A method for screening key obstacles in an intelligent driving system

Abstract

This invention discloses a method for screening key obstacles in an intelligent driving system. The method establishes a topological relationship between the vehicle, obstacles, and lanes based on the vehicle's environmental information. Filters are set within this topological relationship, and the appropriate filter is selected based on the vehicle's driving intentions to obtain relevant key obstacle information. Corresponding labels are then assigned to the obstacles, and the key obstacle information, labels, and corresponding topological structures are output. This method can filter out obstacles that significantly impact the vehicle based on different driving intentions, improving the efficiency of obstacle screening in intelligent driving systems. Furthermore, it can handle the screening of obstacles on lane lines, reducing safety hazards caused by incorrect or missed screening of lane-crossing obstacles. Simultaneously, by using single or combined filters, it can screen out obstacles that the vehicle needs to focus on in a given scenario, further improving the efficiency of obstacle screening in intelligent driving systems and reducing safety hazards caused by incorrect or missed screening of lane-crossing obstacles.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent vehicles, and particularly relates to a key obstacle screening method in an intelligent driving system. BACKGROUND

[0002] In recent years, intelligent driving vehicles have gradually become one of the research hotspots in the world. With the improvement of the intelligent driving system level of vehicles and the progress of automatic driving related technologies, people are increasingly expecting a safe intelligent driving system to land as soon as possible to bring convenience to people's travel.

[0003] In the entire intelligent driving system, the key obstacle refers to an obstacle of interest to an intelligent vehicle (host vehicle); the key obstacle screening refers to screening the obstacle of interest to the host vehicle from the set of obstacles within the sensor range obtained by the upstream perception module, reducing redundant calculations, and affecting the effect of the downstream decision planning control module, and is an indispensable key function. Therefore, reasonable and accurate screening of the key obstacle will greatly improve the efficiency and safety of intelligent driving.

[0004] In the prior art, most of the key obstacle screening methods screen the obstacle that has the greatest impact on the host vehicle lane by lane, that is, the obstacle closest to the host vehicle in each lane. Although this method simply and roughly solves the obstacle screening, it screens out the obstacles that are not closest to the host vehicle from the obstacle layer, which also screens out more possibilities of intelligent driving schemes, resulting in the decision planning space of intelligent driving being compressed, and the decision planning being too conservative and single. In addition, since the lane screening will cause the obstacles located on the lane line to be unable to be processed or screened incorrectly, safety hazards are caused. SUMMARY

[0005] To solve the above problems, the main purpose of the present application is to design a key obstacle screening method in an intelligent driving system, based on the requirements of different driving intentions, a series of filters are designed, and the obstacles that have a great impact on the host vehicle under different intentions are screened out through the combination of the filters, solving the limitation problem caused by the re-screening of obstacles based on the generated trajectory in the prior art, and the problem of safety hazards caused by the lane screening that causes the obstacles located on the lane line to be unable to be processed or screened incorrectly.

[0006] In order to achieve the above purpose, the application adopts the following technical scheme:

[0007] A key obstacle screening method in an intelligent driving system, which screens out the key obstacle that has the most significant impact on the host vehicle according to the driving scene and driving intention of the host vehicle, the key obstacle refers to the obstacle closest to the host vehicle, and the specific steps of the method include the following:

[0008] Step 1: obtaining the environment information of the intelligent vehicle from the environment perception layer in the intelligent driving system of the host vehicle;

[0009] Step 2: establishing the topological relationship between the host vehicle and the obstacles and the lanes based on the environment information obtained in step 1;

[0010] Step 3: setting a plurality of filters for the topological relationship established in step 2;

[0011] Step 4: selecting the filters set in step 3 to screen the corresponding key obstacle information according to the different driving intentions of the host vehicle;

[0012] Step 5: labeling the key obstacle screened in step 4, and the label information includes the filter information corresponding to the key obstacle;

[0013] Step 6: outputting the key obstacle information obtained in step 4, the key obstacle label information obtained in step 5, and the topological structure information between the host vehicle and the lanes corresponding to the key obstacle, that is, completing the screening and output of the key obstacle once.

[0014] As a further description of the application, in step 1, the environment information of the intelligent vehicle includes the information of all obstacles and lane information within the current perception range of the host vehicle, and the environment information of the intelligent vehicle is stored in the database of the intelligent driving system.

[0015] As a further description of the application, in step 2, the topological relationship between the host vehicle and the obstacles and the lanes includes the topological relationship between the host vehicle and the obstacles, and the topological relationship between the host vehicle, the obstacles and the lanes.

[0016] The topological relationship between the host vehicle and the obstacles includes the position relationship in space and the speed relationship in time between the obstacles and the host vehicle; and the topological relationship between the host vehicle, the obstacles and the lanes includes the position relationship between the host vehicle, the obstacles and the lane boundary, the lane line and the reference line.

[0017] As a further description of the application, the position relationship in space between the obstacles and the host vehicle includes the front obstacle, the rear obstacle, the front-front obstacle, the rear-rear obstacle, the left obstacle, the left-front obstacle, the left-rear obstacle, the right obstacle, the right-front obstacle, the right-rear obstacle relative to the host vehicle, and the longitudinal distance and the lateral distance of the obstacles from the host vehicle.

[0018] As a further description of the application, the topological relationship between the lanes includes the current lane of the host vehicle, the left lane and the left-left lane on the left side of the current lane, and the right lane and the right-right lane on the right side of the current lane.

[0019] As a further description of the present application, in step 3, the filter includes a longitudinal distance filter, a lateral distance filter, a lane filter, and a lead obstacle filter.

[0020] As a further description of the present application, the longitudinal distance filter is capable of screening out the obstacles in front of the host vehicle within a longitudinal range of s front and the obstacles in the rear of the host vehicle within a longitudinal range of s back , and filtering out the remaining obstacles outside the range of s back , s front .

[0021] The lateral distance filter is capable of screening out the obstacles in the left of the host vehicle within a lateral range of l upper and the obstacles in the right of the host vehicle within a lateral range of l lower , and filtering out the remaining obstacles outside the range of l lower , l upper .

[0022] The lane filter is capable of screening out all the obstacles in the designated lane, and filtering out the obstacles in the remaining lanes.

[0023] The lead obstacle filter is capable of screening out the first obstacle in front of the host vehicle in the lane of interest, and filtering out the remaining obstacles.

[0024] As a further description of the present application, in step 4, the selected filter in step 3 is used to screen out the key obstacle, including using the filters in step 3 individually or in combination.

[0025] As a further description of the present application, in step 5, the label is specifically represented as follows:

[0026]

[0027] wherein, "longitudinal distance filter" represents that the obstacle is screened out by the longitudinal distance filter; "lateral distance filter" represents that the obstacle is screened out by the lateral distance filter; "lane filter" represents that the obstacle is screened out by the lane filter; and "lead obstacle filter" represents that the obstacle is screened out by the lead obstacle filter.

[0028] As a further description of the present application, the label can include the obstacle screened out by one or more filters.

[0029] Compared with the prior art, the present application has the following technical effects:

[0030] This invention provides a method for screening key obstacles in an intelligent driving system. Instead of simply selecting the obstacle closest to the driver vehicle in each lane, it selects individual or combined filters based on the needs of different driving intentions to screen out obstacles that have a significant impact on the driver vehicle under different intentions. This improves the screening efficiency of obstacles in the intelligent driving system and reduces safety hazards caused by incorrect or missed screening of obstacles that cross the lane. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the overall screening method of the present invention;

[0032] Figure 2 This is a schematic diagram of the vehicle coordinate system in this invention;

[0033] Figure 3 This is a schematic diagram illustrating the screening of key obstacles when the main vehicle intends to change lanes to the left in this invention. Detailed Implementation

[0034] The present invention will now be described in detail with reference to the accompanying drawings:

[0035] A method for screening key obstacles in an intelligent driving system, referenced Figures 1-3 As shown, this method filters out the key obstacles that have the most significant impact on the driving of the main vehicle based on the driving scenario and driving intention of the main vehicle. The key obstacles refer to the obstacles that are closest to the main vehicle.

[0036] like Figure 1 As shown, the specific steps of this method include the following:

[0037] Step 1: Obtain environmental information of the intelligent vehicle from the environmental perception layer of the main vehicle's intelligent driving system;

[0038] Step 2: Based on the environmental information obtained in Step 1, establish the topological relationship between the main vehicle and obstacles and lanes;

[0039] Step 3: Set up several filters for the topology established in Step 2;

[0040] Step 4: Based on the different driving intentions of the main vehicle, select the filter set in Step 3 to filter and obtain the corresponding key obstacle information;

[0041] Step 5: Label the key obstacles selected in Step 4. The label information includes obtaining the filter information corresponding to the key obstacle.

[0042] Step 6: Output the key obstacle information obtained in Step 4, the key obstacle label information obtained in Step 5, and the topology information between the main vehicle and the lane corresponding to the key obstacle, thus completing one key obstacle screening and output.

[0043] Specifically, this embodiment will analyze each of the above steps in detail, and the analysis content is as follows:

[0044] In step 1, the environmental information of the intelligent vehicle includes information on all obstacles and lane information within the current perception range of the main vehicle, and the environmental information of the intelligent vehicle is stored in the database of the intelligent driving system.

[0045] The obstacle information mentioned above includes, but is not limited to, obstacle ID, length, width, height, position coordinates, speed, and heading; lane information includes, but is not limited to, lane data of structured lanes, lane width, lane length, lane line position, road boundary, slope, and road marking information; it should be noted that, in order to meet the needs of intelligent vehicles to collect the above information, the intelligent vehicle equipped with this embodiment should be equipped with a vehicle information collection device, which is used to collect the above environmental information.

[0046] Second, in step 2, the topological relationship between the main vehicle and the obstacle and the lane includes: the topological relationship between the main vehicle and the obstacle, and the topological relationship between the main vehicle, the obstacle and the lane.

[0047] 1. Topological relationship between the main vehicle and obstacles

[0048] The topological relationship between the main vehicle and the obstacles includes the spatial positional relationship and temporal velocity relationship between the obstacles and the main vehicle; the spatial positional relationship between the obstacles and the main vehicle includes obstacles in front of the main vehicle, obstacles behind the main vehicle, obstacles in front of the ...

[0049] The spatial positional relationship between the obstacle vehicle and the main vehicle in this embodiment includes topological types such as front vehicle, rear vehicle, front-front vehicle, rear-rear vehicle, left vehicle, left-front vehicle, left-rear vehicle, right vehicle, right-front vehicle, and right-rear vehicle; the longitudinal distance Δs and lateral distance Δl between the obstacle and the main vehicle are calculated using the following formulas:

[0050] Δs=s obs -s ego

[0051] Δl=l obs -l ego

[0052] Among them, s ego , l ego The longitudinal and lateral coordinates of the main vehicle, s obs , l obs The longitudinal and lateral coordinates of the obstacle's projection onto the reference line where the main vehicle is located.

[0053] 2. Topological relationships between the main vehicle, obstacles, and lanes

[0054] The topological relationships between the main vehicle, obstacles, and lanes include the positional relationships between the main vehicle, obstacles, lane boundaries, lane lines, and reference lines; the topological relationships between lanes include the current lane of the lane where the main vehicle is located, the left lane to the left of the current lane, the left-left lane, the right lane to the right of the current lane, and the right-right lane.

[0055] Third, in step 3, the filters include longitudinal distance filters, lateral distance filters, lane filters, and nearest preceding vehicle filters.

[0056] The aforementioned longitudinal distance filter can filter obstacles within the longitudinal range of the main vehicle, the lateral distance filter can filter obstacles within the lateral range of the main vehicle, the lane filter can filter obstacles in a specified lane, and the nearest preceding vehicle filter can filter the first obstacle in front of the main vehicle in the lane of interest that is closest to the main vehicle.

[0057] It should be noted that the lateral and longitudinal directions in this embodiment are directions in the Frenet coordinate system. Specifically, the Frenet coordinate system uses the lane centerline as a reference. S represents the direction of the longitudinal guide line, which in this embodiment refers to the direction of the road centerline where the vehicle travels, i.e., longitudinal; L represents the direction perpendicular to the lane centerline, i.e., lateral. Figure 2 As shown.

[0058] More specifically, this embodiment will analyze the four filters mentioned above, as follows:

[0059] 1. Longitudinal Distance Filter: Capable of filtering the area in front of the vehicle within its longitudinal range. front Obstacles within range and behind s back Obstacles within range are filtered out (s back s front Other obstacles outside the designated area.

[0060] The set of obstacles Ω filtered by the longitudinal distance filter obs_lon for:

[0061] {Ω obs_lon |Ω obs_lon ∈Ω, s obs ∈(s back s front )}

[0062] Where Ω is the total set of all obstacles within the sensing range, s obs The longitudinal coordinates of the obstacle's projection onto the reference line where the main vehicle is located.

[0063] 2. Lateral Distance Filter: This filter can identify vehicles that are horizontally positioned to the left of the main vehicle. upper Within the range and to the right l lower Obstacles within the range, and filter them out (l lower , l upper Other obstacles outside the designated area.

[0064] The set of obstacles Ω filtered by the lateral distance filter obs_lat for:

[0065] {Ω obs_lat |Ω obs_lat ∈Ω, l obs ∈(l lower , l upper )}

[0066] Where Ω represents the total set of all obstacles within the sensing range, l obs The longitudinal coordinates of the obstacle's projection onto the reference line where the main vehicle is located.

[0067] It should be noted that the aforementioned lateral range is not constrained by lane lines. That is, even if there is an obstacle that crosses the lane lines, as long as the lateral coordinate of the obstacle's projection onto the current lane reference line is within (l lower , l upper Within the specified range, obstacles located on the lane lines (over the lines) can be screened out, avoiding the inability to handle or the incorrect screening of such safety hazards.

[0068] 3. Lane Filter: It can filter out all obstacles in a specified lane and filter out obstacles in other lanes;

[0069] The set of obstacles Ω filtered by the lane filter obs_lan for:

[0070] {Ω obs_lan |Ω obs_lan ∈Ω, lane obs =lane0}

[0071] Where Ω represents the total set of all obstacles within the sensing range, and lane obs The lane where the obstacle is located; lane0 indicates the lane specified according to user requirements.

[0072] 4. Nearest vehicle filter: It can filter out the first obstacle in front of the vehicle in the lane of interest that is closest to the vehicle and filter out the other obstacles.

[0073] The latest set of obstacles filtered out by the front vehicle filter Ω obs_lea for:

[0074] {Ω obs_lea |Ω obs_lea∈Ωlane obs =lane0,s obs =s upper_bound}

[0075] Where Ω is the total set of all obstacles within the sensing range, s upper_bound Indicates greater than s ego The longitudinal coordinate of the first obstacle.

[0076] Fourth, in step 4, the selection of the filter set in step 3 for screening includes using the filter set in step 3 alone or in combination to screen key obstacles.

[0077] V. In step 5, the label is specifically represented as follows:

[0078]

[0079] The label content can include obstacles obtained by filtering with one or more filters. "longitudinal distance filter" indicates that the obstacle was obtained by filtering with a longitudinal distance filter; "lateral distance filter" indicates that the obstacle was obtained by filtering with a lateral distance filter; "lane filter" indicates that the obstacle was obtained by filtering with a lane filter; "lead obstacle filter" indicates that the obstacle was obtained by filtering with the nearest preceding vehicle filter.

[0080] The labels mentioned above are used to indicate which filter(s) the obstacle was obtained by. If an obstacle was obtained by using multiple filters, the obstacle will be labeled with the labels of all the filters used.

[0081] 6. The key obstacle information obtained in step 4 and the key obstacle label information obtained in step 5, as well as the topology information between the main vehicle and the lane corresponding to the key obstacle, are output to the downstream module, thus completing the screening and output of key obstacles.

[0082] like Figure 3 As shown, the specific implementation process of step 4 above in this embodiment is as follows:

[0083] In the diagram, obstacles (obstacle vehicles) and the main vehicle are represented by rectangles. Numbers 1, 2, 3, 4, 6, 7, 8, 9, and 10 represent all obstacle vehicles within the current sensing range. Vehicle 5 is the main vehicle, and its lane is the current lane, with vehicles 6 ahead, 7 ahead of it, and 4 behind it. The left lane contains vehicles 2 ahead of it on the left, 3 behind it on the left, and 1 ahead of it on the left. The right lane contains vehicles 9 ahead of it on the right, 8 ahead of it on the right, and 10 behind it on the right. There are no obstacles in the left-left and right-right lanes. The scenario depicts the main vehicle having a left lane change intention sent by the upstream intelligent driving system, meaning its target lane is the left lane. In this scenario, the main vehicle no longer needs to focus on all obstacles within the sensing range; it only needs to focus on a few key obstacle vehicles to improve efficiency and reduce computational complexity.

[0084] Traditional key obstacle screening methods only focus on the obstacle closest to the driver in each lane, i.e. Figure 3 The obstacles are numbered 2, 3, 6, 4, 9, and 10. However, this traditional method is too mechanical and lacks flexibility for the following reasons:

[0085] 1. This scenario is a main vehicle changing lanes to the left, so obstacles in the right lane have a relatively small impact. Obstacles numbered 9 and 10 can be disregarded as critical obstacles.

[0086] 2. The impact of the rear vehicle 4 on the lane-changing maneuver of the main vehicle is also relatively small, and it should not be considered a critical obstacle;

[0087] 3. Traditional screening methods do not identify the left-front vehicle 1 as a critical obstacle. However, after the main vehicle intends to change lanes to the left, it can choose to go to the section between obstacles 2 and 3, or it should also be able to choose to go to the section between obstacles 1 and 2. For example, after the main vehicle intends to change lanes to the left, the speed of the left-front vehicle 2 is less than the speed of the main vehicle, that is:

[0088] v2 < v5

[0089] The forced lane change and following of vehicle 2 would cause the main vehicle to brake suddenly, reducing traffic efficiency and comfort. Traditional screening methods, by not considering obstacle 1, block out more possibilities for intelligent driving solutions, compressing the intelligent driving decision-making and planning space and resulting in overly conservative and simplistic decisions.

[0090] The screening method provided in this embodiment, in the case of Figure 3 In this example, obstacle 6 can be filtered out using the nearest preceding vehicle filter in step 3 above, that is:

[0091] s upper_bound =s6

[0092] Where s6 is the longitudinal coordinate of obstacle 6; then, using a combination of lane filter and longitudinal distance filter, obstacles 1, 2, and 3 are filtered to the left lane, where the lane filter specifies the left lane ID, and the longitudinal filter filters obstacles 100 meters in front of and 50 meters behind the main vehicle, that is:

[0093]

[0094] Thus, in the left lane change scenario, the screening method proposed in this embodiment selects obstacles 1, 2, 3, and 6 as key obstacles.

[0095] In such Figure 3 If the main vehicle has a left lane change intention sent by the upstream intelligent driving system, the specific labels of the filtered obstacles are as follows: obstacles 1, 2, and 3 are labeled "longitudinal distance filter" and "lane filter"; obstacle 6 is labeled "lead obstacle filter".

[0096] Based on the disclosed embodiments above, this invention innovatively proposes a method for screening key obstacles in an intelligent driving system. This method can screen out obstacles that have a significant impact on the main vehicle based on the different driving intentions of the vehicle, thereby improving the efficiency of obstacle screening in the intelligent driving system. Furthermore, this invention can handle the screening of obstacles on lane lines, reducing safety hazards caused by incorrect or missed screening of obstacles that cross the lane lines. At the same time, the designed series of filters can screen out obstacles that the main vehicle needs to pay attention to in different scenarios by using a single filter or a combination of different filters, thereby improving the efficiency of obstacle screening in the intelligent driving system and reducing safety hazards caused by incorrect or missed screening of obstacles that cross the lane lines.

[0097] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Any other modifications or equivalent substitutions made by those skilled in the art to the technical solutions of the present invention, as long as they do not depart from the spirit and scope of the technical solutions of the present invention, should be covered within the scope of the claims of the present invention.

Claims

1. A method for screening key obstacles in an intelligent driving system, characterized in that: This method identifies the most significant obstacles affecting the vehicle's movement based on the driving scenario and the driver's intentions. The key obstacles are those closest to the vehicle. The specific steps of this method include the following: Step 1: Obtain environmental information of the intelligent vehicle from the environmental perception layer of the main vehicle's intelligent driving system; Step 2: Based on the environmental information obtained in Step 1, establish the topological relationship between the main vehicle and obstacles and lanes; Step 3: Based on the topology established in Step 2, set up several filters, including longitudinal distance filter, lateral distance filter, lane filter, and nearest preceding vehicle filter; Step 4: Based on the different driving intentions of the main vehicle, select the filter set in Step 3 to filter and obtain the corresponding key obstacle information; Step 5: Label the key obstacles selected in Step 4. The label information includes obtaining the filter information corresponding to the key obstacle. Step 6: Output the key obstacle information obtained in Step 4, the key obstacle label information obtained in Step 5, and the topology information between the main vehicle and the lane corresponding to the key obstacle, thus completing one key obstacle screening and output.

2. The method for screening key obstacles in an intelligent driving system according to claim 1, characterized in that: In step 1, the environmental information of the intelligent vehicle includes information on all obstacles and lane information within the current perception range of the main vehicle, and the environmental information of the intelligent vehicle is stored in the database of the intelligent driving system.

3. The method for screening key obstacles in an intelligent driving system according to claim 1, characterized in that: In step 2, the topological relationship between the main vehicle and the obstacle and the lane includes: the topological relationship between the main vehicle and the obstacle, and the topological relationship between the main vehicle, the obstacle and the lane; The topological relationship between the main vehicle and the obstacle includes the spatial positional relationship and the temporal speed relationship between the obstacle and the main vehicle; the topological relationship between the main vehicle, the obstacle and the lane includes the positional relationship between the main vehicle, the obstacle and the lane boundary, lane line and reference line.

4. The method for screening key obstacles in an intelligent driving system according to claim 3, characterized in that: The spatial relationship between the obstacle and the main vehicle includes obstacles in front of the main vehicle, obstacles behind the main vehicle, obstacles in front of the ... rear of the front of the front of the front of the rear of the 5. The method for screening key obstacles in an intelligent driving system according to claim 3, characterized in that: The topological relationship between lanes includes the current lane of the lane where the main vehicle is located, the left lane to the left of the current lane, the left-left lane, the right lane to the right of the current lane, and the right-right lane.

6. The method for screening key obstacles in an intelligent driving system according to claim 1, characterized in that: The aforementioned longitudinal distance filter can filter the area in front of the main vehicle within its longitudinal range. Obstacles within range and behind Filter out obstacles within the range. Other obstacles outside the designated area; The aforementioned lateral distance filter can filter out items from the left side of the main vehicle's lateral direction. Within and to the right Obstacles within range, and filter them out. Other obstacles outside the designated area; The lane filter described above can filter out all obstacles in a specified lane and filter out obstacles in other lanes. The nearest preceding vehicle filter can identify the first obstacle in front of the vehicle in the lane of interest that is closest to the vehicle, and filter out the remaining obstacles.

7. The method for screening key obstacles in an intelligent driving system according to claim 1, characterized in that: In step 4, the selection of the filter set in step 3 for screening includes using the filter set in step 3 alone or in combination to screen key obstacles.

8. The method for screening key obstacles in an intelligent driving system according to claim 1, characterized in that: In step 5, the label is specifically represented as follows: ； Among them, "longitudinal distance filter" indicates that the obstacle was filtered by the longitudinal distance filter; "lateral distance filter" indicates that the obstacle was filtered by the lateral distance filter; "lane filter" indicates that the obstacle was filtered by the lane filter; and "lead obstacle filter" indicates that the obstacle was filtered by the nearest preceding vehicle filter.

9. The method for screening key obstacles in an intelligent driving system according to claim 8, characterized in that: The labels may include obstacles obtained by filtering through one or more filters.

Images