Method and apparatus for planning vehicle driving trajectories

Abstract

This application relates to autonomous driving technology, and more particularly to a method and apparatus for planning vehicle trajectories in application scenarios involving vulnerable road users. The method for planning a vehicle trajectory according to one aspect of this application includes the following steps: determining the type of vulnerable road user in the vehicle's direction of motion and the predicted trajectory of the vulnerable road user in the next time period; determining a potential activity area of ​​the vulnerable road user in the next time period based on the type of vulnerable road user and the predicted trajectory; generating a avoidance transition zone surrounding the potential activity area; and selecting the candidate planning trajectory with the minimum cost from multiple candidate planning trajectories for the vehicle in the next time period that can bypass the potential activity area as the optimized planning trajectory, wherein the candidate planning trajectory crossing the avoidance transition zone has an additional avoidance cost compared to the candidate planning trajectory not crossing the avoidance transition zone.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of autonomous driving, and in particular, to a method and device for planning a driving trajectory of a vehicle in an application scenario where a vulnerable road user is present. BACKGROUND

[0002] A vulnerable road user (VRU) refers to a participant in road traffic who is vulnerable to injury due to lack of safety protection, mainly including pedestrians and two-wheeled vehicle users. According to relevant statistics, the number of VRU deaths accounts for more than half of the number of road traffic accident deaths worldwide, and therefore, improving the safety protection of VRUs is becoming a key issue in the field of active collision avoidance for vehicles.

[0003] In the field of autonomous driving, current research on active collision avoidance for vulnerable road users mainly focuses on VRU identification and motion prediction. However, due to the complexity of application scenarios, merely relying on improving the accuracy of VRU identification and motion prediction cannot meet the diverse application requirements. SUMMARY

[0004] An object of the present application is to provide a method and device for planning a driving trajectory of a vehicle, which can simultaneously improve the safety and comfort of autonomous driving.

[0005] According to one aspect of the present application, a method for planning a driving trajectory of a vehicle is provided, comprising the following steps:

[0006] determining a type of a vulnerable road user in a direction of motion of a ego vehicle and a predicted motion trajectory of the vulnerable road user in a next time period;

[0007] determining a potential activity area of the vulnerable road user in the next time period based on the type of the vulnerable road user and the predicted motion trajectory;

[0008] generating an avoidance transition area surrounding the potential activity area;

[0009] selecting a candidate planning trajectory with a minimum cost from a plurality of candidate planning trajectories of the ego vehicle in the next time period that can bypass the potential activity area as an optimized planning trajectory, wherein a candidate planning trajectory that traverses the avoidance transition area has an additional avoidance cost compared to a candidate planning trajectory that does not traverse the avoidance transition area.

[0010] Optionally, in the above method, the step of determining the type of the vulnerable road user and the predicted motion trajectory comprises:

[0011] receiving environmental state data of the ego vehicle;

[0012] determining a type and a motion state of the vulnerable road user based on the environmental state data;

[0013] determining the predicted motion trajectory based at least on the type and the motion state of the vulnerable road user.

[0014] Optionally, in the above method, the vulnerable road user is one or more of the following: a pedestrian, a non-motorized vehicle, and a motorcycle, and the motion state of the vulnerable road user includes one or more of the following: a speed, an acceleration, a motion direction, a direction light state, and a historical motion trajectory.

[0015] Optionally, in the above method, the potential activity region and the evasion transition region are represented as areas in a ST map or a SL map.

[0016] Further, the step of generating the evasion transition region includes:

[0017] determining a boundary of the evasion transition region based on at least one of the following: the type of the vulnerable road user, a user driving history, and a user setting;

[0018] determining an attribute value of each position in the evasion transition region based on at least one of the following: the type of the vulnerable road user, a user driving history, and a user setting, the attribute value representing a contribution component of the corresponding position to an additional evasion cost.

[0019] Further, the attribute value of each position decreases as a minimum distance between the position and a boundary of the potential activity region increases.

[0020] In addition to one or more of the features described above, in the above method, the step of selecting the optimized planning trajectory includes:

[0021] generating a plurality of candidate planning trajectories that are able to bypass the potential activity region;

[0022] calculating a base cost of the candidate planning trajectories using a cost function;

[0023] for a candidate planning trajectory that passes through the evasion transition region, determining a sum of the base cost and an additional evasion cost of the candidate planning trajectory as a corresponding cost;

[0024] determining a candidate planning trajectory with a minimum cost as the optimized planning trajectory.

[0025] Further, the additional evasion cost is a sum of contribution components of positions in the evasion transition region that are passed through by the candidate planning trajectory.

[0026] Further, the parameters of the cost function are determined based on at least one of a type of the vulnerable road user, a motion state of the vulnerable road user, a user setting, and a user driving history.

[0027] Further, the planned trajectory is set to reduce a longitudinal speed of the ego vehicle and increase a lateral distance between the ego vehicle and the vulnerable road user when approaching the vulnerable road user, and to increase the longitudinal speed of the ego vehicle and decrease the lateral distance between the ego vehicle and the vulnerable road user when moving away from the vulnerable road user.

[0028] According to another aspect of the present application, there is provided an apparatus for planning a trajectory of a vehicle, comprising:

[0029] a memory;

[0030] a processor coupled to the memory; and

[0031] a computer program stored on the memory and executable on the processor, by which the method as described above is implemented.

[0032] According to still another aspect of the present application, there is provided an apparatus for planning a trajectory of a vehicle, comprising:

[0033] a perception and prediction module configured to determine a type of a vulnerable road user in a motion direction of an ego vehicle and a predicted motion trajectory of the vulnerable road user in a next time period;

[0034] a planning decision module configured to perform the following operations:

[0035] determine a potential activity area of the vulnerable road user in the next time period based on the type of the vulnerable road user and the predicted motion trajectory;

[0036] generate an evasion transition area surrounding the potential activity area;

[0037] select a candidate planning trajectory with a minimum cost from a plurality of candidate planning trajectories of the ego vehicle in the next time period that can bypass the potential activity area as an optimized planning trajectory, wherein a candidate planning trajectory that traverses the evasion transition area has an additional evasion cost compared to a candidate planning trajectory that does not traverse the evasion transition area.

[0038] In some embodiments of the present application, more safety redundancy is introduced by employing the avoidance transition region. In addition, since the additional avoidance cost is taken into account, it is more likely to select a trajectory that involves longitudinal deceleration and lateral avoidance in the planning process, thus better matching the driving habits of human users. Furthermore, since the basic framework of existing path planning techniques can be followed, the existing mature algorithms can be fully utilized (e.g., the existing cost function can be directly used to calculate the base cost of the candidate planning path without modification), which is advantageous for reducing development costs and shortening the development cycle. BRIEF DESCRIPTION OF DRAWINGS

[0039] The above and / or other aspects and advantages of the present application will become more apparent and more readily appreciated from the following description of various aspects of the present application, taken in conjunction with the accompanying drawings in which like reference numerals refer to like elements throughout. The accompanying drawings include:

[0040] Figure 1 A flowchart of a method for planning a vehicle travel trajectory according to some embodiments of the present application.

[0041] Figure 2 A flowchart of a method for determining the type and predicted motion trajectory of a vulnerable road user according to some other embodiments of the present application.

[0042] Figure 3 A schematic diagram of a potential activity region and an avoidance transition region.

[0043] Figure 4 An exemplary potential activity region and avoidance transition region represented in a ST graph are shown.

[0044] Figure 5 A flowchart of a method for generating an avoidance transition region according to still some other embodiments of the present application.

[0045] Figure 6 A flowchart of a method for generating an optimized planning trajectory according to still some other embodiments of the present application.

[0046] FIG. 7 shows an example of an optimized planning trajectory obtained using the above-described embodiments and using an existing planning strategy.

[0047] Figure 8 A schematic block diagram of a typical computer system.

[0048] Figure 9 A schematic block diagram of an apparatus for planning a vehicle travel trajectory. DETAILED DESCRIPTION

[0049] The present application will be understood more fully from the following detailed description, taken in conjunction with the accompanying drawings, in which:

[0050] In this specification, terms such as "comprising" and "including" mean that the technical solution of the present application does not exclude other units and steps not directly or explicitly listed in the specification and claims.

[0051] Unless otherwise specified, terms such as "first" and "second" do not indicate the order of units in time, space, size, etc., but only distinguish the units.

[0052] In this application, "vulnerable road users" can be broadly understood as dangerous objects that are prone to road traffic injuries, usually referring to pedestrians, non-motor vehicles, motorcycles, etc.

[0053] Figure 1 A flowchart of a method for planning a vehicle driving trajectory according to some embodiments of the present application. Exemplarily, the steps of the method are performed by an autonomous driving domain controller. In a typical domain controller-based vehicle architecture, the autonomous driving domain controller is responsible for data processing operations and judgment capabilities required for autonomous driving, such as including processing capabilities for data obtained from millimeter wave radar, camera, laser radar, GPS, inertial navigation, etc.

[0054] Figure 1 The method shown includes the following steps:

[0055] Step 101 : Determining the type of vulnerable road user and the predicted motion trajectory

[0056] In this step, the type and predicted motion trajectory of the vulnerable road user can be determined by means of Figure 2 The subprogram shown is used to determine the type and predicted motion trajectory of the vulnerable road user.

[0057] Specifically, in step 201, the autonomous driving domain controller of the ego vehicle receives the environmental state data of the ego vehicle. The environment generally includes static elements such as road layout and lane structure, and dynamic elements such as other vehicles, pedestrians and other types of road users. Static elements can be obtained through high-definition maps containing lane-level information, and dynamic elements can be obtained using sensors such as cameras, millimeter wave radars, laser radars, etc.

[0058] Figure 2The method flowchart then proceeds to step 202. In this step, the ego vehicle's autonomous domain controller determines the type and motion state of the vulnerable road user in the ego vehicle's motion direction based on the environment state data (e.g., images captured by cameras, point cloud data acquired by radars, and lane data extracted from high-definition maps, etc.). Examples of the motion state described herein include, but are not limited to, the vulnerable road user's speed, acceleration, motion direction, direction light state, and historical motion trajectory, etc.

[0059] The autonomous domain controller then proceeds to step 203 to predict the vulnerable road user's motion trajectory in the next time period. The length of the time period can vary depending on the application scenario, and its order of magnitude can be, for example, hundreds of milliseconds or seconds. The prediction of the motion trajectory is based at least on the type and motion state of the vulnerable road user. In addition, other factors such as road layout, lane structure, and traffic light state, etc. can also be introduced in the prediction of the motion trajectory. Illustratively, when the vulnerable road user is a pedestrian and far away from an intersection, the predicted motion trajectory is usually limited within the range of the sidewalk.

[0060] Step 102: Determination of the potential activity area

[0061] Figure 3 An example of a potential activity area is shown. As shown, the ego vehicle EV is driving on the road, and the area VRU identified in dark color in the right front of the ego vehicle is the area associated with the vulnerable road user or its potential activity area in the next time period. Figure 3

[0062] In some embodiments, the potential activity area can be represented as an area in an ST graph or an SL graph. The ST graph is a two-dimensional coordinate plane with time (T) as the horizontal axis and the longitudinal distance (S) of the planned path as the vertical axis. The SL graph is a two-dimensional coordinate plane with the longitudinal distance (S) of the planned path as the horizontal axis and the lateral distance (L) of the planned path as the vertical axis, which can be used in combination with the ST graph to generate a three-dimensional planned trajectory (longitudinal dimension, lateral dimension, and time dimension) of the ego vehicle. By constructing the ST graph, the velocity planning can be projected onto a two-dimensional plane to be solved in an optimization manner, and the predicted trajectory of the obstacle can also be projected onto the two-dimensional plane to make a reasonable decision. In addition, in the ST graph, the slope of the curve represents the derivative of the S value with respect to time (i.e., the speed), so the relationship between the speed and the path can be constructed through the S value, and thus the two can be better integrated into a trajectory.

[0063] Figure 4 An example of a potential activity area represented in an ST graph is shown. In the example shown, the potential activity area VRU is identified as a dark area. Figure 4

[0064] ​​In step 102, the autonomous driving domain controller determines a potential activity region according to the type of the vulnerable road user and the predicted motion trajectory determined in step 101. Exemplarily, since the vulnerable road user is an object occupying a certain space, the size of the space occupied by the vulnerable road user can be estimated from its type and the possible activity range (i.e. the potential activity region) of the vulnerable road user in the next time period can be obtained in combination with the predicted motion trajectory.

[0065] Step 103: Generation of the avoidance transition area

[0066] Existing VRU active collision avoidance strategies mainly consider safety, that is, taking no collision with the vulnerable road user as the optimization objective of trajectory planning and motion control. However, in actual driving, only considering safety is not enough. For example, according to the commonly used planning strategy, it is only necessary to ensure that the planned trajectory of the ego vehicle does not cross the potential activity region of the VRU, and no further planning (such as deceleration or acceleration) is made on the speed. However, this way is different from the driving behavior habit of human beings (when the ego vehicle approaches the vulnerable road user, the human driver generally subconsciously reduces the vehicle speed and / or performs a slight lateral avoidance operation), thereby bringing about an adverse user experience. In addition, the existing collision avoidance planning method needs to take relatively aggressive braking and steering control when dealing with the sudden behavior (such as sudden change of motion direction or motion speed) of the vulnerable road user, which will have an adverse effect on the driving safety and comfort.

[0067] In some embodiments of the present application, safety is improved and driving comfort is improved by introducing an avoidance transition region in trajectory planning. The avoidance transition region described herein is a region surrounding or surrounding the potential activity region, and when the planned trajectory passes through the avoidance transition region, an additional avoidance cost will be increased compared with the region outside the avoidance transition region. The additional avoidance cost will affect the optimization result of the planned trajectory, that is, the planned trajectory passing through the avoidance transition region will be less likely to be selected as the optimized planned trajectory.

[0068] Figure 3 An example of the avoidance transition region is shown. Referring to Figure 3 In the region identified in dark color, the VRU is surrounded by the region A (avoidance transition region). Figure 4 An example of the avoidance transition region represented in the ST graph is shown. Referring to Figure 4 In the example shown, the potential activity region VRU is also surrounded by the region A (avoidance transition region).

[0069] In step 103, the autonomous driving domain controller generates an avoidance transition region surrounding the potential activity region. The avoidance transition region can be generated by means of Figure 5 The subprogram shown can be used to generate the avoidance transition region.

[0070] Specifically, in step 501, the autonomous driving domain controller obtains the boundary of the avoidance transition area, for example, by extending the boundary of the potential activity area. The degree of boundary extension can be determined based on various factors. In some embodiments, the factors considered include at least one of the type of vulnerable road user, user settings, and user profile. For example, for more mobile or unpredictable vulnerable road users (e.g., children, non-motorized vehicles, and motorcycles), the boundary extension coefficient of the potential activity area (e.g., the ratio of the area of ​​the avoidance transition area to the area of ​​the potential activity area, or the ratio of the perimeter of the avoidance transition area to the area of ​​the potential activity area) can be set larger. Alternatively, a user's driving history can be analyzed to obtain their driving habits, thereby obtaining a customized boundary extension coefficient for that user. Furthermore, the boundary extension coefficient can also be set by the user.

[0071] It should be pointed out that, although in Figure 5 In the illustrated embodiment, the expansion of the potential active region is isotropic, but this is not mandatory. In some applications, the expansion can be anisotropic, meaning the degree of expansion varies in different directions.

[0072] As mentioned above, additional avoidance costs are incurred when the planned trajectory passes through the avoidance transition zone. The contribution of each location to the additional avoidance cost is represented by assigning attribute values ​​to each location within the avoidance transition zone. Figure 4 An example of the attribute value is shown. See also Figure 4 The attribute values ​​at each location decrease as the minimum distance between that location and the boundary of the potential activity area increases. It should be noted that... Figure 4 The linear relationship between the attribute value and the minimum distance, as well as the possible values ​​of the attribute, given in the example are all exemplary. For instance, the relationship between the attribute value and the minimum distance can also be made non-linear depending on the needs of the application scenario.

[0073] After executing step 501, Figure 5 The method then proceeds to step 502, in which the autonomous driving domain controller determines attribute values ​​for each location within the avoidance transition zone. In some embodiments, the attribute values ​​may be determined based on at least one of the type of vulnerable road user, user driving history, and user settings.

[0074] Step 104: Generation of the planned trajectory

[0075] In this step, you can use Figure 6 The subroutine shown is for generating optimized trajectories for the self-driving vehicle.

[0076] Specifically, in step 601, the autonomous driving domain controller generates a plurality of candidate planning trajectories that can bypass the potential active region. There are various planning trajectory strategies that can be used to generate these candidate planning trajectories, which are not described in detail here. In some embodiments, the candidate planning trajectories are set to reduce the longitudinal speed of the ego vehicle and increase the lateral distance between the ego vehicle and the vulnerable road user when approaching the vulnerable road user, and to increase the longitudinal speed of the ego vehicle and reduce the lateral distance between the ego vehicle and the vulnerable road user when moving away from the vulnerable road user.

[0077] Subsequently, step 602 is entered, in which the autonomous driving domain controller calculates the base cost of the plurality of candidate planning trajectories generated in step 601 using a cost function. As described above, an additional avoidance cost will be added when the planning trajectory passes through the avoidance transition region. The base cost described here refers to the cost of the candidate planning trajectory without taking into account the additional avoidance cost. When calculating the base cost, a cost function commonly used in current trajectory planning techniques can be optionally used. The cost function can take into account various factors, for example, including but not limited to reaching the target, smoothness, collision avoidance, centripetal acceleration, lateral offset, comfort, etc. In addition, in some embodiments, the parameters of the cost function (such as the cost coefficient within a specified distance, etc.) can be determined or adjusted based on at least one of the type of the vulnerable road user, the motion state, the user settings, and the user driving history.

[0078] Then step 603 is entered, in which for each candidate planning trajectory that passes through the avoidance transition region, the sum of the corresponding base cost and the additional avoidance cost is determined as the corresponding cost.

[0079] In some embodiments, for each candidate planning trajectory that passes through the avoidance transition region, its additional avoidance cost can be determined in the following manner:

[0080]

[0081] wherein Extra_Cost i is the additional avoidance cost of the i-th candidate planning trajectory that passes through the avoidance transition region, assuming that the i-th candidate planning trajectory that passes through the avoidance transition region passes through n positions within the avoidance transition region, μ ij is the attribute value of the j-th position among the n positions, i.e., the contribution component of the j-th position to the additional avoidance cost.

[0082] After step 603 is performed, the cost of each candidate planning trajectory that has taken into account the additional avoidance cost will be obtained. Therefore, Figure 6 The method flow shown enters step 604. In this step, the autonomous driving domain controller determines the candidate planning trajectory with the minimum cost as the optimized planning trajectory.

[0083] In the above shown embodiments, the adoption of the avoidance transition region instead of the potentially active region amounts to the introduction of more safety redundancy. At the same time, since the additional avoidance cost is taken into account for the candidate planning path that crosses the avoidance transition region, it is more likely for the planning process to select a trajectory that involves longitudinal deceleration and lateral avoidance, thus better matching the driving habits of human users. Moreover, the above embodiments can follow the basic framework of existing path planning techniques, thus can take full advantage of the existing mature algorithms (e.g. the existing cost function can be directly used to calculate the base cost of the candidate planning path without modification), which is beneficial for the reduction of development cost and the shortening of development cycle.

[0084] Figure 7A and 7B respectively show examples of the optimized planning trajectories obtained by adopting the above embodiments and by adopting the existing planning strategy, wherein, Figure 7A and 7B the left half of each figure is a top view and the right half is a front view. By comparing Figure 7A and 7B it can be seen that, during the travel of the ego vehicle EV from right to left, when the ego vehicle EV approaches the vulnerable road user VRU, the optimized planning trajectory provided according to the above embodiments involves a small-amplitude lateral avoidance operation.

[0085] Figure 8 is a schematic block diagram of a typical computer system, which can be used to implement the above-mentioned device for planning a vehicle travel trajectory or the autonomous driving domain controller. As shown in Figure 8 , the computer system 80 comprises a memory 810 (e.g. a non-volatile memory such as a flash memory, a ROM, a hard drive, a magnetic disk, an optical disk), a processor 820 and a computer program 830.

[0086] The memory 810 stores the computer program 830 which can be executed by the processor 820. The processor 820 is configured to run the computer program 830 stored on the memory 820. By running the computer program 830, one or more steps included in the above-mentioned method can be implemented. Figures 1-6

[0087] Figure 9 is a schematic block diagram of a device for planning a vehicle travel trajectory, which can be used to implement the above-mentioned autonomous driving domain controller. As shown in Figure 9 , the device for planning a vehicle travel trajectory 90 comprises a perception and prediction module 910 and a planning decision module 920.

[0088] The perception and prediction module 910 is configured to determine the type of the vulnerable road user in the motion direction of the ego vehicle and the predicted motion trajectory of the vulnerable road user in the next time period. ​

[0089] The planning decision module 920 is configured to perform the following operations:

[0090] determine a potential activity area of the vulnerable road user in the next time period based on the type of the vulnerable road user and the predicted motion trajectory;

[0091] generate an evasion transition area surrounding the potential activity area;

[0092] select a candidate planning trajectory with minimum cost from a plurality of candidate planning trajectories of the ego vehicle in the next time period that can bypass the potential activity area as the optimized planning trajectory, wherein a candidate planning trajectory that traverses the evasion transition area has additional evasion cost compared to a candidate planning trajectory that does not traverse the evasion transition area.

[0093] In addition to the features described above, the perception and prediction module 910 and the planning decision module 920 can be configured to have the features described above by means of Figures 1-6 part or all of the embodiments described above.

[0094] According to another aspect of the present application, there is also provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, can implement one or more steps of the method described above by means of Figures 1-6 the embodiments described above.

[0095] The computer readable storage medium described in the present application includes various types of computer storage media, which can be any available media that can be accessed by a general or special purpose computer. For example, the computer readable storage medium can include a RAM, a ROM, an EPROM, an E2PROM, a register, a hard disk, a removable disk, a CD-ROM or other optical disk storage, a magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code means in the form of instructions or data structures and that can be accessed by a general or special purpose computer, or a general or special purpose processor. Combinations of the above should also be included within the scope of the computer readable storage medium. The exemplary storage medium is coupled to the processor so that the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium can be integral to the processor. The processor and the storage medium can reside in an ASIC. The ASIC can reside in a user terminal. In the alternative, the processor and the storage medium can reside as discrete components in a user terminal.

[0096] Those skilled in the art will understand that the various illustrative logical blocks, modules, circuits, and algorithm steps described herein can be implemented as electronic hardware, computer software, or combinations of both.

[0097] To show the interchangability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. The functionality described with respect to a particular illustrative component, block, module, circuit, or step can be embodied in whole or in part in hardware, software, firmware, or any combination thereof. Not all components, blocks, modules, circuits, and steps are necessarily required, and not all components, blocks, modules, circuits, and steps can necessarily be employed to implement the functionality described. Functionality may

[0098] Although only a few embodiments of the present application have been described above, it should be understood that many variations and modifications of the preferred embodiment and applications of the processes described above are possible. For example, although the present application has been described in the context of a single application, the application can be implemented in the context of multiple applications. Furthermore, the processes described above can be implemented in the context of a single computer or multiple computers. Other sequences of operations can also be possible. Accordingly, other implementations and applications of the present application are within the scope of the following claims.

[0099] The embodiments and examples set forth herein were presented in order to best explain the embodiments and their particular application and to thereby enable those skilled in the art to make and use the present application. However, those skilled in the art will recognize that the foregoing description and examples have been presented for the purposes of illustration and example only. The description as set forth is not intended to be exhaustive or to limit the application to the precise forms disclosed.

Claims

1. A method for planning a driving trajectory of a vehicle, characterized in that, comprising the steps of: determining a type of a vulnerable road user in a direction of motion of the ego vehicle and a predicted trajectory of the vulnerable road user in a next time period; determining a potential activity region of the vulnerable road user in the next time period based on the type and the predicted trajectory of the vulnerable road user; generating an evasion transition region surrounding the potential activity region; selecting a candidate planning trajectory with a minimum cost from a plurality of candidate planning trajectories of the ego vehicle in the next time period that can bypass the potential activity region as an optimized planning trajectory, wherein a candidate planning trajectory that traverses the evasion transition region has an additional evasion cost compared to a candidate planning trajectory that does not traverse the evasion transition region; the potential activity region and the evasion transition region are represented as regions in an ST map or an SL map, wherein the ST map has a time (T) as a horizontal axis and a longitudinal distance (S) of a planning path as a vertical axis, and the SL map has the longitudinal distance (S) of the planning path as a horizontal axis and a lateral distance (L) of the planning path as a vertical axis.

2. The method of claim 1, wherein, the step of determining the type and the predicted trajectory of the vulnerable road user comprises: receiving environment state data of the ego vehicle; determining the type and a motion state of the vulnerable road user from the environment state data; determining the predicted trajectory based on at least the type and the motion state of the vulnerable road user.

3. The method of claim 2, wherein, the vulnerable road user is one or more of a pedestrian, a non-motorized vehicle, and a motorcycle, and the motion state of the vulnerable road user comprises one or more of a speed, an acceleration, a direction of motion, a direction light state, and a historical trajectory of motion.

4. The method of claim 1, wherein, the step of generating the evasion transition region comprises: determining a boundary of the evasion transition region based on at least one of the type of the vulnerable road user, a user driving history, and a user setting; determining an attribute value of each location in the evasion transition region based on at least one of the type of the vulnerable road user, the user driving history, and the user setting, the attribute value representing a contribution component of the corresponding location to the additional evasion cost.

5. The method of claim 4, wherein, the attribute value of each location decreases as a minimum distance between the location and the boundary of the potential activity region increases.

6. The method of any of claims 1-5, wherein, the step of selecting the optimized planning trajectory comprises: generating a plurality of candidate planning trajectories that can bypass the potential activity region; calculating a base cost of the candidate planning trajectories using a cost function; for a candidate planning trajectory that traverses the evasion transition region, determining a sum of the base cost and the additional evasion cost of the corresponding candidate planning trajectory as a corresponding cost; determining the candidate planning trajectory with the minimum cost as the optimized planning trajectory.

7. The method of claim 6, wherein, the additional evasion cost is a sum of contribution components of locations in the evasion transition region traversed by the candidate planning trajectory.

8. The method of claim 6, wherein, a parameter of the cost function is determined based on at least one of the type of the vulnerable road user, the motion state, the user setting, and the user driving history.

9. The method of claim 6, wherein, The planned trajectory is configured to decrease longitudinal speed of the ego vehicle and increase lateral distance between the ego vehicle and the vulnerable road user when approaching the vulnerable road user, and to increase longitudinal speed of the ego vehicle and decrease lateral distance between the ego vehicle and the vulnerable road user when moving away from the vulnerable road user.

10. An apparatus for planning a trajectory of a vehicle, comprising: a memory; a processor coupled to the memory; and a computer program stored on the memory and executable on the processor, by which the method of any one of claims 1-9 is implemented.

11. An apparatus for planning a trajectory of a vehicle, comprising: a perception and prediction module configured to determine a type of a vulnerable road user in a direction of motion of an ego vehicle and a predicted trajectory of the vulnerable road user in a next time period; a planning decision module configured to: determine a potential activity region of the vulnerable road user in the next time period based on the type of the vulnerable road user and the predicted trajectory; generate an evasion transition region that encloses the potential activity region; select a candidate planned trajectory with a minimum cost as an optimized planned trajectory from a plurality of candidate planned trajectories of the ego vehicle in the next time period that can avoid the potential activity region, wherein a candidate planned trajectory that traverses the evasion transition region has an additional evasion cost compared to a candidate planned trajectory that does not traverse the evasion transition region; the potential activity region and the evasion transition region are represented as regions in an ST or SL map; wherein the ST map has time (T) as a horizontal axis and longitudinal distance (S) of a planned path as a vertical axis, and the SL map has longitudinal distance (S) of a planned path as a horizontal axis and lateral distance (L) of a planned path as a vertical axis.

12. A computer readable storage medium having stored thereon a computer program, which computer program is executed by a processor to implement the method of any one of claims 1-9.

Images