Intersection lane selection method, device and equipment and readable storage medium

Abstract

The embodiment of the invention discloses an intersection lane selection method, device and equipment and a readable storage medium. The method comprises the following steps: acquiring map data corresponding to a current intersection, navigation information, own vehicle state information and obstacle state information; based on the map data, the navigation information, the self-vehicle state information and the obstacle state information, lane evaluation values of candidate lanes corresponding to the intersection where the self-vehicle is driven out are determined, and for each candidate lane, a passing evaluation value of the candidate lane is determined based on the lane evaluation value of the candidate lane; determining the regulation and control cost of each candidate lane based on a current trajectory planning algorithm and a control tracking algorithm; and determining an optimal lane in the candidate lanes according to the traffic evaluation value and the regulation and control cost of each candidate lane. By applying the scheme provided by the embodiment of the invention, the accuracy of intersection lane selection of the automatic driving vehicle can be improved, and the driving safety of the automatic driving vehicle is further ensured.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology. Specifically, it relates to a method, device, equipment and readable storage medium for selecting lanes at intersections. Background Art

[0002] During the driving process of an autonomous vehicle, when it passes through an intersection (such as an intersection or a roundabout), when determining the lane to enter the intersection, there are usually multiple exit lanes to choose from. In this case, since the road conditions of different lanes are different, such as the traffic flow size, obstacle distribution, etc., therefore, which lane to choose to exit will directly affect the traffic efficiency, safety when passing through the intersection and driving experience of the autonomous vehicle.

[0003] The known methods for selecting lanes at intersections usually learn the optimal lane selection results from expert driving data based on a network. However, when learning the optimal lane selection results from expert driving data, it is only obtained based on the evaluation of the complete driving behavior of the driver under a specific scenario. For a specific autonomous driving software system, this result may not necessarily be the optimal one. Specifically, after selecting the exit lane, trajectory planning and control tracking need to be carried out based on the selected lane. Trajectory planning and control tracking are completed by separate trajectory generation modules and control tracking modules. The trajectory generated by the trajectory generation module cannot guarantee to be close to the expert trajectory, and at the same time, the tracking of the trajectory by the control tracking module cannot maintain a high control accuracy like an expert driver. This may lead to the following problems: for a given optimal lane selection decision, there may be a situation where the vehicle collides with obstacles such as the road edge, fence, cone barrel, etc. during the intersection due to unreasonable trajectory planning results or excessive control tracking errors. Therefore, how to improve the accuracy of lane selection at intersections of autonomous vehicles and thus ensure the driving safety of autonomous vehicles has become an urgent technical problem to be solved. Summary of the Invention

[0004] This application provides a method, device, equipment and readable storage medium for selecting lanes at intersections to improve the accuracy of lane selection at intersections of autonomous vehicles and thus ensure the driving safety of autonomous vehicles. The specific technical solutions are as follows.

[0005] In a first aspect, an embodiment of this application provides a method for selecting lanes at intersections, and the method includes:

[0006] Obtain the map data corresponding to the current intersection, as well as navigation information, the state information of the vehicle itself, and the state information of obstacles;

[0007] Based on the map data, the navigation information, the ego-vehicle state information, and the obstacle state information, determine the lane evaluation values of each candidate lane corresponding to the ego-vehicle exiting the intersection, and for each candidate lane, determine the passing evaluation value of the candidate lane based on the lane evaluation value of the candidate lane;

[0008] Based on the current trajectory planning algorithm and control tracking algorithm, determine the planning and control costs of each candidate lane;

[0009] According to the passing evaluation values and planning and control costs of each candidate lane, determine the optimal lane among each candidate lane.

[0010] In the embodiments of the present application, when selecting lanes at an intersection, not only the lane evaluation values of each candidate lane, that is, the current road conditions of each candidate lane, can be considered, but also the planning and control costs of each candidate lane can be considered based on the current trajectory planning algorithm and control tracking algorithm. That is to say, if each candidate lane is used as the lane for exiting the intersection, the corresponding planning and control results can be obtained. Therefore, by selecting the optimal lane based on the lane evaluation values and planning and control costs of each candidate lane, it can be ensured that the road conditions of the lane for exiting the intersection finally determined are relatively good, and at the same time, the computing power of the current planning and control algorithm can be satisfied, avoiding the situation that the autonomous driving vehicle collides with obstacles such as road edges, fences, and cones during passing through the intersection, thereby improving the accuracy of the lane selection of the autonomous driving vehicle at the intersection and further ensuring the driving safety of the autonomous driving vehicle.

[0011] Optionally, before the step of determining the passing evaluation value of each candidate lane based on the lane evaluation value of the candidate lane, the method further includes:

[0012] Calculate the safety evaluation value of each candidate lane; the safety evaluation value includes: the trajectory safety evaluation result, and / or, the intersection collision risk;

[0013] The step of determining the passing evaluation value of each candidate lane based on the lane evaluation value of the candidate lane includes:

[0014] For each candidate lane, determine the passing evaluation value of the candidate lane based on the lane evaluation value and the safety evaluation value of the candidate lane.

[0015] Optionally, when the safety evaluation value includes the trajectory safety evaluation result and the intersection collision risk, the step of calculating the safety evaluation value of each candidate lane includes:

[0016] For each candidate lane, generate a planned trajectory from the current lane where the ego-vehicle is located to the candidate lane, and calculate the trajectory safety evaluation result corresponding to the planned trajectory;

[0017] Obtain the intersection collision risk values corresponding to each pre-stored intersection, and determine the intersection collision risk corresponding to the current intersection according to the intersection collision risk values corresponding to each intersection.

[0018] Optionally, the step of determining the lane evaluation values of each candidate lane corresponding to the vehicle exiting the intersection based on the map data, the navigation information, the vehicle state information, and the obstacle state information includes:

[0019] Determine the current lane where the vehicle is located, and according to the map data and the navigation information, determine each candidate lane corresponding to the vehicle exiting the intersection, and the target lane corresponding to the vehicle before reaching the next intersection;

[0020] For each candidate lane, calculate the first lane evaluation value from the current lane to the candidate lane according to the vehicle state information and the obstacle state information; and calculate the second lane evaluation value from the candidate lane to the target lane according to the vehicle state information and the obstacle state information; the first lane evaluation value includes a first traffic efficiency evaluation value, and / or, a first traffic safety evaluation value; the second lane evaluation value includes a second traffic efficiency evaluation value, and / or, a second traffic safety evaluation value;

[0021] For each candidate lane, calculate the lane evaluation value of the candidate lane according to the first lane evaluation value and the second lane evaluation value.

[0022] Optionally, the obstacle state information includes static obstacle state information and dynamic obstacle state information; when the first lane evaluation value includes a first traffic efficiency evaluation value and a first traffic safety evaluation value, the step of calculating the first lane evaluation value from the current lane to the candidate lane for each candidate lane according to the vehicle state information and the obstacle state information includes:

[0023] For each candidate lane, determine the smooth continuous curve length from the current lane to the candidate lane, and based on the smooth continuous curve length, determine the first traffic efficiency evaluation value from the current lane to the candidate lane; the first traffic efficiency evaluation value is negatively correlated with the smooth continuous curve length;

[0024] For each of the candidate lanes, calculate the static structure risk coefficient corresponding to the candidate lane according to the static obstacle state information, and calculate the dynamic traffic flow risk coefficient corresponding to the candidate lane according to the ego-vehicle state information and the dynamic obstacle state information; and calculate the first traffic safety evaluation value from the current lane to the candidate lane based on the static structure risk coefficient and the dynamic traffic flow risk coefficient; wherein, the first traffic safety evaluation value is negatively correlated with the static structure risk coefficient and the dynamic traffic flow risk coefficient; the static structure risk coefficient is related to the relative position between the static obstacle and the lane and the probability that the static obstacle is observed; the dynamic traffic flow risk coefficient is related to the time to collision between the ego-vehicle and the dynamic obstacle.

[0025] Calculate the first lane evaluation value from the current lane to the candidate lane according to the first traffic efficiency evaluation value and the first traffic safety evaluation value; the first lane evaluation value is positively correlated with both the first traffic efficiency evaluation value and the first traffic safety evaluation value.

[0026] Optionally, when the second lane evaluation value includes a second traffic efficiency evaluation value and a second traffic safety evaluation value, the step of calculating the second lane evaluation value from the candidate lane to the target lane according to the ego-vehicle state information and the obstacle state information includes:

[0027] Determine the minimum number of lane changes from the candidate lane to the target lane, and calculate the expected travel time and the expected lane change time corresponding to each lane passed by the ego-vehicle when passing from the candidate lane to the target lane through the minimum number of lane changes according to the ego-vehicle state information and the obstacle state information; wherein, the expected travel time corresponding to any lane is related to the expected travel distance of the ego-vehicle in that lane and the obstacle distribution in that lane; the expected lane change time corresponding to any lane is positively correlated with the length of the lane changeable interval in that lane.

[0028] Calculate the second traffic efficiency evaluation value from the candidate lane to the target lane according to the minimum number of lane changes, the expected travel time and the expected lane change time corresponding to each lane; the second traffic efficiency evaluation value is negatively correlated with the minimum number of lane changes, the expected travel time and the expected lane change time.

[0029] Calculate the second traffic safety evaluation value from the candidate lane to the target lane according to the minimum number of lane changes; the second traffic safety evaluation value is negatively correlated with the minimum number of lane changes.

[0030] Calculate the second lane evaluation value from the candidate lane to the target lane according to the second traffic efficiency evaluation value and the second traffic safety evaluation value; the second lane evaluation value is positively correlated with both the second traffic efficiency evaluation value and the second traffic safety evaluation value.

[0031] Optionally, the step of calculating the lane evaluation value of each candidate lane according to the first lane evaluation value and the second lane evaluation value includes:

[0032] For each candidate lane, determine a first weight for the first lane evaluation value and a second weight for the second lane evaluation value;

[0033] Based on the first weight and the second weight, weight the first lane evaluation value and the second lane evaluation value, and use the weighted result as the lane evaluation value of the candidate lane.

[0034] Optionally, the step of determining the lane evaluation values of the candidate lanes corresponding to the vehicle exiting the intersection based on the map data, the navigation information, the vehicle state information, and the obstacle state information includes:

[0035] Input the map data, the navigation information, the vehicle state information, and the obstacle state information into a pre-trained lane selection model to obtain the probability values of the candidate lanes corresponding to the vehicle exiting the intersection; the lane selection model is pre-trained according to expert driving data;

[0036] Use the probability values of the candidate lanes as the lane evaluation values corresponding to the candidate lanes.

[0037] Optionally, the step of determining the regulation and control cost of each candidate lane based on the current trajectory planning algorithm and control tracking algorithm includes:

[0038] Statistical the first historical lateral error of the trajectory planning algorithm and the second historical lateral error of the control tracking algorithm;

[0039] According to the first historical lateral error and the second historical lateral error, calculate the regulation and control cost of each candidate lane, and the regulation and control cost is positively correlated with both the first historical lateral error and the second historical lateral error.

[0040] Optionally, the step of determining the regulation and control cost of each candidate lane based on the current trajectory planning algorithm and control tracking algorithm includes:

[0041] For each candidate lane, generate multiple candidate trajectories from the current lane where the vehicle is located to the candidate lane based on the trajectory planning algorithm;

[0042] For each of the candidate trajectories, calculate the planning cost of the candidate trajectory, and use the control tracking algorithm to control and track the candidate trajectory to obtain the corresponding control error, and calculate the tracking cost of the candidate trajectory according to the control error; the tracking cost is positively correlated with the control error;

[0043] For each of the candidate trajectories, calculate the trajectory cost of the candidate trajectory according to the planning cost and the tracking cost of the candidate trajectory; the trajectory cost is positively correlated with the planning cost and positively correlated with the tracking cost;

[0044] Determine the candidate trajectory with the lowest trajectory cost, and use the trajectory cost of the candidate trajectory as the planning and control cost of the candidate lane.

[0045] Optionally, the step of determining the optimal lane among the candidate lanes according to the traffic evaluation values and the planning and control costs of the candidate lanes includes:

[0046] Among the candidate lanes, determine the candidate lanes whose planning and control costs meet the preset requirements as the alternative lanes;

[0047] Among the alternative lanes, determine the alternative lane with the highest traffic evaluation value as the optimal lane.

[0048] In a second aspect, an embodiment of the present application provides an intersection lane selection device, and the device includes:

[0049] A data acquisition module, configured to acquire map data corresponding to the current intersection, as well as navigation information, self-vehicle state information, and obstacle state information;

[0050] A traffic evaluation value determination module, configured to determine the lane evaluation values of the candidate lanes corresponding to the self-vehicle exiting the intersection based on the map data, the navigation information, the self-vehicle state information, and the obstacle state information, and for each candidate lane, determine the traffic evaluation value of the candidate lane based on the lane evaluation value of the candidate lane;

[0051] A planning and control cost determination module, configured to determine the planning and control costs of the candidate lanes based on the current trajectory planning algorithm and control tracking algorithm;

[0052] An optimal lane determination module, configured to determine the optimal lane among the candidate lanes according to the traffic evaluation values and the planning and control costs of the candidate lanes.

[0053] Optionally, the device further includes:

[0054] A safety evaluation value calculation module, configured to calculate the safety evaluation value of each candidate lane; the safety evaluation value includes: a trajectory safety evaluation result, and / or, an intersection collision risk;

[0055] The traffic evaluation value determination module is specifically configured to:

[0056] For each of the candidate lanes, based on the lane evaluation value and the safety evaluation value of the candidate lane, determine the traffic evaluation value of the candidate lane.

[0057] Optionally, when the safety evaluation value includes a trajectory safety evaluation result and an intersection collision risk, the safety evaluation value calculation module is specifically configured to:

[0058] For each of the candidate lanes, generate a planned trajectory from the current lane where the vehicle is located to the candidate lane, and calculate the trajectory safety evaluation result corresponding to the planned trajectory;

[0059] Obtain the intersection collision risk values corresponding to each intersection stored in advance, and determine the intersection collision risk corresponding to the current intersection according to the intersection collision risk values corresponding to each intersection.

[0060] Optionally, the traffic evaluation value determination module includes:

[0061] A lane determination sub-module, configured to determine the current lane where the vehicle is located, and according to the map data and the navigation information, determine each candidate lane corresponding to the vehicle exiting the intersection, and the target lane corresponding to the vehicle before reaching the next intersection;

[0062] A lane evaluation value calculation sub-module, configured to, for each of the candidate lanes, calculate a first lane evaluation value from the current lane to the candidate lane according to the vehicle state information and the obstacle state information; and calculate a second lane evaluation value from the candidate lane to the target lane according to the vehicle state information and the obstacle state information; the first lane evaluation value includes a first traffic efficiency evaluation value, and / or, a first traffic safety evaluation value; the second lane evaluation value includes a second traffic efficiency evaluation value, and / or, a second traffic safety evaluation value;

[0063] A lane evaluation value combination sub-module, configured to, for each of the candidate lanes, calculate the lane evaluation value of the candidate lane according to the first lane evaluation value and the second lane evaluation value.

[0064] Optionally, the obstacle state information includes static obstacle state information and dynamic obstacle state information; when the first lane evaluation value includes a first traffic efficiency evaluation value and a first traffic safety evaluation value, the lane evaluation value calculation sub-module is specifically configured to:

[0065] For each of the candidate lanes, determine the length of the smooth continuous curve from the current lane to that candidate lane, and based on the length of the smooth continuous curve, determine the first traffic efficiency evaluation value from the current lane to that candidate lane; the first traffic efficiency evaluation value is negatively correlated with the length of the smooth continuous curve;

[0066] For each of the candidate lanes, calculate the static structure risk coefficient corresponding to that candidate lane according to the static obstacle state information, and calculate the dynamic traffic flow risk coefficient corresponding to that candidate lane according to the host vehicle state information and the dynamic obstacle state information; and calculate the first traffic safety evaluation value from the current lane to that candidate lane based on the static structure risk coefficient and the dynamic traffic flow risk coefficient; wherein, the first traffic safety evaluation value is negatively correlated with the static structure risk coefficient and the dynamic traffic flow risk coefficient; the static structure risk coefficient is related to the relative position of the static obstacle and the lane and the probability that the static obstacle is observed; the dynamic traffic flow risk coefficient is related to the collision time between the host vehicle and the dynamic obstacle;

[0067] Calculate the first lane evaluation value from the current lane to that candidate lane according to the first traffic efficiency evaluation value and the first traffic safety evaluation value; the first lane evaluation value is positively correlated with both the first traffic efficiency evaluation value and the first traffic safety evaluation value.

[0068] Optionally, when the second lane evaluation value includes a second traffic efficiency evaluation value and a second traffic safety evaluation value, the lane evaluation value calculation sub-module is specifically configured to:

[0069] Determine the minimum number of lane changes from that candidate lane to the target lane, and according to the host vehicle state information and the obstacle state information, calculate the expected travel time and the expected lane change time corresponding to each lane passed by the host vehicle when passing from that candidate lane to the target lane through the minimum number of lane changes; wherein, the expected travel time corresponding to any lane is related to the expected travel distance of the host vehicle in that lane and the obstacle distribution in that lane; the expected lane change time corresponding to any lane is positively correlated with the length of the lane changeable interval of that lane;

[0070] Calculate the second traffic efficiency evaluation value from that candidate lane to the target lane according to the minimum number of lane changes, the expected travel time and the expected lane change time corresponding to each lane; the second traffic efficiency evaluation value is negatively correlated with the minimum number of lane changes, the expected travel time and the expected lane change time;

[0071] Calculate the second traffic safety evaluation value from that candidate lane to the target lane according to the minimum number of lane changes; the second traffic safety evaluation value is negatively correlated with the minimum number of lane changes;

[0072] Calculate a second lane evaluation value from the candidate lane to the target lane according to the second traffic efficiency evaluation value and the second traffic safety evaluation value; the second lane evaluation value is positively correlated with both the second traffic efficiency evaluation value and the second traffic safety evaluation value.

[0073] Optionally, the lane evaluation value combination sub-module is specifically configured to:

[0074] For each candidate lane, determine a first weight of the first lane evaluation value and a second weight of the second lane evaluation value;

[0075] Weight the first lane evaluation value and the second lane evaluation value based on the first weight and the second weight, and use the weighted result as the lane evaluation value of the candidate lane.

[0076] Optionally, the traffic evaluation value determination module is specifically configured to:

[0077] Input the map data, the navigation information, the self-vehicle state information, and the obstacle state information into a pre-trained lane selection model to obtain probability values of each candidate lane corresponding to the self-vehicle exiting the intersection; the lane selection model is pre-trained according to expert driving data;

[0078] Use the probability values of each candidate lane as the lane evaluation values of the corresponding candidate lanes.

[0079] Optionally, the regulation and control cost determination module is specifically configured to:

[0080] Statistically calculate a first historical lateral error of the trajectory planning algorithm and a second historical lateral error of the control tracking algorithm;

[0081] Calculate the regulation and control cost of each candidate lane according to the first historical lateral error and the second historical lateral error, and the regulation and control cost is positively correlated with both the first historical lateral error and the second historical lateral error.

[0082] Optionally, the regulation and control cost determination module is specifically configured to:

[0083] For each candidate lane, generate multiple candidate trajectories from the current lane where the self-vehicle is located to the candidate lane based on the trajectory planning algorithm;

[0084] For each candidate trajectory, calculate the planning cost of the candidate trajectory, and use the control tracking algorithm to control and track the candidate trajectory to obtain a corresponding control error, and calculate the tracking cost of the candidate trajectory according to the control error; the tracking cost is positively correlated with the control error;

[0085] For each of the candidate trajectories, calculate the trajectory cost of the candidate trajectory according to the planning cost and the tracking cost of the candidate trajectory; the trajectory cost is positively correlated with the planning cost and positively correlated with the tracking cost;

[0086] Determine the candidate trajectory with the lowest trajectory cost, and use the trajectory cost of the candidate trajectory as the planning and control cost of the candidate lane.

[0087] Optionally, the optimal lane determination module is specifically configured to:

[0088] Among the candidate lanes, determine the candidate lanes whose planning and control costs meet the preset requirements as the alternative lanes;

[0089] Among the alternative lanes, determine the alternative lane with the highest traffic evaluation value as the optimal lane.

[0090] In a third aspect, an embodiment of the present application provides a computer device, including: a memory and a processor, the memory and the processor are coupled;

[0091] The memory is used to store one or more computer instructions;

[0092] The processor is used to execute the one or more computer instructions to implement the intersection lane selection method as described in the first aspect.

[0093] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which one or more computer instructions are stored, and the instructions are executed by a processor to implement the intersection lane selection method as described in the first aspect above.

[0094] In a fifth aspect, the present application provides a computer program product, the computer program product includes a computer program, and when the computer program is executed by a processor, it implements the intersection lane selection method as described in the first aspect. Description of the Drawings

[0095] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0096] Figure 1 It is a schematic diagram of the intersection lane selection and trajectory planning and tracking results;

[0097] Figure 2 It is a schematic flowchart of an intersection lane selection method in an embodiment of the present application;

[0098] Figure 3 Schematic diagram of the process for determining the lane evaluation value of each candidate lane in the embodiment of the present application;

[0099] Figure 4 Schematic diagram of a target lane determination result in the embodiment of the present application;

[0100] Figure 5 Schematic diagram of another target lane determination result in the embodiment of the present application;

[0101] Figure 6 Schematic diagram of the minimum lane change times result of each lane in the embodiment of the present application;

[0102] Figure 7 Schematic diagram of a structure of the intersection lane selection device provided in the embodiment of the present application;

[0103] Figure 8 Schematic diagram of a structure of the computer device provided in the embodiment of the present application. Detailed implementation manners

[0104] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0105] It should be noted that the terms "including" and "having" in the embodiments of the present application and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or devices.

[0106] In the known intersection lane selection method, it is usually based on a network to learn the optimal lane selection result from expert driving data. However, after selecting the exit lane, trajectory planning and control tracking need to be performed based on the selected lane. Trajectory planning and control tracking are completed by separate trajectory generation modules and control tracking modules. The trajectory generated by the trajectory generation module cannot guarantee to be close to the expert trajectory, and at the same time, the tracking of the trajectory by the control tracking module cannot maintain a high control accuracy like an expert driver. This may lead to the following problems: for a given optimal lane selection decision, there may be a situation where the vehicle collides with obstacles such as the road edge, fence, and cone barrel during the intersection due to unreasonable trajectory planning results or excessive control tracking errors.

[0107] As shown Figure 1 in the figure, according to the navigation information, the autonomous vehicle 110 needs to go straight through the current intersection. The lane where the autonomous vehicle is currently located, that is, the incoming intersection lane, is Lane A. According to the map data or the real-time perception data of the sensor, the available outgoing intersection lanes include Lane A′, Lane B′, and Lane C′. The expert lane learned based on the network is Lane A′, and the corresponding expert trajectory is Trajectory 120. However, when the outgoing intersection lane is determined to be Lane A′, the planned trajectory obtained by the current trajectory planning algorithm of the autonomous vehicle is Trajectory 130, and after control tracking, the obtained tracking trajectory is Trajectory 140. It can be seen that the distance between the final trajectory 140 and the static obstacle on the road edge is very small. When the autonomous vehicle travels according to Trajectory 140, a collision between the autonomous vehicle and the obstacle will occur.

[0108] It can be seen that relying solely on the expert trajectory for outgoing intersection lane selection cannot ensure the safety of the autonomous vehicle during driving. It is necessary to consider the computing power of the trajectory planning algorithm and the control tracking algorithm at the same time. Based on this, the embodiments of the present application disclose a method, device, equipment, and readable storage medium for intersection lane selection to improve the accuracy of intersection lane selection of autonomous vehicles, thereby ensuring the driving safety of autonomous vehicles. The embodiments of the present application will be described in detail below.

[0109] Figure 2 FIG. is a schematic flowchart of a method for intersection lane selection provided by an embodiment of the present application. This method can be applied to an autonomous vehicle. Specifically, it can be applied to a processor in an autonomous vehicle, such as Figure 2 shown in the figure. The method may include the following steps:

[0110] S210: Obtain the map data corresponding to the current intersection, as well as navigation information, self-vehicle state information, and obstacle state information.

[0111] During the driving process of the autonomous vehicle, path planning and the like can be performed based on the pre-stored or real-time loaded map data. Among them, the map data may include a navigation map, a high-precision map, etc., which are all acceptable. The above navigation information may be the driving route determined by the autonomous vehicle according to the starting point and the ending point set by the user; the self-vehicle state information and the obstacle state information can be obtained through perception. Specifically, the self-vehicle state information may include, for example: position, speed, heading angle, acceleration, etc.; the obstacles may include all dynamic obstacles (vehicles, pedestrians, etc.) and static obstacles (green plants, road edges, cones, etc.) in the surrounding environment of the self-vehicle that may affect the driving of the self-vehicle. Each obstacle state information may include its position, speed, etc.

[0112] S220: Based on the map data, navigation information, the state information of the host vehicle, and the state information of the obstacles, determine the lane evaluation values of the candidate lanes corresponding to the host vehicle exiting the intersection, and for each candidate lane, determine the passing evaluation value of the candidate lane based on the lane evaluation value of the candidate lane.

[0113] When an autonomous vehicle travels to any intersection, according to the navigation information, it may need to travel in different directions after passing through the intersection, such as going straight, turning left, turning right, etc. And it can be understood that when the autonomous vehicle travels in any direction, there are usually multiple lanes that can be traveled in that direction (which can be called candidate lanes), and the conditions of each candidate lane are usually not exactly the same, such as the current traffic flow of each candidate lane is different, the distance from the current lane of the host vehicle is different, etc., which will result in different subsequent driving conditions (travel time, safety, etc.) after the autonomous vehicle selects different candidate lanes. In the embodiments of the present application, in order to improve the driving experience and safety of the user driving the autonomous vehicle, the optimal candidate lane corresponding to the intersection can be selected when the autonomous vehicle travels to the intersection.

[0114] Specifically, the autonomous vehicle can first determine the driving direction corresponding to its exit from the intersection according to the map data and the navigation information, and then regard each lane of the road in that direction as a candidate lane. Further, the lane evaluation values of the candidate lanes can be determined based on the map data, the state information of the host vehicle, and the state information of the obstacles. That is to say, the candidate lanes can be evaluated according to the road conditions of the candidate lanes, and then the optimal candidate lane can be selected according to the evaluation results of the candidate lanes as the lane that the vehicle is about to enter. For each candidate lane, its lane evaluation value can reflect the degree to which the candidate lane is currently suitable for the host vehicle to pass. The higher the lane evaluation value, the more suitable the lane is for the host vehicle to pass.

[0115] In one implementation, as Figure 3 shown, the process of determining the lane evaluation values of the candidate lanes may include the following steps:

[0116] S310: Determine the current lane where the host vehicle is located, and according to the map data and the navigation information, determine the candidate lanes corresponding to the host vehicle exiting the intersection, and the target lane corresponding to the host vehicle before reaching the next intersection.

[0117] It can be understood that when determining the lane evaluation values of the candidate lanes, in addition to considering the road conditions of the candidate lanes themselves (traffic flow, obstacle distribution, etc.), the positional relationship between the candidate lanes and the current lane where the host vehicle is located, and the target lane corresponding to the host vehicle before reaching the next intersection is also particularly important. For example, the closer the candidate lane is to the current lane, the shorter the passing time will be, and the closer the candidate lane is to the target lane, the fewer the number of lane changes required, and the higher the driving safety of the host vehicle, etc.

[0118] Specifically, the autonomous vehicle can determine the driving direction corresponding to its exit based on the map data and navigation information, and then regard each lane of the road in that direction as a candidate lane. Among them, the above driving directions can include going straight, turning left, turning right, etc. When determining the target lane corresponding to the vehicle before reaching the next intersection, specifically, it can be determined according to the driving direction of the vehicle when it reaches the next intersection. For example, when the driving direction of the vehicle when it reaches the next intersection is turning left, the target lane can be determined as the left-turn lane in the road or the lane closer to the left-turn lane; when the driving direction of the vehicle when it reaches the next intersection is turning right, the target lane can be determined as the right-turn lane in the road or the lane closer to the right-turn lane; when the driving direction of the vehicle when it reaches the next intersection is going straight, the target lane can be determined as the straight lane in the road. When there are multiple straight lanes, the straight lane closest to the vehicle's current lane can be selected, or a lane selection rule can be preset. For example, it is possible to determine the leftmost lane or the rightmost lane as the target lane.

[0119] As Figure 4 shown in the intersection, the current lane of the vehicle 410 is lane A, the driving direction at the current intersection is going straight, and the candidate lanes available after passing through the intersection are lane A′, lane B′, and lane C′. When the driving direction of the vehicle 410 when it reaches the next intersection is turning left, the target lane can be determined as lane A′; when the driving direction of the vehicle 410 when it reaches the next intersection is turning right, the target lane can be determined as lane C′; when the driving direction of the vehicle 410 when it reaches the next intersection is going straight, the target lane can be determined as lane B′.

[0120] As Figure 5 shown in the intersection, the current lane of the vehicle 510 is lane C, the driving direction at the current intersection is turning right, and the candidate lanes available after passing through the intersection are lane A′, lane B′, and lane C′. The navigation information shows that the driving direction when reaching the next intersection is turning left. Therefore, the target lane corresponding to before reaching the next intersection can be determined as lane A′.

[0121] S320: For each candidate lane, calculate the first lane evaluation value from the current lane to the candidate lane according to the vehicle state information and the obstacle state information; and calculate the second lane evaluation value from the candidate lane to the target lane according to the vehicle state information and the obstacle state information.

[0122] In the embodiments of the present application, when selecting a lane, the efficiency and safety of passing can be mainly considered. Therefore, the above first lane evaluation value may include a first passing efficiency evaluation value and / or a first passing safety evaluation value; the second lane evaluation value may also include a second passing efficiency evaluation value and / or a second passing safety evaluation value. It can be understood that the passing efficiency evaluation value can identify the driving efficiency of the vehicle itself. The higher the passing efficiency evaluation value of any candidate lane, the higher the passing efficiency of the vehicle itself, and the more suitable the candidate lane is for the vehicle to pass; the passing safety evaluation value identifies the driving safety of the vehicle itself. The higher the passing safety evaluation value of any candidate lane, the higher the passing safety of the vehicle itself, and the more suitable the candidate lane is for the vehicle to pass.

[0123] Specifically, the above obstacle state information may include static obstacle state information and dynamic obstacle state information. For each candidate lane, when calculating the first lane evaluation value from the current lane to the candidate lane, for each candidate lane, the smooth continuous curve length from the current lane to the candidate lane can be determined, and based on the smooth continuous curve length, the first passing efficiency evaluation value from the current lane to the candidate lane can be determined; the first passing efficiency evaluation value is negatively correlated with the smooth continuous curve length; for each candidate lane, according to the static obstacle state information, the static structure risk coefficient corresponding to the candidate lane can be calculated, and according to the vehicle state information and the dynamic obstacle state information, the dynamic traffic flow risk coefficient corresponding to the candidate lane can be calculated; and based on the static structure risk coefficient and the dynamic traffic flow risk coefficient, the first passing safety evaluation value from the current lane to the candidate lane can be calculated; wherein, the first passing safety evaluation value is negatively correlated with the static structure risk coefficient and the dynamic traffic flow risk coefficient; the static structure risk coefficient is related to the relative position of the static obstacle and the lane and the probability of the static obstacle being observed; the dynamic traffic flow risk coefficient is related to the collision time between the vehicle itself and the dynamic obstacle; then, according to the first passing efficiency evaluation value and the first passing safety evaluation value, the first lane evaluation value from the current lane to the candidate lane can be calculated; the first lane evaluation value is positively correlated with both the first passing efficiency evaluation value and the first passing safety evaluation value.

[0124] For each candidate lane, when determining the smooth continuous curve length from the current lane to the candidate lane, first, based on the map data, the attitude points of the current lane and the candidate lane can be determined. The attitude point can be a vector, including at least 4 parameters, namely: x coordinate, y coordinate, heading angle, and curvature value; then, based on any curve fitting method, the connection line between the two attitude points can be determined, and the length of the obtained connection line is the smooth continuous curve length from the current lane to the candidate lane. Among them, the above curve fitting method may include, for example, a third-order Bezier curve or a cubic spiral curve, etc.

[0125] The length of the smooth continuous curve from the current lane to each candidate lane can identify the distance from the current lane to each candidate lane. Further, this distance can identify the passing time of the host vehicle from the current lane to each candidate lane. Therefore, based on the calculated smooth continuous curve length, the first passing efficiency evaluation value from the current lane to each candidate lane can be determined, where the first passing efficiency evaluation value is negatively correlated with the smooth continuous curve length. That is to say, the greater the length of the smooth continuous curve from the current lane to any candidate lane, the longer the passing time required for the host vehicle to travel from the current lane to that candidate lane, and the lower the passing efficiency.

[0126] It can be understood that the relative position between the static obstacle and the lane, as well as the probability of the static obstacle being observed, are related to the lane safety. For example, the greater the relative position between the static obstacle and the lane, that is, the farther the distance between the static obstacle and the lane, the greater the lane safety. Therefore, the above-mentioned static structure risk coefficient is related to the relative position between the static obstacle and the lane and the probability of the static obstacle being observed. The time to collision (TTC) between the host vehicle and the dynamic obstacle can reflect the relative position between the host vehicle and the dynamic obstacle. For example, the longer the collision time, the higher the passing safety of the host vehicle. Therefore, the above-mentioned dynamic traffic flow risk coefficient is related to the collision time between the host vehicle and the dynamic obstacle. Moreover, the greater the static structure risk coefficient and the dynamic traffic flow risk coefficient, the lower the passing safety of the host vehicle. Therefore, the first passing safety evaluation value is negatively correlated with both the static structure risk coefficient and the dynamic traffic flow risk coefficient.

[0127] When calculating the first lane evaluation value, for example, the sum of the first passing efficiency evaluation value and the first passing safety evaluation value corresponding to any candidate lane can be used as the first lane evaluation value from the current lane to that candidate lane; or, weights for the first passing efficiency evaluation value and the first passing safety evaluation value can be set, and then the weighted result of the first passing efficiency evaluation value and the first passing safety evaluation value can be used as the first lane evaluation value. This is all acceptable.

[0128] For each candidate lane, when calculating the second-lane evaluation value from the candidate lane to the target lane, the minimum number of lane changes from the candidate lane to the target lane can be determined, and based on the vehicle state information and obstacle state information, when the vehicle travels from the candidate lane to the target lane through the minimum number of lane changes, calculate the expected travel time and expected lane change time corresponding to each lane passed; among them, the expected travel time corresponding to any lane is related to the expected driving distance of the vehicle in that lane and the obstacle distribution in that lane; the expected lane change time corresponding to any lane is positively correlated with the length of the lane changeable interval; then, based on the minimum number of lane changes, the expected travel time and expected lane change time corresponding to each lane, calculate the second traffic efficiency evaluation value from the candidate lane to the target lane; the second traffic efficiency evaluation value is negatively correlated with the minimum number of lane changes, expected travel time, and expected lane change time; and calculate the second traffic safety evaluation value from the candidate lane to the target lane according to the minimum number of lane changes; the second traffic safety evaluation value is negatively correlated with the minimum number of lane changes; finally, based on the second traffic efficiency evaluation value and the second traffic safety evaluation value, calculate the second-lane evaluation value from the candidate lane to the target lane; the second-lane evaluation value is positively correlated with the second traffic efficiency evaluation value and the second traffic safety evaluation value.

[0129] The minimum number of lane changes from each candidate lane to the target lane, that is, the minimum number of lane changes required for the vehicle to travel from the candidate lane to the target lane, and there will be no situation of lane return during the lane change. As Figure 6 shown, it shows the minimum number of lane changes for each candidate lane. When the target lane is lane a, the minimum number of lane changes for each candidate lane is as follows: the minimum number of lane changes for lane a is 0, the minimum number of lane changes for lane b is 1, the minimum number of lane changes for lane c is 2, the minimum number of lane changes for lane d is 3, and the minimum number of lane changes for lane d is 4.

[0130] It can be understood that many traffic accidents occur during the vehicle lane change process, such as vehicle rear-end collisions caused during the vehicle lane change process. Therefore, the number of lane changes directly affects the safety of vehicle driving. In the embodiments of the present application, the safety of each candidate lane can be measured based on the number of vehicle lane changes. Specifically, the second traffic safety evaluation value from each candidate lane to the target lane can be calculated according to the minimum number of lane changes; among them, the second traffic safety evaluation value is negatively correlated with the minimum number of lane changes. That is to say, the smaller the minimum number of lane changes, the higher the vehicle traffic safety, and the larger the second traffic safety evaluation value.

[0131] The above obstacle distribution may include the number of obstacles, the arrangement form of obstacles, and the obstacle flow rate, etc. The lane-changing interval of any lane may include the dotted line length between this lane and the next lane. The fewer the minimum number of lane changes, the shorter the expected passing time, and the shorter the expected lane-changing time, the higher the passing efficiency of the host vehicle. Therefore, the second passing efficiency evaluation value is negatively correlated with the minimum number of lane changes, the expected passing time, and the expected lane-changing time.

[0132] When calculating the second lane evaluation value, for example, the sum of the second passing efficiency evaluation value and the second passing safety evaluation value corresponding to any candidate lane can be used as the second lane evaluation value from this candidate lane to the target lane; or, the weights of the second passing efficiency evaluation value and the second passing safety evaluation value can be set, and then the weighted result of the second passing efficiency evaluation value and the second passing safety evaluation value can be used as the second lane evaluation value, which is all acceptable.

[0133] S330: For each candidate lane, calculate the lane evaluation value of this candidate lane according to the first lane evaluation value and the second lane evaluation value.

[0134] Specifically, for each candidate lane, the first weight of the first lane evaluation value and the second weight of the second lane evaluation value can be determined; then, based on the first weight and the second weight, the first lane evaluation value and the second lane evaluation value are weighted, and the weighted result is used as the lane evaluation value of this candidate lane. Among them, the first weight and the second weight can be set to the same value or different values, which are all acceptable, and the embodiments of the present application do not limit their specific values.

[0135] In a specific embodiment, the lane evaluation value F(N) of any candidate lane can be calculated according to the following formula: F(N) = F 1 (N) + F 2 (N).

[0136] F 1 (N) is used to measure the passing efficiency, and F 2 (N) is used to measure the passing safety.

[0137] Among them, F 1 (N) = G 1 (N) + H 1 (N) = g 1 (d) + h 1 (c) + h2(s, p 1 ) + h 3 (p 2 )

[0138] G 1 (N) represents the efficiency cost from the current lane S to the candidate lane N, measured by the expected passing time, G 1(N) is negatively correlated with d, denoted by g 1 (d), where d is the length of the smooth continuous curve from the current lane S to the candidate lane N.

[0139] H 1 (N) represents the efficiency cost from the candidate lane N to the target lane E, measured by the expected travel time. Among them, H 1 (N) is negatively correlated with c, where c is the minimum number of lane changes from the candidate lane N to the target lane E, denoted by h 1 (c); at the same time, H 1 (N) is related to the expected travel time on each lane n passed through the shortest number of lane changes c from the candidate lane N to the target lane E. The expected travel time t(n) is a function of the expected driving length s on lane n and the obstacle distribution p 1 on the lane n, that is: t(n) = h2(s, p 1 ); at the same time, H 1 (N) is related to the expected lane change time on lane n in the direction of the target lane after passing through the shortest number of lane changes c from the candidate lane N to the target lane E. The expected lane change time is negatively correlated with the available lane change interval distribution p 2 (the length of the dotted line between lanes), that is h 3 (p 2 ).

[0140] F 2 (N) = G 2 (N) + H 2 (N) = g 2 (r 1 , r 2 ) + h 2 (c)

[0141] G 2 (N) represents the safety cost from the current lane S to the candidate lane N, measured by the expected collision probability. G 2 (N) is related to the road static structure risk coefficient r 1 , and the dynamic traffic flow risk coefficient r 2 from S to N, denoted by g 2 (r 1 , r 2 ).

[0142] H 2 (N) represents the safety cost from the candidate lane N to the target lane E, measured by the expected collision or expected collision risk probability. H 2 (N) is negatively correlated with c, where c is the minimum number of lane changes from the candidate lane N to the target lane E, denoted by h 2 (c).

[0143] In another implementation, the lane evaluation values of each candidate lane can also be determined based on a pre-trained network model. Specifically, a certain amount of expert driving data can be collected in advance. The expert driving data includes intersection driving data, that is, when the autonomous vehicle passes through an intersection, the corresponding lane selection result when leaving the intersection. Based on the network model trained with this expert driving data, lane selection at intersections can be performed.

[0144] When performing lane selection at an intersection, the obtained map data, navigation information, self-vehicle status information, and obstacle status information can be input into the pre-trained lane selection model. The lane selection model can then output the probability values of each candidate lane corresponding to the self-vehicle leaving the intersection. The magnitudes of the probability values of each candidate lane indicate the probabilities of each candidate lane being suitable for passage. Furthermore, the probability values of each candidate lane can be used as the lane evaluation values corresponding to each candidate lane.

[0145] After determining the lane evaluation values of each candidate lane, in one implementation, the lane evaluation values of each candidate lane can be directly used as their passage evaluation values.

[0146] It can be understood that when determining the passage evaluation values of each candidate lane, in addition to the lane conditions of each candidate lane itself, there may be other factors that affect the vehicle passage conditions of the lane. Therefore, in one implementation, before determining the passage evaluation value based on the lane evaluation value of the candidate lane, the safety evaluation value of each candidate lane can also be calculated. The safety evaluation value can include: the trajectory safety evaluation result, and / or, the intersection collision risk. Then, for each candidate lane, based on the lane evaluation value and safety evaluation value of the candidate lane, the passage evaluation value of the candidate lane can be determined.

[0147] Among them, when calculating the trajectory safety evaluation result of each candidate lane, for each candidate lane, a planned trajectory from the current lane where the self-vehicle is located to the candidate lane can be generated, and the trajectory safety evaluation result corresponding to the planned trajectory can be calculated. For example, based on the current trajectory planning module of the autonomous vehicle, for each candidate lane, a planned trajectory from the current lane where the self-vehicle is located to the candidate lane can be generated, and for this planned trajectory, considering its safety, smoothness, etc. comprehensively, the corresponding trajectory safety evaluation result can be calculated.

[0148] It can be understood that different intersections have different corresponding safety levels due to their complexity, intersection structure, etc. In the embodiments of the present application, for each intersection, the corresponding intersection collision risk value can be determined and stored according to historical vehicle driving safety conditions, etc. For example, each intersection can be numbered, and the intersection numbers and the corresponding collision risk values can be stored correspondingly. Thus, when selecting an intersection lane, the intersection collision risk values corresponding to each intersection stored in advance can be obtained, and the intersection collision risk corresponding to the current intersection can be determined.

[0149] The larger the trajectory safety assessment result is, the higher the safety level of the candidate lane is, and the more suitable it is for the host vehicle to pass through; while the larger the intersection collision risk is, the lower the safety level of the candidate lane is, and the less suitable it is for the host vehicle to pass through; therefore, the safety assessment value of each candidate lane is positively correlated with the trajectory safety assessment result and negatively correlated with the intersection collision risk. In one implementation manner, for example, the difference between the trajectory safety assessment result and the intersection collision risk of any candidate lane can be used as the safety assessment value of the candidate lane.

[0150] When determining the passing assessment values of each candidate lane, for each candidate lane, the sum of the lane assessment value and the safety assessment value of the candidate lane can be used as the passing assessment value of the candidate lane.

[0151] Comprehensively considering the lane assessment value and the safety assessment value of each candidate lane to determine the passing assessment value of each candidate lane can more comprehensively evaluate each candidate lane, and thus can improve the accuracy of lane selection when performing lane selection.

[0152] S230: Based on the current trajectory planning algorithm and control tracking algorithm, determine the regulatory and control costs of each candidate lane.

[0153] As described above, only considering the lane of the candidate lane at the intersection may result in an excessive subsequent control tracking error, leading to traffic accidents. Therefore, in the embodiments of the present application, in order to ensure the safety of the autonomous vehicle during driving, the regulatory and control costs of each candidate lane can be comprehensively considered after determining the passing assessment values of each candidate lane.

[0154] Specifically, the regulatory and control costs of each candidate lane can be determined based on the current trajectory planning algorithm and control tracking algorithm. That is, for each candidate lane, the trajectory error corresponding to when the host vehicle travels from the current lane to the candidate lane can be calculated, so as to screen out the candidate lanes that do not meet the driving safety of the host vehicle.

[0155] In one implementation, the first historical lateral error of the trajectory planning algorithm and the second historical lateral error of the control tracking algorithm can be statistically analyzed. Then, based on the first historical lateral error and the second historical lateral error, the regulation and control cost of each candidate lane is calculated, and the regulation and control cost is positively correlated with both the first historical lateral error and the second historical lateral error. For example, the sum of the first historical lateral error and the second historical lateral error can be determined as the regulation and control cost of each candidate lane.

[0156] After lane selection, the autonomous vehicle needs to control the vehicle's driving based on its current trajectory planning algorithm and control tracking algorithm. The errors of the trajectory planning algorithm and the control tracking algorithm will directly affect whether each candidate lane allows the vehicle to pass safely. For example, when the regulation and control cost of any candidate lane is too high, such as Figure 1 As shown, when the leftmost lane is determined as the exit lane, the actual driving trajectory 140 of the vehicle will have too small a distance from the static obstacle on the road edge, which may lead to dangerous accidents. Therefore, the first historical lateral error of the trajectory planning algorithm and the second historical lateral error of the control tracking algorithm can accurately reflect the computing power of the current trajectory planning algorithm and control tracking algorithm. Calculating the regulation and control cost of each candidate lane based on the first historical lateral error and the second historical lateral error can accurately evaluate each candidate lane.

[0157] When calculating the regulation and control cost of each candidate lane based on the first historical lateral error and the second historical lateral error, in one implementation, the error distribution of trajectory planning and control tracking can be statistically analyzed based on historical data to determine the mean and variance of the error, and then the regulation and control cost can be determined based on the determined mean and variance of the error.

[0158] To calculate the regulation and control cost of each candidate lane more accurately, in another implementation, for each candidate lane, multiple candidate trajectories from the current lane where the vehicle is located to the candidate lane can be generated based on the trajectory planning algorithm. For each candidate trajectory, the planning cost of the candidate trajectory is calculated, and the control tracking algorithm is used to control and track the candidate trajectory to obtain the corresponding control error. The tracking cost is calculated based on the control error; the tracking cost is positively correlated with the control error; for each candidate trajectory, the trajectory cost of the candidate trajectory is calculated based on the planning cost and the tracking cost of the candidate trajectory; the trajectory cost is negatively correlated with the planning cost and positively correlated with the tracking cost; the candidate trajectory with the lowest trajectory cost is determined, and the trajectory cost of the candidate trajectory is used as the regulation and control cost of the candidate lane.

[0159] When calculating the planning cost of each candidate trajectory, the implementation environment can be the vehicle end or the cloud. The specific calculation method can be based on a method that considers factors such as safety, body feeling, and efficiency, or the planning cost of the candidate trajectory can be given by a trajectory evaluation network.

[0160] Control and track the candidate trajectories to obtain the control errors of the candidate trajectories. The specific method can be: constructing a vehicle dynamics model and performing control pre-tracking based on the vehicle dynamics model; or predicting the control error based on a neural network, where the input of the neural network is the candidate trajectory and the output is the corresponding control error. The implementation environment of the above method can be the vehicle side or the cloud side. The cloud side can utilize sufficient computing resources to perform fast and accurate simulation calculations or model inferences, and immediately feedback the calculation results to the vehicle side through the 5G network.

[0161] That is to say, for each candidate lane, multiple candidate trajectories can be generated, and the planning cost of each candidate trajectory can be calculated. The planning cost can identify the performance of the corresponding candidate trajectory (such as smoothness, safety, etc.). The greater the planning cost, the worse the performance of the candidate trajectory; the smaller the planning cost, the better the performance of the candidate trajectory. Moreover, the candidate trajectories can be controlled and tracked, and the corresponding control errors can be calculated. The magnitude of the control error can identify the trackability of the corresponding candidate trajectory. The greater the control error, the worse the trackability of the candidate trajectory; the smaller the control error, the stronger the trackability of the candidate trajectory. Finally, the trajectory cost of the candidate trajectory can be calculated based on the planning cost and tracking cost of the candidate trajectory. The trajectory cost is positively correlated with the planning cost and positively correlated with the tracking cost. For example, the sum of the tracking cost and the trajectory cost can be determined as the regulation and control cost of each candidate lane. And the candidate trajectory with the lowest trajectory cost is determined, that is, the trajectory with good performance and strong trackability, and its trajectory cost is used as the regulation and control cost of the candidate lane.

[0162] When calculating the planning cost of each candidate trajectory, the safety, smoothness, etc. of the candidate trajectory can be comprehensively considered, and the trajectory safety evaluation result can be calculated as the planning cost of the candidate trajectory.

[0163] S240: Determine the optimal lane among the candidate lanes according to the traffic evaluation values and regulation and control costs of the candidate lanes.

[0164] The traffic evaluation values of the candidate lanes identify the degree to which the candidate lanes are suitable for vehicle passage. The greater the traffic evaluation value, the more suitable the candidate lane is for the host vehicle to pass; while its regulation and control cost identifies the total error size of the corresponding trajectory planning and control tracking when the host vehicle passes through the exit of the candidate lane. The greater the regulation and control cost, the greater the total error.

[0165] In the embodiments of the present application, after calculating the traffic evaluation value and the planning and control cost of each candidate lane, the optimal lane can be determined among the candidate lanes according to the traffic evaluation value and the planning and control cost of each candidate lane. For example, among the candidate lanes, the candidate lanes whose planning and control costs meet the preset requirements can be first determined as alternative lanes; then, among the alternative lanes, the alternative lane with the highest traffic evaluation value can be determined as the optimal lane.

[0166] Among them, the above preset requirements may include, for example, that when the vehicle passes through the exit of the candidate lane, under the influence of the planning and control cost of the candidate lane, the vehicle will not collide with static obstacles. After screening the candidate lanes through the planning and control cost, the determined alternative lanes are all lanes that can ensure the safe passage of the vehicle. Further, the alternative lane with the highest traffic evaluation value, that is, the lane with the highest vehicle safety and the highest traffic efficiency, is used as the optimal lane, so as to take into account the safety and traffic efficiency of the vehicle's driving and improve the accuracy of lane determination.

[0167] In the embodiments of the present application, when selecting an intersection lane, not only the lane evaluation value of each candidate lane, that is, the current road conditions of each candidate lane, can be considered, but also the planning and control cost of each candidate lane can be considered based on the current trajectory planning algorithm and control tracking algorithm. That is to say, if each candidate lane is used as the exit lane, the corresponding planning and control results can be considered. Therefore, by selecting the optimal lane based on the lane evaluation value and the planning and control cost of each candidate lane, it can be ensured that the road conditions of the finally determined exit lane are better, and at the same time, it can meet the computing power of the current planning and control algorithm, avoiding the situation that the autonomous vehicle collides with obstacles such as road edges, fences, and cones during passing through the intersection, thereby improving the accuracy of the intersection lane selection of the autonomous vehicle and further ensuring the driving safety of the autonomous vehicle.

[0168] Figure 7 The structure diagram of an intersection lane selection device provided by the embodiments of the present application is shown. The device includes:

[0169] A data acquisition module 710, configured to acquire map data corresponding to the current intersection, as well as navigation information, vehicle state information, and obstacle state information;

[0170] A traffic evaluation value determination module 720, configured to determine the lane evaluation value of each candidate lane corresponding to the vehicle exiting the intersection based on the map data, the navigation information, the vehicle state information, and the obstacle state information, and for each candidate lane, determine the traffic evaluation value of the candidate lane based on the lane evaluation value of the candidate lane;

[0171] The trajectory planning and control cost determination module 730 is configured to determine the trajectory planning and control costs of the candidate lanes based on the current trajectory planning algorithm and control tracking algorithm;

[0172] The optimal lane determination module 740 is configured to determine the optimal lane from the candidate lanes according to the traffic evaluation values and trajectory planning and control costs of the candidate lanes.

[0173] Optionally, the apparatus further includes:

[0174] The safety evaluation value calculation module is configured to calculate the safety evaluation value of each candidate lane; the safety evaluation value includes: the trajectory safety evaluation result, and / or, the intersection collision risk;

[0175] The traffic evaluation value determination module 720 is specifically configured to:

[0176] For each candidate lane, determine the traffic evaluation value of the candidate lane based on the lane evaluation value and safety evaluation value of the candidate lane.

[0177] Optionally, when the safety evaluation value includes the trajectory safety evaluation result and the intersection collision risk, the safety evaluation value calculation module is specifically configured to:

[0178] For each candidate lane, generate a planned trajectory from the current lane where the vehicle is located to the candidate lane, and calculate the trajectory safety evaluation result corresponding to the planned trajectory;

[0179] Obtain the intersection collision risk values corresponding to each intersection stored in advance, and determine the intersection collision risk corresponding to the current intersection according to the intersection collision risk values corresponding to each intersection.

[0180] Optionally, the traffic evaluation value determination module 720 includes:

[0181] The lane determination sub-module is configured to determine the current lane where the vehicle is located, and determine the candidate lanes corresponding to the vehicle exiting the intersection and the target lane corresponding to the vehicle before reaching the next intersection according to the map data and the navigation information;

[0182] The lane evaluation value calculation sub-module is configured to, for each candidate lane, calculate the first lane evaluation value from the current lane to the candidate lane according to the vehicle state information and the obstacle state information; and calculate the second lane evaluation value from the candidate lane to the target lane according to the vehicle state information and the obstacle state information; the first lane evaluation value includes the first traffic efficiency evaluation value, and / or, the first traffic safety evaluation value; the second lane evaluation value includes the second traffic efficiency evaluation value, and / or, the second traffic safety evaluation value;

[0183] The lane evaluation value combination sub-module is used to calculate the lane evaluation value of each candidate lane according to the first lane evaluation value and the second lane evaluation value.

[0184] Optionally, the obstacle state information includes static obstacle state information and dynamic obstacle state information; when the first lane evaluation value includes a first traffic efficiency evaluation value and a first traffic safety evaluation value, the lane evaluation value calculation sub-module is specifically configured to:

[0185] For each candidate lane, determine the length of the smooth continuous curve from the current lane to the candidate lane, and based on the length of the smooth continuous curve, determine the first traffic efficiency evaluation value from the current lane to the candidate lane; the first traffic efficiency evaluation value is negatively correlated with the length of the smooth continuous curve;

[0186] For each candidate lane, calculate the static structure risk coefficient corresponding to the candidate lane according to the static obstacle state information, and calculate the dynamic traffic flow risk coefficient corresponding to the candidate lane according to the vehicle state information and the dynamic obstacle state information; and calculate the first traffic safety evaluation value from the current lane to the candidate lane based on the static structure risk coefficient and the dynamic traffic flow risk coefficient; wherein, the first traffic safety evaluation value is negatively correlated with the static structure risk coefficient and the dynamic traffic flow risk coefficient; the static structure risk coefficient is negatively correlated with both the relative position of the static obstacle and the lane and the probability that the static obstacle is observed; the dynamic traffic flow risk coefficient is negatively correlated with the collision time between the vehicle and the dynamic obstacle;

[0187] Calculate the first lane evaluation value from the current lane to the candidate lane according to the first traffic efficiency evaluation value and the first traffic safety evaluation value; the first lane evaluation value is positively correlated with both the first traffic efficiency evaluation value and the first traffic safety evaluation value.

[0188] Optionally, when the second lane evaluation value includes a second traffic efficiency evaluation value and a second traffic safety evaluation value, the lane evaluation value calculation sub-module is specifically configured to:

[0189] Determine the minimum number of lane changes from the candidate lane to the target lane, and according to the vehicle state information and the obstacle state information, calculate the expected travel time and the expected lane change time corresponding to each lane passed by the vehicle when changing lanes from the candidate lane to the target lane through the minimum number of lane changes; wherein, the expected travel time corresponding to any lane is related to the expected driving distance of the vehicle in the lane and the obstacle distribution in the lane; the expected lane change time corresponding to any lane is positively correlated with the length of the lane changeable interval;

[0190] Calculate a second traffic efficiency evaluation value from the candidate lane to the target lane according to the minimum number of lane changes, the expected travel time and the expected lane change time corresponding to each lane; the second traffic efficiency evaluation value is negatively correlated with the minimum number of lane changes, the expected travel time and the expected lane change time;

[0191] Calculate a second traffic safety evaluation value from the candidate lane to the target lane according to the minimum number of lane changes; the second traffic safety evaluation value is negatively correlated with the minimum number of lane changes;

[0192] Calculate a second lane evaluation value from the candidate lane to the target lane according to the second traffic efficiency evaluation value and the second traffic safety evaluation value; the second lane evaluation value is positively correlated with both the second traffic efficiency evaluation value and the second traffic safety evaluation value.

[0193] Optionally, the lane evaluation value combination sub-module is specifically configured to:

[0194] For each candidate lane, determine a first weight of the first lane evaluation value and a second weight of the second lane evaluation value;

[0195] Based on the first weight and the second weight, weight the first lane evaluation value and the second lane evaluation value, and use the weighted result as the lane evaluation value of the candidate lane.

[0196] Optionally, the traffic evaluation value determination module 720 is specifically configured to:

[0197] Input the map data, the navigation information, the self-vehicle state information, and the obstacle state information into a pre-trained lane selection model to obtain the probability values of each candidate lane when the self-vehicle exits the intersection; the lane selection model is pre-trained according to expert driving data;

[0198] Use the probability values of each candidate lane as the lane evaluation values of the corresponding candidate lanes.

[0199] Optionally, the regulation and control cost determination module 730 is specifically configured to:

[0200] Statistical the first historical lateral error of the trajectory planning algorithm and the second historical lateral error of the control tracking algorithm;

[0201] Calculate the regulation and control cost of each candidate lane according to the first historical lateral error and the second historical lateral error, and the regulation and control cost is positively correlated with both the first historical lateral error and the second historical lateral error.

[0202] Optionally, the regulation and control cost determination module 730 is specifically configured to:

[0203] For each of the candidate lanes, based on the trajectory planning algorithm, multiple candidate trajectories from the current lane where the vehicle is located to the candidate lane are generated;

[0204] For each of the candidate trajectories, calculate the planning cost of the candidate trajectory, and use the control tracking algorithm to control and track the candidate trajectory to obtain the corresponding control error, and calculate the tracking cost of the candidate trajectory according to the control error; the tracking cost is positively correlated with the control error;

[0205] For each of the candidate trajectories, calculate the trajectory cost of the candidate trajectory according to the planning cost and tracking cost of the candidate trajectory; the trajectory cost is negatively correlated with the planning cost and positively correlated with the tracking cost;

[0206] Determine the candidate trajectory with the lowest trajectory cost, and use the trajectory cost of the candidate trajectory as the regulation and control cost of the candidate lane.

[0207] Optionally, the optimal lane determination module 740 is specifically configured to:

[0208] Among the candidate lanes, determine the candidate lanes whose regulation and control costs meet the preset requirements as alternative lanes;

[0209] Among the alternative lanes, determine the alternative lane with the highest traffic evaluation value as the optimal lane.

[0210] In the embodiments of the present application, when selecting lanes at an intersection, not only the lane evaluation values of each candidate lane, that is, the current road conditions of each candidate lane, can be considered, but also the regulation and control costs of each candidate lane can be considered based on the current trajectory planning algorithm and control tracking algorithm. That is to say, if each candidate lane is used as the exit lane, the corresponding planning and control results can be obtained. Therefore, by selecting the optimal lane based on the lane evaluation values and regulation and control costs of each candidate lane, it can ensure that the road conditions of the finally determined exit lane are better, and at the same time meet the computing power of the current planning and control algorithm, avoiding the situation where an autonomous vehicle collides with obstacles such as road edges, fences, and cones during passing through the intersection, thereby improving the accuracy of the autonomous vehicle's intersection lane selection and further ensuring the driving safety of the autonomous vehicle.

[0211] The above device embodiments correspond to the method embodiments and have the same technical effects as the method embodiments. For specific descriptions, please refer to the method embodiments. The device embodiments are obtained based on the method embodiments. For specific descriptions, please refer to the method embodiment part and will not be elaborated here.

[0212] Next, a computer device provided by the embodiments of the present application will be introduced. Please refer to Figure 8 ,Figure 8 A schematic structural diagram of a computer device provided by an embodiment of the present application. The computer device includes:

[0213] One or more processors 40;

[0214] The processor 40 is coupled to a storage device 41, and the storage device 41 is used to store one or more programs.

[0215] When the one or more programs are executed by the one or more processors 40, the electronic device implements the technical solution of a lane selection method at an intersection as Figures 2 to 6 described.

[0216] An embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the technical solution of a lane selection method at an intersection as Figures 2 to 6 described.

[0217] The present application provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the technical solution of a lane selection method at an intersection as Figures 2 to 6 described.

[0218] Those of ordinary skill in the art can understand that the drawings are only schematic diagrams of one embodiment, and the modules or processes in the drawings are not necessarily essential for implementing the present application.

[0219] Those of ordinary skill in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment as described in the embodiment, or can be correspondingly changed and located in one or more devices different from this embodiment. The modules in the above embodiments can be combined into one module, or further split into multiple sub-modules.

[0220] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An intersection lane selection method, characterized in that, the method includes: obtaining map data corresponding to the current intersection, as well as navigation information, self-vehicle status information, and obstacle status information; based on the map data, the navigation information, the self-vehicle status information, and the obstacle status information, determining the lane evaluation value of each candidate lane for the self-vehicle to exit the intersection, and for each candidate lane, determining the passing evaluation value of the candidate lane based on the lane evaluation value of the candidate lane; based on the current trajectory planning algorithm and control tracking algorithm, determining the regulation and control cost of each candidate lane; determining the optimal lane among the candidate lanes according to the passing evaluation values and regulation and control costs of the candidate lanes.

2. The method according to claim 1, characterized in that, before the step of determining the passing evaluation value of each candidate lane based on the lane evaluation value of the candidate lane, the method further includes: calculating the safety evaluation value of each candidate lane; the safety evaluation value includes: trajectory safety evaluation result, and / or, intersection collision risk; the step of determining the passing evaluation value of each candidate lane based on the lane evaluation value of the candidate lane includes: for each candidate lane, determining the passing evaluation value of the candidate lane based on the lane evaluation value and safety evaluation value of the candidate lane.

3. The method according to claim 2, characterized in that, when the safety evaluation value includes a trajectory safety evaluation result and an intersection collision risk, the step of calculating the safety evaluation value of each candidate lane includes: for each candidate lane, generating a planned trajectory from the current lane where the self-vehicle is located to the candidate lane, and calculating the trajectory safety evaluation result corresponding to the planned trajectory; obtaining the intersection collision risk values corresponding to each intersection stored in advance, and determining the intersection collision risk corresponding to the current intersection according to the intersection collision risk values corresponding to each intersection.

4. The method according to claim 1, characterized in that, the step of determining the lane evaluation value of each candidate lane for the self-vehicle to exit the intersection based on the map data, the navigation information, the self-vehicle status information, and the obstacle status information includes: determining the current lane where the self-vehicle is located, and according to the map data and the navigation information, determining each candidate lane for the self-vehicle to exit the intersection, and the target lane corresponding to the self-vehicle before reaching the next intersection; for each candidate lane, calculating a first lane evaluation value from the current lane to the candidate lane according to the self-vehicle status information and the obstacle status information; and calculating a second lane evaluation value from the candidate lane to the target lane according to the self-vehicle status information and the obstacle status information; the first lane evaluation value includes a first passing efficiency evaluation value, and / or, a first passing safety evaluation value; the second lane evaluation value includes a second passing efficiency evaluation value, and / or, a second passing safety evaluation value; for each candidate lane, calculating the lane evaluation value of the candidate lane according to the first lane evaluation value and the second lane evaluation value.

5. The method according to claim 4, wherein, the obstacle state information includes static obstacle state information and dynamic obstacle state information; when the first lane evaluation value includes a first traffic efficiency evaluation value and a first traffic safety evaluation value, the step of calculating, for each of the candidate lanes, the first lane evaluation value from the current lane to the candidate lane according to the vehicle state information and the obstacle state information includes: for each of the candidate lanes, determining the length of the smooth continuous curve from the current lane to the candidate lane, and based on the length of the smooth continuous curve, determining the first traffic efficiency evaluation value from the current lane to the candidate lane; the first traffic efficiency evaluation value is negatively correlated with the length of the smooth continuous curve; for each of the candidate lanes, calculating the static structure risk coefficient corresponding to the candidate lane according to the static obstacle state information, and calculating the dynamic traffic flow risk coefficient corresponding to the candidate lane according to the vehicle state information and the dynamic obstacle state information; and calculating the first traffic safety evaluation value from the current lane to the candidate lane based on the static structure risk coefficient and the dynamic traffic flow risk coefficient; wherein, the first traffic safety evaluation value is negatively correlated with the static structure risk coefficient and the dynamic traffic flow risk coefficient; the static structure risk coefficient is related to the relative position between the static obstacle and the lane and the probability that the static obstacle is observed; the dynamic traffic flow risk coefficient is related to the collision time between the vehicle and the dynamic obstacle; calculating the first lane evaluation value from the current lane to the candidate lane according to the first traffic efficiency evaluation value and the first traffic safety evaluation value; the first lane evaluation value is positively correlated with both the first traffic efficiency evaluation value and the first traffic safety evaluation value.

6. The method according to claim 4, wherein, when the second lane evaluation value includes a second traffic efficiency evaluation value and a second traffic safety evaluation value, the step of calculating the second lane evaluation value from the candidate lane to the target lane according to the vehicle state information and the obstacle state information includes: determining the minimum number of lane changes from the candidate lane to the target lane, and according to the vehicle state information and the obstacle state information, calculating the expected travel time and the expected lane change time corresponding to each lane passed by the vehicle when changing lanes from the candidate lane to the target lane through the minimum number of lane changes; wherein, the expected travel time corresponding to any lane is related to the expected travel distance of the vehicle in that lane and the obstacle distribution in that lane; the expected lane change time corresponding to any lane is positively correlated with the length of the lane changeable interval of that lane; calculating the second traffic efficiency evaluation value from the candidate lane to the target lane according to the minimum number of lane changes, the expected travel time and the expected lane change time corresponding to each lane; the second traffic efficiency evaluation value is negatively correlated with the minimum number of lane changes, the expected travel time and the expected lane change time; Calculate a second traffic safety evaluation value from the candidate lane to the target lane according to the minimum number of lane changes; the second traffic safety evaluation value is negatively correlated with the minimum number of lane changes; Calculate a second lane evaluation value from the candidate lane to the target lane according to the second traffic efficiency evaluation value and the second traffic safety evaluation value; the second lane evaluation value is positively correlated with both the second traffic efficiency evaluation value and the second traffic safety evaluation value.

7. The method according to claim 4, characterized in that the step of calculating the lane evaluation value of each candidate lane according to the first lane evaluation value and the second lane evaluation value includes: For each candidate lane, determine a first weight for the first lane evaluation value and a second weight for the second lane evaluation value; Based on the first weight and the second weight, weight the first lane evaluation value and the second lane evaluation value, and use the weighted result as the lane evaluation value of the candidate lane.

8. The method according to claim 1, characterized in that the step of determining the lane evaluation value of each candidate lane when the vehicle exits the intersection based on the map data, the navigation information, the vehicle state information, and the obstacle state information includes: Input the map data, the navigation information, the vehicle state information, and the obstacle state information into a pre-trained lane selection model to obtain the probability values of each candidate lane when the vehicle exits the intersection; the lane selection model is pre-trained according to expert driving data; Use the probability values of each candidate lane as the lane evaluation values of the corresponding candidate lanes.

9. The method according to any one of claims 1-8, characterized in that the step of determining the regulation and control cost of each candidate lane based on the current trajectory planning algorithm and control tracking algorithm includes: Statistical the first historical lateral error of the trajectory planning algorithm and the second historical lateral error of the control tracking algorithm; According to the first historical lateral error and the second historical lateral error, calculate the regulation and control cost of each candidate lane, and the regulation and control cost is positively correlated with both the first historical lateral error and the second historical lateral error.

10. The method according to any one of claims 1-8, characterized in that the step of determining the regulation and control cost of each candidate lane based on the current trajectory planning algorithm and control tracking algorithm includes: For each candidate lane, generate multiple candidate trajectories from the current lane where the vehicle is located to the candidate lane based on the trajectory planning algorithm; For each candidate trajectory, calculate the planning cost of the candidate trajectory, and use the control tracking algorithm to control and track the candidate trajectory to obtain the corresponding control error, and calculate the tracking cost of the candidate trajectory according to the control error; the tracking cost is positively correlated with the control error; For each candidate trajectory, calculate the trajectory cost of the candidate trajectory according to the planning cost and the tracking cost of the candidate trajectory; the trajectory cost is positively correlated with the planning cost and positively correlated with the tracking cost; Determine the candidate trajectory with the lowest trajectory cost, and use the trajectory cost of this candidate trajectory as the planning and control cost of this candidate lane.

11. The method according to any one of claims 1-8, characterized in that the step of determining the optimal lane among the candidate lanes according to the traffic evaluation values and the planning and control costs of the candidate lanes includes: Among the candidate lanes, determine the candidate lanes whose planning and control costs meet the preset requirements as alternative lanes; Among the alternative lanes, determine the alternative lane with the highest traffic evaluation value as the optimal lane.

12. An intersection lane selection device, characterized in that the device includes: a data acquisition module for acquiring map data corresponding to the current intersection, as well as navigation information, self-vehicle state information, and obstacle state information; a traffic evaluation value determination module for determining the lane evaluation value of each candidate lane for the self-vehicle to drive out of the intersection based on the map data, the navigation information, the self-vehicle state information, and the obstacle state information, and for each candidate lane, determining the traffic evaluation value of this candidate lane based on the lane evaluation value of this candidate lane; a planning and control cost determination module for determining the planning and control costs of the candidate lanes based on the current trajectory planning algorithm and control tracking algorithm; an optimal lane determination module for determining the optimal lane among the candidate lanes according to the traffic evaluation values and the planning and control costs of the candidate lanes.

13. The device according to claim 12, characterized in that the device further includes: a safety evaluation value calculation module for calculating the safety evaluation value of each candidate lane; the safety evaluation value includes: a trajectory safety evaluation result, and / or, an intersection collision risk; the traffic evaluation value determination module is specifically configured to: For each candidate lane, determine the traffic evaluation value of this candidate lane based on the lane evaluation value and the safety evaluation value of this candidate lane.

14. The device according to claim 13, characterized in that when the safety evaluation value includes a trajectory safety evaluation result and an intersection collision risk, the safety evaluation value calculation module is specifically configured to: For each candidate lane, generate a planned trajectory from the current lane where the self-vehicle is located to this candidate lane, and calculate the trajectory safety evaluation result corresponding to the planned trajectory; Obtain the intersection collision risk values corresponding to each intersection stored in advance, and determine the intersection collision risk corresponding to the current intersection according to the intersection collision risk values corresponding to each intersection.

15. The device according to claim 12, characterized in that the traffic evaluation value determination module includes: a lane determination sub-module for determining the current lane where the self-vehicle is located, and according to the map data and the navigation information, determining each candidate lane for the self-vehicle to drive out of the intersection, and the target lane corresponding to the self-vehicle before reaching the next intersection; The lane evaluation value calculation sub-module is used to calculate the first lane evaluation value from the current lane to each candidate lane according to the ego-vehicle state information and the obstacle state information; and calculate the second lane evaluation value from each candidate lane to the target lane according to the ego-vehicle state information and the obstacle state information; the first lane evaluation value includes a first traffic efficiency evaluation value and / or a first traffic safety evaluation value; the second lane evaluation value includes a second traffic efficiency evaluation value and / or a second traffic safety evaluation value; The lane evaluation value combination sub-module is used to calculate the lane evaluation value of each candidate lane according to the first lane evaluation value and the second lane evaluation value.

16. The apparatus according to claim 15, wherein, the obstacle state information includes static obstacle state information and dynamic obstacle state information; when the first lane evaluation value includes a first traffic efficiency evaluation value and a first traffic safety evaluation value, the lane evaluation value calculation sub-module is specifically configured to: For each candidate lane, determine the length of the smooth continuous curve from the current lane to the candidate lane, and based on the length of the smooth continuous curve, determine the first traffic efficiency evaluation value from the current lane to the candidate lane; The first traffic efficiency evaluation value is negatively correlated with the length of the smooth continuous curve; For each candidate lane, calculate the static structure risk coefficient corresponding to the candidate lane according to the static obstacle state information, and calculate the dynamic traffic flow risk coefficient corresponding to the candidate lane according to the ego-vehicle state information and the dynamic obstacle state information; And calculate the first traffic safety evaluation value from the current lane to the candidate lane based on the static structure risk coefficient and the dynamic traffic flow risk coefficient; wherein, the first traffic safety evaluation value is negatively correlated with the static structure risk coefficient and the dynamic traffic flow risk coefficient; the static structure risk coefficient is related to the relative position of the static obstacle and the lane and the probability that the static obstacle is observed; the dynamic traffic flow risk coefficient is related to the collision time between the ego-vehicle and the dynamic obstacle; Calculate the first lane evaluation value from the current lane to the candidate lane according to the first traffic efficiency evaluation value and the first traffic safety evaluation value; the first lane evaluation value is positively correlated with both the first traffic efficiency evaluation value and the first traffic safety evaluation value.

17. The apparatus according to claim 15, wherein, when the second lane evaluation value includes a second traffic efficiency evaluation value and a second traffic safety evaluation value, the lane evaluation value calculation sub-module is specifically configured to: Determine the minimum number of lane changes from the candidate lane to the target lane, and calculate the expected travel time and expected lane change time corresponding to each lane passed by the host vehicle when changing lanes from the candidate lane to the target lane through the minimum number of lane changes according to the host vehicle state information and the obstacle state information; wherein, the expected travel time corresponding to any lane is related to the expected driving distance of the host vehicle in that lane and the obstacle distribution in that lane; the expected lane change time corresponding to any lane is positively correlated with the length of the lane changeable interval in that lane; Calculate the second traffic efficiency evaluation value from the candidate lane to the target lane according to the minimum number of lane changes, the expected travel time and expected lane change time corresponding to each lane; the second traffic efficiency evaluation value is negatively correlated with the minimum number of lane changes, the expected travel time and the expected lane change time; Calculate the second traffic safety evaluation value from the candidate lane to the target lane according to the minimum number of lane changes; the second traffic safety evaluation value is negatively correlated with the minimum number of lane changes; Calculate the second lane evaluation value from the candidate lane to the target lane according to the second traffic efficiency evaluation value and the second traffic safety evaluation value; the second lane evaluation value is positively correlated with both the second traffic efficiency evaluation value and the second traffic safety evaluation value.

18. The device according to any one of claims 12-17, wherein, the regulation and control cost determination module is specifically configured to: For each of the candidate lanes, generate multiple candidate trajectories from the current lane where the host vehicle is located to the candidate lane based on the trajectory planning algorithm; For each of the candidate trajectories, calculate the planning cost of the candidate trajectory, and use the control tracking algorithm to control and track the candidate trajectory to obtain the corresponding control error, and calculate the tracking cost of the candidate trajectory according to the control error; The tracking cost is positively correlated with the control error; For each of the candidate trajectories, calculate the trajectory cost of the candidate trajectory according to the planning cost and tracking cost of the candidate trajectory; The trajectory cost is positively correlated with the planning cost and positively correlated with the tracking cost; Determine the candidate trajectory with the lowest trajectory cost, and use the trajectory cost of the candidate trajectory as the regulation and control cost of the candidate lane.

19. A computer device, wherein, comprising: a memory and a processor, the memory and the processor are coupled; the memory is used to store one or more computer instructions; the processor is used to execute the one or more computer instructions to implement the intersection lane selection method according to any one of claims 1 to 11.

20. A readable storage medium, on which one or more computer instructions are stored, wherein, the instructions are executed by the processor to implement the intersection lane selection method according to any one of claims 1 to 11.

Images