the road end. The dynamic signal lamp data includes information such as the current state (red light, green light, yellow light) and the remaining time of the signal lamp.

[0063] In step S120, the current signal lamp scene and the corresponding signal lamp data of the automatic driving vehicle are determined according to the multi-sensor perception data and the dynamic signal lamp data.

[0064] The dynamic signal lamp data is from the road end, and the multi-sensor perception data is from the vehicle end. The multi-sensor perception data, such as visual perception data, can also provide signal lamp recognition information. Therefore, by comprehensively processing and analyzing the multi-sensor perception data and the dynamic signal lamp data, the current actual signal lamp scene of the automatic driving vehicle can be determined, such as a green light, a red light, or a yellow light scene, and the corresponding signal lamp data, such as the countdown data of the signal lamp, can be determined. This approach avoids the inaccuracy of a single information source and improves the accuracy and reliability of signal lamp recognition.

[0065] In step S130, the passing strategy of the automatic driving vehicle in the intersection area is determined according to the positioning data and the current signal lamp scene and the corresponding signal lamp data of the automatic driving vehicle.

[0066] In combination with the positioning data obtained in step S110, the specific position and driving direction of the automatic driving vehicle in the intersection area are determined. At the same time, considering the current signal lamp scene and the corresponding signal lamp data determined in step S120, the passing strategy of the automatic driving vehicle in the intersection area is determined. The passing strategy may, for example, include deceleration and braking, normal passing, etc.

[0067] In step S140, the current control parameters of the automatic driving vehicle are determined by using a preset decision control algorithm according to the passing strategy of the automatic driving vehicle in the intersection area, and the passing of the automatic driving vehicle in the intersection area is controlled according to the current control parameters.

[0068] According to the passing strategy of the automatic driving vehicle in the intersection area determined in the foregoing steps, the current control parameters of the vehicle are calculated by using a preset decision control algorithm. These control parameters include throttle opening, brake force, steering angle, etc., which are used to accurately control the driving state of the vehicle. The calculated current control parameters are sent to the execution mechanism of the vehicle, such as the engine control system, the braking system, and the steering system, and the passing of the automatic driving vehicle in the intersection area is controlled in real time according to these parameters, so as to ensure that the vehicle safely and smoothly passes through the intersection according to the predetermined passing strategy.

[0069] The above decision control algorithm may, for example, adopt a longitudinal DPQP+MPC cooperative mode, specifically including:

[0070] (1) Decision layer: longitudinal DPQP speed planning

[0071] Dynamic programming (DP) stage: state space definition and cost function design; quadratic programming (QP) stage: the rough solution of DP is used as the initial value of QP, and the solution is refined. The solution, i.e., a series of trajectory points, is sent to the control layer.

[0072] (2) Control layer: MPC (model predictive control) trajectory tracking

[0073] A bicycle model is used to construct a vehicle model; a prediction time domain, a control time domain, and a sampling period are determined for rolling optimization; finally, EKF (extended Kalman filter) is used to fuse GPS and IMU data to correct the state estimation value in real time, and finally control the throttle and steering angle.

[0074] Through the deep cooperation of longitudinal DPQP and MPC, the deficiencies of traditional methods in real-time performance, control accuracy, and comfort are compensated.

[0075] The intersection passing method of the automatic driving vehicle in the embodiments of the application realizes intelligent passing control of the automatic driving vehicle in the intersection area by comprehensively obtaining multi-sensor perception data, positioning data, and road end dynamic signal lamp data, and performing scene recognition, passing strategy determination, and control parameter calculation based on the data. The method can accurately judge the scene in which the vehicle is located and develop a reasonable passing strategy according to the real-time road conditions and signal lamp state, thereby reducing the risk of traffic accidents caused by human judgment errors or untimely reactions, and effectively improving the safety and efficiency of the automatic driving vehicle in intersection passing.

[0076] In some embodiments of the application, the receiving of the dynamic signal lamp data of the intersection area in which the automatic driving vehicle is located from the road end comprises: receiving the dynamic signal lamp data of the intersection area in which the automatic driving vehicle is located from the road end based on an optimized V2I communication protocol; wherein the optimized V2I communication protocol is defined as dynamically adjusting the transmission frequency and transmission field of the signal lamp data based on the distance mode of the automatic driving vehicle to the intersection area.

[0077] V2I refers to communication between a vehicle and infrastructure, which is used for information exchange between a vehicle and road infrastructure (such as a traffic light, a roadside unit RSU) to optimize traffic flow. The embodiments of the application define an optimized V2I communication protocol in the automatic driving system and the road end communication module. The core rule of the protocol is to dynamically adjust the transmission frequency and transmission field of the signal lamp data based on the distance mode of the automatic driving vehicle to the intersection area.

[0078] Based on the V2I communication protocol, the autonomous vehicle can send the positioning data output by its positioning system to the road terminal in real time, and the road terminal can calculate the distance from the vehicle to the intersection area (such as the intersection stop line) in real time. According to the calculated distance, the distance mode of the autonomous vehicle currently in is determined according to the preset distance threshold. For example, when the vehicle is more than 200m away from the intersection, it is determined to be a long distance mode; when the vehicle is between 50m and 200m away from the intersection, it is determined to be a medium distance mode; and when the vehicle is less than or equal to 50m away from the intersection, it is determined to be a short distance mode.

[0079] Of course, how to define and divide the above distance modes can be adjusted flexibly by those skilled in the art according to actual needs, and here is not limited.

[0080] When the autonomous vehicle is in the long distance mode, the road terminal device can send signal lamp data to the autonomous vehicle according to a lower transmission frequency, such as a transmission frequency of 1Hz. The data packet contains signal lamp ID and basic state (red / green / yellow) information. When the vehicle enters the medium distance mode, the road terminal device increases the transmission frequency of the signal lamp data, for example, adjusts to 5Hz, and adds a countdown field (such as accuracy 0.5 seconds) in the data packet. When the vehicle enters the short distance mode, the road terminal device further increases the transmission frequency, for example, adjusts to 10Hz, and further supplements lane association information (such as left turn / dedicated light) in the data packet.

[0081] During the vehicle driving process, as the distance between the vehicle and the intersection changes continuously, the distance is continuously detected and the distance mode is determined in real time. Once the distance mode changes, the road terminal device adjusts the transmission frequency and transmission field of the signal lamp data according to the new mode, to ensure that the data transmission matches the actual needs of the vehicle.

[0082] On the one hand, by dynamically adjusting the transmission frequency of the signal lamp data according to the distance between the vehicle and the intersection, a lower frequency transmission is adopted in the long distance mode, which reduces unnecessary data transmission and avoids network congestion, thereby reducing communication delay; in the short distance mode, the transmission frequency is increased to ensure that the vehicle can obtain the latest signal lamp information in time, and the real-time of the information is ensured.

[0083] On the other hand, according to different distance modes, the transmission field is dynamically adjusted, only the basic signal lamp state information is transmitted in the long distance mode, the countdown information is added in the medium distance mode, and the lane association information is supplemented in the short distance mode, so that the transmitted data is always closely related to the current needs of the vehicle, the effectiveness and utilization of the data are improved, and the transmission of useless data is reduced.

[0084] The dynamic adjustment manner rationally utilizes communication resources and computing resources, avoids resource waste caused by transmission of a large amount of data in all cases, and reduces the processing burden of the vehicle decision system, thereby improving the performance and stability of the entire automatic driving system.

[0085] In some embodiments of the present application, the multi-sensor perception data includes visual perception data, and determining the current signal light scene and corresponding signal light data in which the automatic driving vehicle is located according to the multi-sensor perception data and the dynamic signal light data includes: performing signal light identification according to the visual perception data to obtain a first visual-based signal light identification result; comparing the first visual-based signal light identification result with the dynamic signal light data; if the first visual-based signal light identification result is inconsistent with the dynamic signal light data, performing auxiliary verification using the visual perception data, and determining the current signal light scene and corresponding signal light data in which the automatic driving vehicle is located according to an auxiliary verification result; otherwise, directly determining the current signal light scene and corresponding signal light data in which the automatic driving vehicle is located according to the first visual-based signal light identification result or the dynamic signal light data.

[0086] During driving, the multi-sensor system carried by the automatic driving vehicle continuously collects surrounding environment information, including visual perception data. At the same time, the vehicle receives dynamic signal light data of the intersection area in which the vehicle is located from the road end through an optimized V2I communication protocol (such as the distance mode dynamic adjustment protocol described in the foregoing embodiments).

[0087] The visual processing module inside the vehicle processes the obtained visual perception data. The module may, for example, use an improved YOLO-v7 model, increase CBAM (convolutional block attention module), and improve the feature extraction capability in complex scenes. At the same time, a color separation algorithm is used to set a dynamic threshold in the HSV color space, thereby distinguishing signal lights from tail lights. Finally, a first visual-based signal light identification result is obtained.

[0088] The first visual-based signal light identification result is compared with the dynamic signal light data received from the road end to determine whether the two are consistent. The comparison mainly includes the color state of the signal light and other related information (such as countdown) that may be included. For example, if the visual identification result is a green light, and the dynamic signal light data shows a red light, it is considered that the two are inconsistent.

[0089] If the first visual-based signal light recognition result is inconsistent with the dynamic signal light data, a redundant auxiliary verification process is started. In addition to the image information used for initial signal light recognition, the visual perception data also contains rich environmental context information. For example, the behavior of surrounding vehicles, the movement of pedestrians, traffic signs at the intersection, etc. These auxiliary information is used to further infer and verify the signal light state. According to the results of auxiliary verification, the current signal light scene and the corresponding signal light data of the autonomous vehicle are determined comprehensively.

[0090] If the first visual-based signal light recognition result is consistent with the dynamic signal light data, the current signal light scene and the corresponding signal light data of the autonomous vehicle can be determined directly according to any one of the results (the first visual-based signal light recognition result or the dynamic signal light data). Because they are consistent, selecting one of the results can meet the demand for determining the signal light scene and data, without the need for complex auxiliary verification process.

[0091] By combining visual perception data and dynamic signal light data, and using the comparison and auxiliary redundant verification mechanism, the advantages of different data sources are fully utilized, the errors and errors that may exist in a single data source are effectively reduced, and the accuracy and reliability of signal light recognition are greatly improved.

[0092] In some embodiments of the present application, the auxiliary verification using the visual perception data and determining the current signal light scene and the corresponding signal light data of the autonomous vehicle according to the auxiliary verification result comprises: detecting the motion state of surrounding vehicles according to the visual perception data; determining a second visual-based signal light recognition result according to the motion state of the surrounding vehicles; and determining the current signal light scene and the corresponding signal light data of the autonomous vehicle according to the second visual-based signal light recognition result.

[0093] The autonomous vehicle can continuously collect visual perception data of the environment around the vehicle through the camera mounted on the vehicle. These data cover scene information in multiple directions in front of the vehicle, side of the vehicle, etc., and contain elements such as surrounding vehicles, pedestrians, traffic signs and signal lights.

[0094] The position and contour of the surrounding vehicles are detected in the visual perception data using a target detection algorithm (such as an improved YOLO series algorithm or Faster R-CNN algorithm, etc.). The position information of the detected surrounding vehicles at different times is tracked and analyzed by optical flow method or motion estimation method based on deep learning, so as to determine the motion state of each vehicle, including acceleration, deceleration, uniform speed driving or static, etc. For example, by calculating the position change of the vehicle in consecutive several frames of images, if the position change gradually increases, it is judged that the vehicle is accelerating; if the position change gradually decreases, it is judged that the vehicle is decelerating.

[0095] Statistical analysis is performed on the motion state of surrounding vehicles. The number of vehicles and their motion state distribution within a certain range (e.g., a certain distance ahead of the lane where the autonomous vehicle is located) are counted. Based on the statistical analysis result, a second visual-based signal light recognition result is determined. According to traffic rules and common driving behavior patterns, if most vehicles are decelerating or stationary, it can be reasonably determined that the current signal light is red; if most vehicles are accelerating through the intersection, it is determined that the current signal light is green.

[0096] The second visual-based signal light recognition result is compared and analyzed with the first visual-based signal light recognition result (if any) and the dynamic signal light data (received from the road end). If the second visual-based signal light recognition result is consistent with the dynamic signal light data, the current signal light scenario and signal light data of the autonomous vehicle are determined based on the second visual-based signal light recognition result and the dynamic signal light data. For example, if the second visual-based signal light recognition result is green and the dynamic signal light data also shows green, it is determined that the current signal light scenario is green, and the corresponding signal light data is the green state and related countdown data.

[0097] By detecting the motion state of surrounding vehicles using visual perception data, the signal light state is further verified, and the behavior correlation between vehicles in the traffic scene is fully utilized. Even if the direct recognition of the signal light by the visual sensor is disturbed (e.g., the signal light is blocked, the light is too strong or too dark, etc., leading to inaccurate recognition), or the dynamic signal light data from the road end communication is abnormal, the motion state of surrounding vehicles often reflects the true signal light situation, thereby improving the accuracy and reliability of the overall signal light recognition.

[0098] The above technical solution provides a redundant signal light recognition mechanism for the autonomous driving system, avoiding the failure of a single recognition method and leading to incorrect decisions by the autonomous vehicle, enhancing the fault tolerance of the system in complex and uncertain environments, and thus improving the safety of autonomous driving.

[0099] In some embodiments of the present application, the determining the passing strategy of the autonomous vehicle at the intersection region according to the positioning data and the current signal light scenario and corresponding signal light data of the autonomous vehicle comprises: calculating the distance of the autonomous vehicle to the stop line according to the positioning data and the stop line data of the intersection region; if the current signal light scenario is a green light scenario, calculating the driving speed safety range of the autonomous vehicle in the current signal light scenario according to the distance of the autonomous vehicle to the stop line and the signal light data, and determining the passing strategy of the autonomous vehicle in the green light scenario according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line and the safety margin of the green light scenario; if the current signal light scenario is a red light scenario or a yellow light scenario, determining the passing strategy of the autonomous vehicle in the red light scenario or the yellow light scenario according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line and the safety margin of the red light scenario or the yellow light scenario.

[0100] According to the current positioning data of the autonomous vehicle and the stop line data of the intersection region, the distance of the autonomous vehicle to the stop line (stop_line_s) can be calculated. In addition, the safety margin in different scenarios is set in advance, for example, the safety margin in the green light scenario (green_buffer) is set to 3m, and the safe parking distance in the red light scenario (red_buffer) is set to 5m. These safety margins are used to ensure the safety of the vehicle when passing through the intersection or parking.

[0101] (1) Passing strategy in green light scenario:

[0102] On the one hand, the safety speed range in the green light scenario can be calculated by the following formula:

[0103] v_min = (stop_line_s - green_buffer) / green_duration_t, (1)

[0104] v_max = (stop_line_s + green_buffer) / green_duration_t, (2)

[0105] where v_min is the minimum safe passing speed, v_max is the maximum allowed speed, green_buffer is the safety margin under green light scenario, and green_duration_t is the remaining green light time. The speed of the autonomous vehicle needs to satisfy v_min≤v≤v_max to ensure that it can safely pass the stop line or stop within the safety margin range at the end of the green light. The safe speed range of the autonomous vehicle under the green light scenario can be used as one of the constraint conditions for subsequent speed planning control.

[0106] On the other hand, the remaining time of the green light countdown is monitored in real time. If the countdown time suddenly decreases (e.g., from 15 seconds to 5 seconds), a recalculation is triggered immediately. At the same time, an acceleration constraint (a≤2.5m / s 2 ) is introduced to avoid dangerous situations caused by emergency acceleration.

[0107] In another aspect, according to the current speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line, and the safety margin of the green light scenario, it is determined whether the autonomous vehicle can pass the stop line within the safety margin at the current speed at the end of the green light, and different passing strategies are adopted, such as normal passing or the need to slow down and stop.

[0108] (2) Passing strategy under red / yellow light scenario:

[0109] According to the current speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line, and the safety margin of the red / yellow light scenario, it is determined whether the autonomous vehicle can drive to the safe stopping distance behind the stop line at the current speed at the end of the red / yellow light, and different passing strategies are adopted, such as the need to slow down and stop or the possibility of normal passing.

[0110] By accurately calculating the speed safety range and passing strategy under different signal light scenarios, combined with the safety margin and dynamic adjustment mechanism, it is ensured that the autonomous vehicle can make reasonable decisions according to the actual situation when driving at the intersection, avoiding traffic accidents caused by inappropriate speed or decision-making errors, and effectively improving the safety of intersection passing.

[0111] The above technical solutions fully consider the characteristics of different signal light scenarios and the actual state of the vehicle, such as speed, planned trajectory length, etc. In the green light scenario, passing is given priority to improve passing efficiency; in the red / yellow light scenario, stopping is given priority to ensure safe stopping. At the same time, the dynamic adjustment strategy can cope with sudden changes in signal light countdown, making the vehicle's decision-making more reasonable and adaptive to actual traffic conditions.

[0112] In some embodiments of the present application, the determination of the passing strategy of the autonomous vehicle at the intersection region according to the positioning data and the current signal light scenario and corresponding signal light data of the autonomous vehicle comprises: determining whether to need to slow down and brake according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line, and the safety margin of the current signal light scenario; if yes, establishing a virtual obstacle for braking, calculating a braking distance and determining a slow-down passing strategy according to the braking distance and the distance of the autonomous vehicle to the stop line; otherwise, determining the passing strategy of the autonomous vehicle under the current signal light scenario as a normal passing strategy.

[0113] The autonomous vehicle obtains the current vehicle speed (vehicle_v) through its own sensor system, and the maximum length of the planned trajectory (planning_max_dis) can be obtained by using a planning module.

[0114] Green light scenario: comparing the size of (stop_line_s+green_buffer) and max(vehicle_v*green_duration_t, planning_max_dis). If stop_line_s+green_buffer>max(vehicle_v*green_duration_t, planning_max_dis), it indicates that the vehicle cannot pass the stop line safety margin at the current vehicle speed when the green light ends, and the vehicle needs to slow down and brake; otherwise, the vehicle can pass normally, and a normal passing strategy is adopted, i.e. the vehicle drives normally at the current speed.

[0115] Red light scenario: comparing the size of (stop_line_s-red_buffer) and max(vehicle_v*red_duration_t, planning_max_dis). If stop_line_s-red_buffer>max(vehicle_v*red_duration_t, planning_max_dis), it means that the vehicle does not need to brake when the red light ends, and the vehicle can choose to pass normally at the current speed; otherwise, the vehicle needs to slow down and brake. The judgment of the yellow light scenario is similar to that of the red light scenario, which is not described here.

[0116] When it is determined that the vehicle needs to slow down and brake (including the cases where the vehicle needs to brake under green light, red light, and yellow light scenarios), a virtual obstacle is established, the position of which is dynamically updated according to real-time data, so as to simulate the parking limit position when the vehicle brakes, and assist the vehicle to perform reasonable brake control.

[0117] As another implementation, in order to improve the efficiency of passing, passing is given priority in the green light scenario; in order to ensure safe stopping, stopping is given priority in the red light / yellow light scenario. Therefore, in the green light scenario, the virtual obstacle can be established when it is actually determined that the vehicle needs to slow down and stop, while in the red light / yellow light scenario, the virtual obstacle model can be established in advance, and the established virtual obstacle model can be directly applied when the vehicle needs to slow down and stop.

[0118] Further, when the vehicle needs to slow down and stop, the braking distance d_brake can be calculated according to the following formula:

[0119] d_brake = v 2 / (2 * a_min) + t_reaction * v, (3)

[0120] where a_min is the maximum deceleration of the vehicle, and t_reaction is the system response time (for example, set to 0.5 seconds).

[0121] Overline permission judgment and strategy determination: compare the sizes of the remaining distance stop_line_s and the braking distance d_brake. If stop_line_s≥d_brake, it means that the vehicle has enough distance to stop smoothly before the stop line, and at this time, a smooth deceleration strategy can be adopted to make the vehicle decelerate smoothly to stop; if stop_line_s<d_brake, it means that the vehicle may not be able to stop completely before the stop line, and under the premise of complying with traffic regulations, slight overline (such as overline distance≤0.5m) is allowed, and deceleration operation is performed according to the corresponding rules.

[0122] By comprehensively considering the current vehicle speed, the planned trajectory, the distance to the stop line, and the safety margin of different signal light scenarios, it is accurately determined whether the vehicle needs to brake, and the braking strategy is reasonably planned, which effectively avoids collisions and other safety accidents caused by improper speed or decision-making errors of the vehicle at the intersection, and greatly improves the safety of the automatic driving vehicle passing through the intersection.

[0123] The corresponding judgment logic and passing strategy are formulated for different signal light scenarios, which can make reasonable decisions according to real-time traffic conditions. In the green light scenario, the remaining green light time and the vehicle driving capability are fully considered to avoid unnecessary stopping; in the red light / yellow light scenario, the vehicle is ensured to stop safely. At the same time, the dynamic updating of the virtual obstacle and the accurate calculation of the braking distance make the decision of the vehicle more adaptive to the actual road condition changes.

[0124] In some embodiments of the present application, the intersection passing method of the autonomous vehicle further comprises: detecting whether the autonomous vehicle has a fault; and in the case that the autonomous vehicle has a fault, dynamically deciding and controlling the autonomous vehicle according to the fault type by using a pre-defined multi-level fault tolerance mechanism and an emergency braking strategy.

[0125] During the operation of the autonomous vehicle, the vehicle can be continuously monitored in real time by a built-in fault diagnosis system for each key subsystem. These subsystems include but are not limited to a computing unit (such as a module related to DPQP solving), a communication module (for V2I communication), and sensors (such as a laser radar, a camera, and a millimeter wave radar). The fault diagnosis system determines whether the vehicle has a fault and the fault type according to a pre-set fault detection algorithm and threshold.

[0126] The present embodiments define the following multi-level fault tolerance mechanism for the main faults that may occur in actual application scenarios:

[0127] Level 1 (DPQP solving timeout): When the fault diagnosis system detects the fault of DPQP solving timeout, the control decision module of the vehicle immediately responds to switch the originally used complex DPQP model to a simplified QP model. The simplified QP model reduces the number of state variables to reduce the computational complexity, thereby avoiding control delay caused by solving timeout and ensuring the timeliness of the vehicle's intersection passing decision.

[0128] Level 2 (V2I communication interruption): Once the V2I communication interruption is detected, the vehicle quickly adjusts its perception and decision strategy. In terms of perception, it switches to a visual-radar fusion mode to perceive the environment using camera and millimeter wave radar data. In terms of signal light state processing, the yellow light is regarded as a red light, and the green light flashing is regarded as a yellow light to enhance safety. For the countdown information, since it cannot be obtained through V2I communication, the vehicle estimates the current countdown time by visually recognizing the current countdown display and combining the historical data interpolation of the countdown change rate in the last 10 seconds, to provide a basis for intersection passing decision.

[0129] Level 3 (partial sensor failure):

[0130] Laser radar failure: If the fault diagnosis system determines that the laser radar is failed, the vehicle immediately closes the occupancy grid map function, because this function highly depends on laser radar data. At this time, the vehicle only relies on the camera and millimeter wave radar for environmental perception, and at the same time reduces the perception range to adapt to the effective detection distance of the remaining sensors, to ensure safe passing decisions within the limited perception range.

[0131] Camera failure: when the camera fails, the vehicle uses millimeter wave radar to detect the motion state of the vehicle queue, such as the driving speed and distance of the vehicle in front. At the same time, combined with V2I data (if communication is not interrupted) to infer the signal light state, make up for the information loss caused by camera failure, and ensure the normal traffic decision of the vehicle at the intersection.

[0132] Level4 (full perception failure): if all perception sensors of the vehicle fail, i.e. full perception failure, the vehicle will trigger the minimum risk state (MRC). At this time, the control module of the vehicle issues instructions to make the vehicle stop on the roadside at a slow speed, and the target speed is set to not more than 10km / h, for example, to ensure that the vehicle can safely stop on the roadside in the case of losing effective perception ability, and avoid causing danger to other traffic participants.

[0133] Of course, it should be noted that how to define and divide the fault level can be adjusted flexibly by those skilled in the art according to actual needs, and is not limited here.

[0134] Through the multi-level fault tolerance mechanism, corresponding measures are taken for different types of faults to ensure that the autonomous vehicle will not lose control immediately when a fault occurs, but can be dynamically adjusted according to the fault level to maintain a certain degree of normal operation ability, greatly reducing the risk of traffic accidents caused by faults, and improving the reliability and safety of the entire autonomous driving system.

[0135] When a fault occurs, the vehicle can quickly switch to the appropriate operating mode and decision-making strategy according to the fault type, avoiding the vehicle from stopping at the intersection due to the fault, thereby ensuring the traffic capacity of the vehicle at the intersection, reducing the impact on traffic flow, maintaining the smooth operation of traffic, and improving the user's ride experience.

[0136] In some embodiments of the present application, based on the pedestrian priority braking strategy, it is detected in real time whether the pedestrian invades the vehicle path and the longitudinal distance from the vehicle is close (even if the signal light is green), and if so, the emergency braking is triggered immediately, i.e. the longitudinal deceleration reaches a certain value, to ensure that the vehicle stops a certain distance in front of the pedestrian.

[0137] The laser radar equipped in the autonomous vehicle can continuously scan the environment around the vehicle, collect laser radar point cloud data, process the collected laser radar point cloud data by using a DBSCAN clustering algorithm, and cluster the point cloud data of different objects (such as pedestrians, non-motor vehicles, vehicles, buildings, etc.) into different clusters. Feature analysis and identification are performed on each cluster to determine whether it represents a pedestrian or a non-motor vehicle. After detecting pedestrians and non-motor vehicles, based on their historical position information and current motion state, a suitable motion model (such as uniform motion model, uniform acceleration motion model, etc.) is used to predict their future motion trajectory, providing a basis for the safety decision of the vehicle.

[0138] The relationship between the motion trajectory of the pedestrian and non-motor vehicle and the driving path of the vehicle is monitored in real time. By comparing and analyzing the predicted motion trajectory of the pedestrian and non-motor vehicle with the pre-planned driving path of the vehicle, it is determined whether the pedestrian or non-motor vehicle has invaded the vehicle path. When a pedestrian is detected to have invaded the vehicle path and the longitudinal distance between the pedestrian and the vehicle is close (even if the signal light is green at this time), an emergency braking system is triggered immediately. The emergency braking system quickly adjusts the braking device of the vehicle, causing the vehicle to generate a larger longitudinal deceleration, for example, which can reach -4 m / s 2 .

[0139] During the emergency braking process, the control module of the vehicle continuously monitors the driving speed of the vehicle and the distance to the pedestrian. By precisely controlling the braking system, it is ensured that the vehicle can safely stop at a certain distance, such as 5 meters, in front of the pedestrian. This requires real-time adjustment of the brake pressure to change the deceleration of the vehicle according to the predetermined strategy, avoiding the vehicle from losing control due to excessive braking or not stopping in time due to insufficient braking.

[0140] The above technical solution effectively avoids the collision between the vehicle and the pedestrian, greatly reduces the risk of injury to the pedestrian, provides reliable safety protection for the pedestrian, enhances the intelligence and adaptability of autonomous driving decision-making, and optimizes traffic safety and smoothness.

[0141] For the convenience of understanding the above embodiments, the specific implementation of the present application is further described taking a L4 level autonomous vehicle in a green light countdown scenario as an example:

[0142] (1) Initialization configuration:

[0143] Load high-precision map and calibrate absolute position of signal light (error ≤10 cm).

[0144] Calibrate sensors: adjust the focal length of the camera to cover a range of 200 m, and the laser radar point cloud registration error is ≤2 cm.

[0145] (2) Decision layer (DPQP):

[0146] Input: V2I module receives signal light data at 10Hz frequency: green light countdown remaining 15 seconds. Map and localization module obtains vehicle position to stop line distance, e.g. 50m. DP stage generates coarse solution (speed curve v(t)), QP stage optimizes to smooth curve (acceleration fluctuation <0.5m / s 2 ) per second).

[0147] (3) Control layer (MPC):

[0148] According to the reference speed v_ref(t) output by DPQP, the optimal control amount (a, δ) is solved by rolling. The actual tracking error is <0.1m, and the acceleration change rate is <1m / s 3 .

[0149] In summary, the intersection passing method of the autonomous vehicle proposed in the application is a "three-layer decision architecture", which realizes efficient and safe intersection passing control through the cooperative optimization of the perception layer, the decision layer and the control layer, wherein:

[0150] (1) Perception layer:

[0151] Mainly used for fusing V2I communication, multi-sensor data (laser radar, millimeter wave radar, camera) and high-precision map, to realize centimeter-level signal light positioning and traffic participant perception. A bidirectional verification mechanism is introduced: the vehicle sends positioning data to the roadside unit (RSU), and the RSU returns a signal light state verification code to ensure data integrity.

[0152] (2) Decision layer:

[0153] Adopting longitudinal dynamic programming and quadratic programming (DPQP), the speed curve is optimized in stages to meet safety and comfort constraints. A multi-level fault tolerance strategy is defined: according to the fault type (communication interruption, sensor failure, pedestrian intrusion), the decision priority is dynamically adjusted.

[0154] (3) Control layer:

[0155] Based on model predictive control (MPC), the optimal control input is solved by rolling to realize trajectory tracking and disturbance suppression.

[0156] The key points and technical effects of the intersection passing method of the autonomous vehicle of the application mainly lie in:

[0157] (1) Multi-sensor fusion perception and recognition method: improve signal light and pedestrian, motor vehicle, non-motor vehicle recognition rate, multi-module dynamic cooperation, improve vehicle intersection passing efficiency.

[0158] (2) Dynamic virtual obstacle: based on countdown and vehicle dynamics model generation, supporting smooth deceleration.

[0159] (3) Real-time speed planning: Deep coordination between longitudinal DPQP and MPC, making up for the shortcomings of traditional methods in real-time performance, control accuracy and comfort.

[0160] (4) Complete emergency response: Define multi-level fault-tolerant rules to ensure minimum risk control in extreme scenarios.

[0161] The application can be applied to automatic driving systems in urban roads, highway ramps and mixed traffic scenarios, and has wide application prospects.

[0162] The application also provides an intersection passing device 200 of an autonomous vehicle, as shown in Figure 2 The structure diagram of the intersection passing device of the autonomous vehicle in the embodiment of the application is provided, and the intersection passing device 200 of the autonomous vehicle comprises an acquisition unit 210, a first determination unit 220, a second determination unit 230 and a decision control unit 240, wherein:

[0163] The acquisition unit 210 is configured to acquire multi-sensor perception data and positioning data of the autonomous vehicle, and receive dynamic traffic light data of an intersection area where the autonomous vehicle is located from a road end;

[0164] The first determination unit 220 is configured to determine a current traffic light scene and corresponding traffic light data of the autonomous vehicle according to the multi-sensor perception data and the dynamic traffic light data;

[0165] The second determination unit 230 is configured to determine a passing strategy of the autonomous vehicle in the intersection area according to the positioning data and the current traffic light scene and corresponding traffic light data of the autonomous vehicle;

[0166] The decision control unit 240 is configured to determine a current control parameter of the autonomous vehicle by using a preset decision control algorithm according to the passing strategy of the autonomous vehicle in the intersection area, and control the passing of the autonomous vehicle in the intersection area according to the current control parameter.

[0167] In some embodiments of the application, the acquisition unit 210 is specifically configured to receive the dynamic traffic light data of the intersection area where the autonomous vehicle is located from the road end based on an optimized V2I communication protocol; wherein the optimized V2I communication protocol is defined as dynamically adjusting the transmission frequency and transmission field of the traffic light data based on the distance mode of the autonomous vehicle to the intersection area.

[0168] In some embodiments of the present application, the multi-sensor perception data includes visual perception data, and the first determination unit 220 is specifically configured to: perform signal light recognition according to the visual perception data to obtain a first visual-based signal light recognition result; compare the first visual-based signal light recognition result with the dynamic signal light data; if the first visual-based signal light recognition result is inconsistent with the dynamic signal light data, perform auxiliary verification using the visual perception data, and determine the current signal light scene and corresponding signal light data of the autonomous vehicle according to an auxiliary verification result; otherwise, directly determine the current signal light scene and corresponding signal light data of the autonomous vehicle according to the first visual-based signal light recognition result or the dynamic signal light data.

[0169] In some embodiments of the present application, the first determination unit 220 is specifically configured to: detect a motion state of a surrounding vehicle according to the visual perception data; determine a second visual-based signal light recognition result according to the motion state of the surrounding vehicle; and determine the current signal light scene and corresponding signal light data of the autonomous vehicle according to the second visual-based signal light recognition result.

[0170] In some embodiments of the present application, the second determination unit 230 is specifically configured to: calculate a distance from the autonomous vehicle to a stop line according to the positioning data and stop line data of the intersection region; if the current signal light scene is a green light scene, calculate a driving speed safety range of the autonomous vehicle in the current signal light scene according to the distance from the autonomous vehicle to the stop line and the signal light data, and determine a passing strategy of the autonomous vehicle in the green light scene according to a current vehicle speed of the autonomous vehicle, a maximum length of a planned trajectory, the distance from the autonomous vehicle to the stop line, and a safety margin of the green light scene; if the current signal light scene is a red light scene or a yellow light scene, determine a passing strategy of the autonomous vehicle in the red light scene or the yellow light scene according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance from the autonomous vehicle to the stop line, and a safety margin of the red light scene or the yellow light scene.

[0171] In some embodiments of the present application, the second determination unit 230 is specifically configured to: determine whether deceleration braking is needed according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance from the autonomous vehicle to the stop line, and a safety margin of the current signal light scene; if yes, establish a virtual obstacle to perform braking, calculate a braking distance, and determine a deceleration passing strategy according to the braking distance and the distance from the autonomous vehicle to the stop line; otherwise, determine that a passing strategy of the autonomous vehicle in the current signal light scene is a normal passing strategy.

[0172] In some embodiments of the present application, the intersection passing device 200 of the autonomous vehicle further comprises a detection unit configured to detect whether the autonomous vehicle has a fault; and a dynamic decision control unit configured to, in the case that the autonomous vehicle has a fault, dynamically decide and control the autonomous vehicle according to the fault type by using a pre-defined multi-level fault-tolerant mechanism and an emergency braking strategy.

[0173] It can be understood that the intersection passing device of the autonomous vehicle described above can realize each step of the intersection passing method of the autonomous vehicle provided in the foregoing embodiments, and the related explanations about the intersection passing method of the autonomous vehicle are all applicable to the intersection passing device of the autonomous vehicle, which will not be repeated here.

[0174] Figure 3 is a structural schematic diagram of an apparatus in an embodiment of the present application. As shown in Figure 3 the apparatus includes one or more processors (or processing units), can further include one or more memories coupled to the processors, and can further include a communication module coupled to the processors.

[0175] The communication module can be used for communication with other devices or apparatuses, such as transmission or reception of data and / or signals. The communication module can have at least one communication module for communication. The communication module can include any interface necessary for communication with other devices. Illustratively, the communication module can be a transceiver, a circuit, a bus, a module, or other types of communication modules.

[0176] The processor can include, but is not limited to, at least one of the following: a general-purpose computer, a special-purpose computer, a microcontroller, a digital signal controller (Digital Signal Processor, DSP), or one or more of a controller-based multi-core controller architecture. The apparatus can have multiple processors, such as application-specific integrated circuit chips, which are time-dependent on a clock synchronized with the main processor.

[0177] The memory can include one or more non-volatile memories and one or more volatile memories. Examples of non-volatile memories include, but are not limited to, at least one of the following: Read-Only-Memory (ROM), Electrically Programmable Read-Only-Memory (EPROM), flash memory, hard disk, Compact Disc (CD), Digital Video Disk (DVD), or other magnetic and / or optical storage. Examples of volatile memories include, but are not limited to, at least one of the following: Random Access Memory (RAM), or other volatile memory that does not persist in the duration of a power failure.

[0178] The computer program includes computer-executable instructions executed by an associated processor. The program can be stored in the ROM. The processor can perform any suitable action and processing by loading the program into the RAM.

[0179] Possible implementations of the present application can be realized by means of a program and therefore the communication device can perform any process as discussed in the preceding embodiments. Possible implementations of the present application can also be realized in hardware or in a combination of software and hardware.

[0180] In some embodiments, the program can be tangibly embodied in a computer-readable storage medium, which can include other storage devices that can be included in or accessed by the device, such as in the memory. The program can be loaded from the computer-readable storage medium to the RAM for execution. The computer-readable storage medium can include any type of tangible non-volatile memory, such as ROM, EPROM, flash memory, hard disk, CD, DVD, and the like.

[0181] The embodiments of the present application further provide a computer readable storage medium having computer instructions or program codes stored thereon, which, when executed by a processor, cause the processor to perform the methods and functions involved in any of the above embodiments. The computer readable medium can be any tangible medium containing or storing a program for or about an instruction execution system, apparatus or device. The computer readable medium can be a computer readable signal medium or a computer readable storage medium. The computer readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared or semiconductor system, apparatus or device, or any suitable combination thereof. The computer readable storage medium can be any available medium accessible by a computer or data storage device such as a server, data center, etc. integrated with one or more available media. More detailed examples of the computer readable storage medium include an electrical connection with one or more wires, a magnetic medium (e.g., a disk, a floppy disk, a hard disk, a magnetic tape, a magnetic storage device), an optical medium (e.g., an optical storage device, a DVD), a semiconductor medium (e.g., a solid-state hard disk), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), or any suitable combination thereof, etc.

[0182] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The embodiments of the present application also provide at least one computer program product tangibly stored on a non-transitory computer readable storage medium. The computer program product includes one or more computer executable instructions, such as instructions included in program modules, which are executed in a device on a real or virtual processor of a target to perform the processes, methods and functions involved in any of the above embodiments. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present application are generated. The computer can be a general purpose computer, a special purpose computer, a computer network or other programmable device. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line) or wireless (such as infrared, wireless, microwave, etc.) means.

[0183] The embodiments of the present application further provide a computer program product, comprising computer programs or instructions, which, when executed on a computer, cause the computer to perform the processes, methods and functions in the above embodiments. Generally, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform particular tasks or implement particular abstract data types. In various embodiments, the functions of program modules can be combined or divided among program modules as desired. Machine-executable instructions for program modules can be executed within a local or distributed device. In a distributed device, program modules can be located in local and remote storage media.

[0184] Generally, various embodiments of the present application can be implemented in hardware or special-purpose circuits, software, logic or any combination thereof. Some aspects can be implemented in hardware, while other aspects can be implemented in firmware or software which can be executed by a controller, microprocessor or other computing device. While various aspects of embodiments of the present disclosure are illustrated and described as block diagrams, flow charts, or using some other pictorial representation, it is well understood that these blocks, apparatus, systems, techniques or methods described herein can be implemented in, as non-limiting examples, hardware, software, firmware, special-purpose circuits or logic, general purpose hardware or controller or other computing devices, or some combination thereof.

[0185] It should be noted that although the embodiments of the present application are described above respectively in connection with the drawings, the above embodiments are not independent of each other, and they can also be combined to obtain other embodiments. The manners, cases, categories and division of embodiments in the embodiments of the present application are only for the convenience of description, and should not constitute special limitation. The features in various manners, categories, cases and embodiments can be combined with each other as long as they are logically consistent. The various embodiments of the present application can be combined arbitrarily to achieve different technical effects. The embodiments of the present application do not list various combinations again.

[0186] In addition, although the operations of the methods of the present disclosure are described in a particular order in the drawings, this does not require or imply that the operations must be performed in that particular order, or that all of the illustrated operations must be performed to achieve the desired results. On the contrary, the steps depicted in the flowcharts can change the order of execution. Additionally or alternatively, some steps can be omitted, combined into one step, and / or decomposed into multiple steps. It should also be noted that the features and functions of two or more devices according to the present disclosure can be embodied in one device. Conversely, the features and functions of one device described above can be further divided into multiple devices.

[0187] It should also be noted that the terms "comprising," "including," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises a... " does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0188] The above description is merely illustrative of the application, and not restrictive. Various modifications and changes can become apparent to those skilled in the art. Incorporating any modification, equivalent substitution, improvement, etc. within the spirit and principle of the application, shall be included in the scope of the claims of the application.

Claims

1. A method of passing through an intersection of an autonomous vehicle, characterized by, The intersection passing method of the automatic driving vehicle comprises: acquiring multi-sensor perception data and positioning data of the automatic driving vehicle, and receiving dynamic signal lamp data of an intersection area where the automatic driving vehicle is located from a road end; determining a current signal lamp scene and corresponding signal lamp data where the automatic driving vehicle is located according to the multi-sensor perception data and the dynamic signal lamp data; determining a passing strategy of the automatic driving vehicle in the intersection area according to the positioning data and the current signal lamp scene and corresponding signal lamp data where the automatic driving vehicle is located; determining a current control parameter of the automatic driving vehicle by using a preset decision control algorithm according to the passing strategy of the automatic driving vehicle in the intersection area, and controlling the passing of the automatic driving vehicle in the intersection area according to the current control parameter.

2. The intersection passing method of the autonomous vehicle according to claim 1, wherein The receiving of the dynamic signal lamp data of the intersection area where the automatic driving vehicle is located from the road end comprises: receiving the dynamic signal lamp data of the intersection area where the automatic driving vehicle is located from the road end based on an optimized V2I communication protocol; wherein the optimized V2I communication protocol is defined as dynamically adjusting a transmission frequency and a transmission field of the signal lamp data based on a distance mode of the automatic driving vehicle to the intersection area.

3. The intersection passing method of the autonomous vehicle according to claim 1, wherein The multi-sensor perception data comprises visual perception data, and the determining of the current signal lamp scene and corresponding signal lamp data where the automatic driving vehicle is located according to the multi-sensor perception data and the dynamic signal lamp data comprises: performing signal lamp recognition according to the visual perception data to obtain a first visual-based signal lamp recognition result; comparing the first visual-based signal lamp recognition result with the dynamic signal lamp data; if the first visual-based signal lamp recognition result is inconsistent with the dynamic signal lamp data, performing auxiliary verification by using the visual perception data, and determining the current signal lamp scene and corresponding signal lamp data where the automatic driving vehicle is located according to an auxiliary verification result; otherwise, directly determining the current signal lamp scene and corresponding signal lamp data where the automatic driving vehicle is located according to the first visual-based signal lamp recognition result or the dynamic signal lamp data.

4. The intersection passing method of the autonomous vehicle according to claim 3, wherein The auxiliary verification by using the visual perception data and the determination of the current signal lamp scene and corresponding signal lamp data where the automatic driving vehicle is located according to an auxiliary verification result comprise: detecting a motion state of a surrounding vehicle according to the visual perception data; determining a second visual-based signal lamp recognition result according to the motion state of the surrounding vehicle; determining the current signal lamp scene and corresponding signal lamp data where the automatic driving vehicle is located according to the second visual-based signal lamp recognition result.

5. The intersection passing method of the autonomous vehicle according to claim 1, wherein The determining of the passing strategy of the automatic driving vehicle in the intersection area according to the positioning data and the current signal lamp scene and corresponding signal lamp data where the automatic driving vehicle is located comprises: calculating a distance of the automatic driving vehicle to a stop line according to the positioning data and stop line data of the intersection area; if the current signal light scene is a green light scene, calculating a driving speed safety range of the autonomous vehicle in the current signal light scene according to the distance of the autonomous vehicle to the stop line and the signal light data, and determining a passing strategy of the autonomous vehicle in the green light scene according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line and the safety margin of the green light scene; if the current signal light scene is a red light scene or a yellow light scene, determining a passing strategy of the autonomous vehicle in the red light scene or the yellow light scene according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line and the safety margin of the red light scene or the yellow light scene.

6. The intersection passing method of an autonomous vehicle according to claim 5, wherein the determining of the passing strategy of the autonomous vehicle in the intersection region according to the positioning data, the current signal light scene and the corresponding signal light data of the autonomous vehicle comprises: determining whether the autonomous vehicle needs to slow down and brake according to the current vehicle speed of the autonomous vehicle, the maximum length of the planned trajectory, the distance of the autonomous vehicle to the stop line and the safety margin of the current signal light scene; if yes, establishing a virtual obstacle to brake, calculating a braking distance and determining a slow-down passing strategy according to the braking distance and the distance of the autonomous vehicle to the stop line; otherwise, determining that the passing strategy of the autonomous vehicle in the current signal light scene is a normal passing strategy.

7. The method according to any one of claims 1 to 6, wherein the intersection passing method of the autonomous vehicle further comprises: detecting whether the autonomous vehicle has a fault; in the case that the autonomous vehicle has a fault, making a dynamic decision and control on the autonomous vehicle according to the fault type by using a pre-defined multi-level fault-tolerant mechanism and an emergency braking strategy.

8. An intersection passing device of an autonomous vehicle, characterized by, the intersection passing device of the autonomous vehicle comprises: an acquisition unit configured to acquire multi-sensor perception data and positioning data of an autonomous vehicle, and receive dynamic signal light data of an intersection region where the autonomous vehicle is located from a road end; a first determination unit configured to determine a current signal light scene and corresponding signal light data of the autonomous vehicle according to the multi-sensor perception data and the dynamic signal light data; a second determination unit configured to determine a passing strategy of the autonomous vehicle in the intersection region according to the positioning data, the current signal light scene and the corresponding signal light data of the autonomous vehicle; a decision control unit configured to determine a current control parameter of the autonomous vehicle by using a pre-set decision control algorithm according to the passing strategy of the autonomous vehicle in the intersection region, and control the passing of the autonomous vehicle in the intersection region according to the current control parameter.

9. An apparatus comprising: a processor; and a memory arranged to store computer executable instructions that, when executed, cause the processor to perform the intersection passing method of the autonomous vehicle according to any one of claims 1-7.

10. A computer program product comprising computer programs / instructions, characterized in that, The computer program / instruction is executed by the processor to implement the intersection passing method of the autonomous vehicle according to any one of claims 1-7.

Images