Intersection passing intelligent decision method and device considering surrounding real-time traffic flow

Abstract

The application discloses an intersection passing intelligent decision-making method and device considering surrounding real-time traffic flow, and the method comprises the following steps: a sensor is arranged on an automatic driving vehicle to acquire state information; the vehicle is placed in a structured road and a lane-level relative relationship is constructed in combination with high-precision map information; the distance of a front intersection is acquired according to the state of the vehicle and the high-precision map, and it is judged whether the distance is greater than a set threshold value; if yes, the vehicle is driven on a normal public road, otherwise, surrounding real-time traffic flow estimation is performed; the required entrance and exit and the corresponding lane are determined according to the passing direction of the front intersection; the preceding vehicle and the left and right vehicle groups are distinguished according to the above information; the vehicle group information is obtained according to the positions and historical trajectory points of the preceding vehicle and the left and right vehicle groups; the overall traffic flow information is obtained by summarizing the current and historical coordinates of the grouped vehicle groups; the driving trend curve is obtained by curve fitting according to the trajectory point set; and the intersection passing mode is selected by a decision-making module according to the surrounding vehicle trend and the exit lane information.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to an intelligent decision-making method and apparatus for intersection traffic that takes into account real-time traffic flow. Background Technology

[0002] As a typical application scenario for artificial intelligence technology, autonomous driving technology has developed rapidly in recent years. The ability to effectively handle highly dynamic and complex traffic scenarios is a key prerequisite for the widespread adoption of autonomous driving technology. Intersection traffic scenarios, due to the presence of multiple road intersections, diverse combinations of roads and signs, and constantly changing traffic signals, result in significant variations in vehicle and pedestrian traffic patterns, making it one of the dynamic and complex scenarios that autonomous vehicles need to overcome.

[0003] In existing technologies, autonomous vehicles, when navigating intersections, pre-select an exit lane based on the lanes they enter, and determine their path accordingly. Their interaction with the environment is primarily considered from a speed perspective. However, when the number of entry lanes differs from the number of exit lanes in the direction the autonomous vehicle is traveling through the intersection, pre-selecting the exit lane can lead to increased lane-changing and merging conflicts, impacting traffic safety and efficiency.

[0004] CN108459588B discloses an autonomous driving method, device, and vehicle. By considering traffic signals in the direction of the autonomous vehicle's travel, it determines the path from the vehicle to the target location at an intersection. It applies different acceleration curves to obtain a driving trajectory including speed and time, and then obtains the optimal trajectory decision based on a driving model, finally arriving smoothly at the intersection. This patent focuses on speed control of the autonomous vehicle before entering the intersection, enabling the vehicle to reach the intersection more smoothly and safely. However, it gives less consideration to the complex interactions that may occur within the intersection, such as merging with traffic flowing in the same direction, thus limiting the ability to navigate complex intersection scenarios.

[0005] CN110471415B discloses a vehicle with an autonomous driving mode and its control method and system. It primarily addresses the passage strategy of autonomous vehicles when traffic lights change at intersections, specifically determining whether to continue through the intersection based on the time it takes for the traffic light to change from yellow to red, combined with the vehicle's current location information. This patent also focuses on the decision of whether to pass through the intersection itself, with less consideration for the autonomous vehicle's passage patterns within the intersection, making it difficult to handle complex and ever-changing real-world intersection scenarios.

[0006] CN109582022B discloses an autonomous driving strategy decision-making system and method, proposing a method for autonomous driving strategy decision-making at intersections. It primarily considers whether the predicted driving intentions of multiple third-party vehicles around the autonomous vehicle overlap with the autonomous vehicle's driving intentions, thereby determining the autonomous vehicle's driving strategy, such as deceleration. This patent analyzes and considers third-party vehicles separately and relies on the accuracy of predictions for third-party vehicles, which may lead to an overly conservative approach, causing the autonomous vehicle to tend to choose unnecessary deceleration or waiting, reducing driving efficiency.

[0007] US Patent 20210278231A1 discloses an autonomous vehicle following system and method utilizing surrounding traffic flow. This system selects multiple nearby vehicles and integrates their driving patterns into a traffic flow trend, then controls the autonomous vehicle to follow this trend, thereby achieving collision-free and smooth driving. However, this patent, after acquiring the traffic flow, only uses following to coordinate the autonomous vehicle's movement with surrounding vehicles to achieve safe and smooth driving. It does not consider whether the vehicle's driving goal aligns with that of other vehicles, especially in situations where different vehicles at intersections exhibit significant differences in their driving patterns. This could lead to the autonomous vehicle being unable to rationally plan its path according to its own driving goal, reducing traffic efficiency. Summary of the Invention

[0008] To address the aforementioned technical problems, this invention proposes an intelligent decision-making method and device for intersection traffic that considers real-time traffic flow in the surrounding area. This method can solve the safety and efficiency problems of traffic flow within intersections, especially at intersections with different numbers of entrance and exit lanes, thereby reducing lane change and merging conflicts.

[0009] To achieve the above objectives, the technical solution of the present invention provides an intelligent decision-making method for intersection traffic considering real-time surrounding traffic flow, comprising the following steps: S1: acquiring the status information of the vehicle and surrounding vehicles through sensors equipped on the autonomous vehicle; S2: placing the vehicle and surrounding vehicles in a structured road by combining high-precision map information, acquiring the lane information of the vehicle and other vehicles, and constructing lane-level relative relationships; S3: acquiring the distance from the vehicle to the intersection ahead based on the vehicle status and high-precision map information, and determining whether the distance is greater than a set threshold. If so, proceeding according to normal public road traffic; otherwise, proceeding to step S4 to estimate the real-time surrounding traffic flow; S4: determining the required entrance and exit, as well as the same-direction traffic lane and exit lane information based on the traffic direction of the intersection ahead; S5: based on real-time perception information and in conjunction with lane information screening... Select vehicles traveling in the same direction as your vehicle. Then, based on the entrance and exit information, as well as the same-direction lane and exit lane information determined in step S4, and combined with the lane information of your vehicle obtained in step S2, distinguish between vehicles in front, vehicles on the left, and vehicles on the right; S6: Based on the current positions and historical trajectory points of the distinguished vehicles in front, on the left, and on the right, obtain information about the vehicle groups in front, on the left, and on the right relative to your vehicle; S7: By summarizing the current and historical coordinates of the corresponding vehicle groups, obtain the overall traffic flow information of the corresponding vehicle groups; S8: By performing curve fitting on the trajectory point set of each vehicle group, obtain the corresponding driving trend curve, thereby obtaining the driving trend curve of the surrounding traffic flow of your vehicle; S9: Select the mode of travel within the intersection based on the surrounding traffic flow driving trend curve and the exit lane information.

[0010] Furthermore, in step S7, when summarizing historical coordinates, only the coordinate information of the corresponding vehicle entering the set threshold range mentioned in step S3 is included.

[0011] Furthermore, in step S8, when performing curve fitting on the trajectory point set, a local coordinate system is first constructed with the vehicle as the center, and the trajectory point sets of different groups of vehicles are transformed into this local coordinate system. Then, a curve fitting problem is constructed in this local coordinate system, and the corresponding fitting curve is obtained by solving it.

[0012] Furthermore, in step S8, assuming the vehicle's trajectory follows a fourth-order polynomial curve, a point on the trajectory is represented by the following expression: Where, x i ,y i Let β be the coordinate value of the i-th trajectory point in the vehicle coordinate system. j (j=0...4) are the fitting parameters; first, all trajectory points of the same group of vehicles are written in vector form. Where X is a polynomial matrix composed of different orders of the x-coordinates of all trajectory points. To fit the parameters, where y is a column vector consisting of the y-coordinates of all trajectory points, the objective function is established as follows: Find the coefficients of the polynomial Next, the starting and ending values ​​of x are set according to the range of the intersection, and the corresponding step size is determined. Then, the corresponding y value is calculated through a polynomial expression, thus obtaining the corresponding fitted traffic flow trend curve.

[0013] Furthermore, in step S1, the sensors equipped with the autonomous vehicle include: LiDAR, high-definition camera, and IMU / GNSS device; the vehicle's state information includes: the vehicle's pose, velocity vector, and acceleration vector; the state information of surrounding vehicles includes: the type, pose, speed, and relative relationship between the other vehicles and the vehicle.

[0014] Furthermore, in step S2, the relative relationships between other vehicles and the vehicle itself include: the vehicle in front in the same lane; the vehicle behind in the same lane; the vehicle in front, the vehicle parallel to, and the vehicle behind in the right lane; the vehicle in front, the vehicle parallel to, and the vehicle behind in the left lane; and when there are more lanes on the same-direction road and other vehicles in the corresponding lanes, the relative relationships between other vehicles and the vehicle itself are further extended to include all lanes in the same direction.

[0015] Furthermore, in step S2, if there is overlap between roads in different directions within the intersection, resulting in multiple roads within the intersection corresponding to the vehicle's location, and making the correspondence between other vehicles and the vehicle unclear, it is necessary to trace back the historical lane information of other vehicles and mark them based on the relative relationship between the last non-intersection lane the vehicle was in before entering the intersection and the lane where the vehicle is located.

[0016] Furthermore, in step S9, the travel method includes: the travel distance, the feasible trajectory, and the selection of the exit lane.

[0017] The technical solution of this invention also provides an intelligent decision-making device for intersection traffic considering real-time surrounding traffic flow, which includes the following modules: a status information acquisition module: acquiring the status information of the vehicle and surrounding vehicles through sensors equipped on the autonomous vehicle; a lane-level relative relationship construction module: placing the vehicle and surrounding vehicles in a structured road by combining high-precision map information, acquiring the lane information of the vehicle and other vehicles, and constructing lane-level relative relationships; a threshold judgment module: acquiring the distance from the vehicle to the intersection ahead based on the vehicle status and high-precision map information, and judging whether the distance is greater than a set threshold. If so, the vehicle proceeds according to normal public road traffic; otherwise, real-time surrounding traffic flow estimation is required; an entrance / exit and lane determination module: when the threshold judgment module determines that real-time surrounding traffic flow estimation is required, determining the required entrance and exit, as well as the same-direction traffic lane and exit lane information based on the direction of traffic at the intersection ahead; and a vehicle differentiation module: based on real-time perception information and in conjunction with lane information... The system filters out vehicles traveling in the same direction as the vehicle. Then, based on the entrance and exit information, as well as the lane information for vehicles traveling in the same direction and the exit lane information determined by the entrance / exit and lane determination module, and combined with the lane information of the vehicle obtained from the lane-level relative relationship construction module, it distinguishes between vehicles in front, vehicles on the left, and vehicles on the right. The vehicle grouping module obtains information about the vehicle groups in front, on the left, and on the right relative to the vehicle based on the current positions and historical trajectory points of the distinguished vehicles. The traffic flow information acquisition module summarizes the current and historical coordinates of the corresponding vehicle groups to obtain the overall traffic flow information for each group. The driving trend curve fitting module performs curve fitting on the trajectory point set of each vehicle group to obtain the corresponding driving trend curve, thereby obtaining the driving trend curve of the surrounding traffic flow of the vehicle. The traffic mode selection module selects the traffic mode within the intersection based on the surrounding traffic flow driving trend curve and the exit lane information.

[0018] The technical solution of the present invention also provides a computer-readable storage medium containing a computer program, characterized in that, when the computer program is executed by one or more processors, it implements the intelligent intersection traffic decision-making method that takes into account the surrounding real-time traffic flow as described above. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the entrance and exit and the corresponding lanes of an embodiment of the present invention;

[0021] Figure 2A and Figure 2B The diagrams show the different driving modes of other vehicles traveling in the same direction as this vehicle;

[0022] Figure 3A and Figure 3B They respectively showed the same as Figure 2A and Figure 2B Example diagram of current and historical trajectory points for vehicle groups corresponding to the two traffic modes;

[0023] Figure 4 This diagram illustrates the overall driving trend obtained from trajectory points of different vehicle groups according to the present invention.

[0024] Figure 5 This is a flowchart of the intelligent intersection traffic decision-making method that takes into account the surrounding real-time traffic flow, as described in this invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The first aspect of this application proposes an intelligent intersection traffic decision-making method that considers real-time surrounding traffic flow, including:

[0027] 1. The current state of the autonomous vehicle and the state of surrounding vehicles are obtained through sensors equipped on the autonomous vehicle.

[0028] The sensors used include various types such as LiDAR, high-definition cameras, and IMU / GNSS devices.

[0029] For autonomous vehicles themselves, their current state contains several key pieces of information, such as the vehicle's pose, velocity vector, and acceleration vector.

[0030] The status information of surrounding vehicles includes the type, position, speed of other vehicles, and their relative relationship with the vehicle itself.

[0031] 2. By combining high-precision map information, the vehicle and surrounding vehicles are placed within a structured road, obtaining lane information for both the vehicle and other vehicles. Based on the vehicle's lane location, lane-level relative relationships are constructed to confirm the relative relationships between the vehicle and other vehicles. Possible relative relationships include:

[0032] The vehicle in front in the same lane, and there may be multiple vehicles, which have a relative front-to-back position relationship;

[0033] There may be multiple vehicles behind in the same lane, and there will be a corresponding relative front-to-back relationship;

[0034] The vehicle in front in the right lane, the vehicle parallel to the vehicle in the right lane, and the vehicle behind the vehicle in the right lane;

[0035] The vehicle in front in the left lane, the vehicle parallel to the vehicle in the left lane, and the vehicle behind the vehicle in the left lane;

[0036] If a road in the same direction has more lanes, and there are other vehicles in those lanes, then it can be further expanded to include right-right lanes, left-left lanes, etc., until all lanes in the same direction are covered.

[0037] Given the overlap of roads leading in different directions within an intersection, resulting in multiple roads corresponding to a vehicle's location, and making the relationship between the vehicle and other vehicles unclear, it is necessary to trace back the historical lane information of other vehicles. Generally, this can be done by marking the relative relationship between the last non-intersection lane the vehicle was in before entering the intersection and the lane the vehicle is in, thereby ensuring a clear judgment of the relative relationship between vehicles.

[0038] 3. Based on the autonomous vehicle's own state and combined with a high-precision map, the distance from the autonomous vehicle to the upcoming intersection can be obtained, denoted as d. distance_to_crossroad If d distance_to_crossroad Greater than the set threshold d threshould If d distance_to_crossroad Less than the set threshold d threshould Then proceed to the surrounding real-time traffic flow estimation step.

[0039] 4. The above-described real-time traffic flow estimation process involves several important steps. First, the entrances and exits of the intersection and their corresponding lane information are determined. Then, based on real-time perception information and lane information, vehicles traveling in the same direction as the current vehicle are selected, thereby constructing the real-time traffic flow. Finally, the decision-making system intelligently determines the traffic mode at the intersection based on the constructed real-time traffic flow. Each step will be described in detail below.

[0040] 5. Determine the vehicle's entrance and exit points based on the required travel direction at the intersection ahead. For example... Figure 1 As shown, taking the case of an autonomous vehicle going straight at an intersection as an example, the lane in which the vehicle is located determines the entrance to the intersection. Based on the vehicle's need to go straight and the location of the entrance, the lane directly opposite can be determined as the exit.

[0041] 6. By utilizing high-precision map information and considering both entrance and traffic direction, it is possible to identify lanes for traffic traveling in the same direction. For example... Figure 1As shown, there are three straight lanes at the entrance: straight lane 1, straight lane 2, and right-turn straight lane, from left to right. It's important to note that the left-turn lane's direction of travel differs from that of autonomous vehicles; therefore, although located at the corresponding entrance, it is excluded from the calculation.

[0042] 7. By utilizing high-precision map information and combining it with exit information, possible exit lanes can be determined. For example... Figure 1 As shown, there are four lanes at the exit, from left to right: left 1, left 2, left 3, and left 4. These lanes can all be used as target lanes for autonomous vehicles.

[0043] 8. When the number of lanes at the entrance and exit differs, autonomous vehicles can choose from multiple lanes at the exit when passing through an intersection. For example, in... Figure 1 In this scenario, depending on the entrance lane where the autonomous vehicle is located, choosing either the second or third lane from the left at the exit is feasible. Whether this choice is appropriate depends primarily on the driving patterns of other vehicles traveling in the same direction. This can be determined through... Figure 2A and Figure 2B This method presents two scenarios. If only each vehicle is considered individually, the decision on traffic flow is easily influenced by the behavior of a single vehicle, leading to unstable decision results. However, this method considers vehicles traveling in the same direction as a whole to obtain stable traffic flow information, which is then input into the decision-making system to provide a more intelligent decision output. For example, for the same intersection, based on the overall driving situation of surrounding vehicles (such as...),... Figure 2A and Figure 2B As shown, autonomous vehicles need to select different exit lanes to reduce intersections with other vehicles, thereby improving the safety and efficiency of traffic flow at intersections.

[0044] 9. Based on the entrance and its corresponding lane information determined in steps 5 to 7 above, and combined with the lane where the autonomous vehicle is located as described in step 2, the vehicles ahead, to the left, and to the right can be identified. For example... Figure 2A and Figure 2B In the diagram, vehicle 3 (also marked "left") is identified as a left-hand vehicle. It's important to note that vehicles 4 and 5 within the intersection are identified based on their historical lane positions before entering the intersection; therefore, both vehicles 4 and 5 can be identified as left-hand vehicles. Vehicle 6 (also marked "right") is a right-hand vehicle. Similarly, vehicles 7 and 8 within the intersection are identified as right-hand vehicles based on their historical lane positions before entering the intersection. Vehicles 1 and 2 are both identified as vehicles in front.

[0045] 10. Based on the current positions and historical trajectory points of the preceding vehicle and the left and right vehicle groups distinguished in step 9, information about the vehicle groups relative to the autonomous vehicle within the intersection can be obtained. This information includes: the preceding vehicle (e.g., ... Figure 2A and Figure 2B Vehicles 1 and 2 in the middle), vehicles on the left (e.g., vehicles on the right) Figure 2A and Figure 2B Vehicles 3, 4, and 5 in the middle and vehicles on the right (such as...) Figure 2A and Figure 2B (Vehicles 6, 7, and 8). Although the driving conditions of individual vehicles may vary, their overall driving trends are relatively stable.

[0046] a. By summarizing the current and historical coordinates of the corresponding vehicle groups, the overall traffic flow information for each group can be obtained. In this process, the historical coordinate set only covers the threshold d mentioned in step 3 when vehicles enter. threshould The information is within a certain range. The purpose of this is to avoid vehicle driving information unrelated to intersection traffic flow from interfering with the overall traffic flow trend.

[0047] b. Figure 3A and Figure 3B Gives a comparison with Figure 2A and Figure 2B Examples of current and historical trajectory points corresponding to the two traffic modes are shown below. The left side shows the overall straight-ahead situation, while the right side shows the overall leftward situation. Each trajectory point is formed by aggregating the trajectories of multiple vehicles in the same group. The density of these points will vary depending on the perceived frame rate in actual applications, but generally, it can ensure that the distance between points is kept within 0.1m.

[0048] c. All trajectory points here represent the actual trajectories traveled by the vehicle and do not contain any predictive information; therefore, their information is deterministic.

[0049] 11. The grouped trajectory points obtained in step 10 are composed of trajectory points from multiple vehicles. Although these trajectory points generally show a consistent trend, there may be some deviations in details. To provide more stable trend information, an additional layer of information processing is needed. One feasible method is to perform curve fitting on each group of trajectory points to obtain a curve that accurately reflects the driving trend.

[0050] Figure 4This demonstrates the overall driving trend obtained from the trajectory points of different vehicle groups. First, a local coordinate system is constructed centered on the autonomous vehicle. The trajectory point sets of different groups of vehicles are transformed into this local coordinate system. Then, a curve fitting problem is constructed in this local coordinate system, and the corresponding fitted curve is obtained by solving it. Least squares fitting will be introduced here as an optional fitting method.

[0051] Assume the vehicle's trajectory roughly follows a polynomial curve of a certain order, let's say order 4. Then, a point on the trajectory can be represented by the following expression: Where x i ,y i Let β be the coordinate value of the i-th trajectory point in the vehicle coordinate system. j (j = 0...4) are the fitting parameters. All trajectory points in the same set can be written in vector form. Where X is a polynomial matrix composed of different orders of the x-coordinates of all trajectory points. Let be the parameters to be fitted, and y be a column vector consisting of the y-coordinates of all trajectory points.

[0052] By establishing the following objective function The coefficients of the polynomial can then be determined. Next, based on the range of the intersection, set the starting and ending values ​​of x and determine the corresponding step size. Then, the corresponding y value can be calculated through a polynomial expression, thus obtaining the corresponding traffic flow fitting trend curve.

[0053] 12. After completing step 11 and obtaining the fitted real-time trend curve of the surrounding traffic flow, the decision-making module can intelligently select the mode of travel within the intersection based on the overall trend of surrounding vehicles and the location information of the exit lanes. This includes the travel distance, feasible trajectory, and selection of the exit lane. This approach reduces the possibility of potential interaction with vehicles traveling in the same direction, thereby improving traffic safety and efficiency.

[0054] The overall process of this invention is as follows: Figure 5 As shown.

[0055] Alternative implementation methods:

[0056] In this invention, information on vehicles traveling in the same direction can be obtained from the historical lane direction or from the similarity between the vehicle's driving posture and that of the vehicle itself.

[0057] In this invention, when obtaining the overall driving trend curve based on the set of vehicle trajectory points, different fitting methods can be used. For example, a similar variant involves introducing a weight vector W based on the reliability of the surrounding vehicle trajectory points and constructing a weighted least squares problem. By solving this problem, a more reasonable trend curve can be obtained when the weight vector W takes reasonable values.

[0058] The use of the overall driving trend curve of the surrounding vehicle group in this invention can also be extended to speed planning to guide autonomous vehicles to maintain coordination with the surrounding traffic flow during intersection passage.

[0059] The beneficial technical effects of the technical solution of this invention are as follows:

[0060] (1) Improved traffic efficiency and avoided unnecessary confluence and interaction;

[0061] (2) Improved traffic safety by taking into account the overall driving patterns of surrounding vehicles and avoiding potential interaction risks;

[0062] (3) It does not rely on probability prediction information;

[0063] (4) It increases the stability and reliability of intersection traffic decisions.

[0064] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent decision-making at intersections that considers real-time traffic flow, characterized in that, Includes the following steps: S1: Acquire the state information of the autonomous vehicle and surrounding vehicles through the sensors equipped on the autonomous vehicle. The state information of the autonomous vehicle includes: the vehicle's pose, velocity vector, and acceleration vector; the state information of surrounding vehicles includes: the type, pose, speed, and relative relationship between the vehicle and the autonomous vehicle. S2: Combine high-precision map information to place the vehicle and surrounding vehicles in a structured road, obtain lane information of the vehicle and other vehicles, and construct lane-level relative relationships; S3: Based on the vehicle status and high-precision map information, obtain the distance from the vehicle to the intersection ahead, and determine whether the distance is greater than the set threshold. If it is, proceed to normal public road driving; otherwise, proceed to step S4 to estimate the surrounding real-time traffic flow. S4: Determine the required entrance and exit, as well as the same-direction traffic lanes and exit lanes, based on the traffic direction at the intersection ahead; S5: Based on real-time perception information and lane information, select vehicles traveling in the same direction as the vehicle. Then, based on the entrance and exit information, as well as the same-direction lane and exit lane information determined in step S4, and combined with the lane information of the vehicle obtained in step S2, distinguish the vehicles in front, the vehicles on the left, and the vehicles on the right. S6: Based on the current position and historical trajectory points of the vehicles in front, on the left, and on the right, obtain information about the vehicle group in front, on the left, and on the right relative to the vehicle itself. S7: By summarizing the current and historical coordinates of the corresponding vehicle groups, the overall traffic flow information of the corresponding vehicle groups can be obtained. S8: By performing curve fitting on the trajectory point set of each group of vehicles, the corresponding driving trend curve is obtained, thereby obtaining the driving trend curve of the surrounding traffic flow of the vehicle. S9: Select the mode of travel within the intersection based on the surrounding traffic flow trend curve and exit lane information. The mode of travel includes: travel distance, feasible trajectory, and selection of exit lane.

2. The method according to claim 1, characterized in that, In step S7, when summarizing historical coordinates, only the coordinate information of the corresponding vehicle entering the set threshold range mentioned in step S3 is included.

3. The method according to claim 1, characterized in that, In step S8, when performing curve fitting on the trajectory point set, a local coordinate system is first constructed with the vehicle as the center, and the trajectory point sets of different groups of vehicles are transformed into this local coordinate system. Then, a curve fitting problem is constructed in this local coordinate system, and the corresponding fitting curve is obtained by solving it.

4. The method according to claim 3, characterized in that, In step S8, assuming the vehicle's trajectory follows a fourth-order polynomial curve, a point on the trajectory is represented by the following expression: ,in, Let i be the coordinates of the i-th trajectory point in the vehicle coordinate system. These are the fitting parameters; First, express all trajectory points of the same vehicle group in vector form. ,in, It is a polynomial matrix composed of different orders of the x-coordinates of all trajectory points. For the parameters that need to be fitted, Let be a column vector consisting of the y-coordinates of all trajectory points. Then, the objective function is established as follows: Find the coefficients of the polynomial. ; Next, the starting and ending values ​​of x are set according to the range of the intersection, and the corresponding step size is determined. Then, the corresponding y value is calculated through a polynomial expression, thus obtaining the corresponding fitted traffic flow trend curve.

5. The method according to claim 1, characterized in that, In step S1, the sensors equipped in the autonomous vehicle include: LiDAR, high-definition camera, and IMU / GNSS device.

6. The method according to claim 1, characterized in that, In step S2, the relative relationships between other vehicles and the vehicle include: the vehicle in front in the same lane; the vehicle behind in the same lane; the vehicle in front, the vehicle parallel to the vehicle in the right lane, and the vehicle behind the vehicle in the right lane; the vehicle in front, the vehicle parallel to the vehicle in the left lane, and the vehicle behind the vehicle in the left lane. Furthermore, when there are more lanes on a road traveling in the same direction and other vehicles are present in those lanes, the relative relationship between the other vehicles and the vehicle is further extended to include all lanes traveling in the same direction.

7. The method according to claim 6, characterized in that, In step S2, if there is overlap between roads in different directions within the intersection, resulting in multiple roads within the intersection corresponding to the vehicle's location, and making the correspondence between other vehicles and the vehicle unclear, it is necessary to trace back the historical lane information of other vehicles and mark them based on the relative relationship between the last non-intersection lane the vehicle was in before entering the intersection and the lane where the vehicle is located.

8. An intelligent intersection traffic decision-making device that considers real-time surrounding traffic flow, characterized in that, Includes the following modules: State information acquisition module: Acquires state information of the autonomous vehicle and surrounding vehicles through sensors equipped on the autonomous vehicle. The state information of the autonomous vehicle includes: the vehicle's pose, velocity vector, and acceleration vector; the state information of surrounding vehicles includes: the type, pose, speed, and relative relationship between the vehicle and the autonomous vehicle. Lane-level relative relationship construction module: Combines high-precision map information to place the vehicle and surrounding vehicles in a structured road, obtains lane information of the vehicle and other vehicles, and constructs lane-level relative relationships; Threshold judgment module: Based on the vehicle status and high-precision map information, obtain the distance from the vehicle to the intersection ahead, and determine whether the distance is greater than the set threshold. If it is, drive on the normal public road; otherwise, estimate the surrounding real-time traffic flow. Entrance / Exit and Lane Determination Module: When the threshold judgment module determines that real-time traffic flow estimation of the surrounding area is required, the required entrance and exit, as well as the same-direction traffic lane and exit lane information, are determined based on the direction of traffic at the intersection ahead. Vehicle differentiation module: Based on real-time perception information and lane information, vehicles traveling in the same direction as the vehicle are filtered out. Then, based on the entrance and exit information, as well as the lane information of the same direction of travel and the exit lane information determined by the entrance and exit and lane determination module, and combined with the lane information of the vehicle obtained in the lane-level relative relationship construction module, the vehicle in front, the vehicle on the left and the vehicle on the right are differentiated. Vehicle grouping module: Based on the current position and historical trajectory points of the vehicles in front, on the left, and on the right, obtain information about the vehicle group in front, on the left, and on the right relative to the vehicle itself. Traffic flow information acquisition module: By summarizing the current and historical coordinates of the corresponding vehicle groups, the overall traffic flow information of the corresponding vehicle groups is obtained. Driving trend curve fitting module: By performing curve fitting on the trajectory point set of each group of vehicles, the corresponding driving trend curve is obtained, thereby obtaining the driving trend curve of the surrounding traffic flow of the vehicle. Traffic mode selection module: Selects the traffic mode within the intersection based on the surrounding traffic flow trend curve and exit lane information. The traffic mode includes: travel distance, feasible trajectory, and exit lane selection.

9. A computer-readable storage medium containing a computer program, characterized in that, When the computer program is executed by one or more processors, it implements the intelligent intersection traffic decision-making method that takes into account the surrounding real-time traffic flow, as described in any one of claims 1-7.

Images