A driving similarity calculation method and device

Abstract

The present invention discloses a driving similarity calculation method and device, comprising: obtaining reference vehicle data and target vehicle data; obtaining reference vehicle driving information based on the reference vehicle's position information, multi-vehicle position topology information, and single-vehicle motion state information; obtaining target vehicle driving information based on the target vehicle's position information, multi-vehicle position topology information, and single-vehicle motion state information; obtaining position similarity, multi-vehicle position topology similarity, topological attribute similarity, and motion state similarity between the reference vehicle and the target vehicle based on the reference vehicle driving information and the target vehicle driving information; and obtaining, based on the aforementioned similarities, a similarity between the reference vehicle and the target vehicle based on the combined single-vehicle and multi-vehicle position topology information. The technical solution of the present invention can achieve driving similarity calculation that comprehensively considers both the single-vehicle motion information and the multi-vehicle position topology information.

Description

Technical Field

[0001] The present invention belongs to the field of autonomous driving technology, and in particular relates to a driving similarity calculation method and device that comprehensively considers the motion information of a single vehicle and the position topology information of multiple vehicles where the vehicle is located. Background Art

[0002] Autonomous vehicles have a rich history, first appearing in the 1980s and experiencing rapid development over the past four decades. Equipped with advanced sensing systems, intelligent control systems, and precise actuation systems, autonomous vehicles possess intelligent perception, autonomous decision-making, and precise execution capabilities within specific traffic environments, enabling safe, comfortable, efficient, and energy-efficient autonomous driving.

[0003] The rapid development of autonomous driving has led to increasing demands for safety and efficiency. Fast, efficient, and reliable computing models can provide strong support for autonomous driving. With the advancement of transportation capabilities around the world, a large number of natural driving datasets have emerged, providing real-world data support for the development of computational models in the autonomous driving field. Furthermore, the types of data included in these datasets can be easily obtained from the sensor systems onboard autonomous vehicles.

[0004] Obtaining data on vehicles and their behaviors that are identical or similar to the baseline vehicle and its driving behaviors can be used to improve autonomous driving decision-making accuracy and control stability. Furthermore, this data can be applied to vehicle driving risk prediction models to improve driving safety.

[0005] Patent CN113902022A discloses a method for obtaining similarity between autonomous driving scenario data samples. This method, based on cosine similarity, places the two objects to be measured in a new coordinate system and calculates similarity to avoid the problem of collinear but distant objects. However, the cosine similarity metric only characterizes the degree of linear correlation and cannot capture nonlinear characteristics. It also fails to consider key influencing factors of the two measured objects.

[0006] Patent CN110969844A discloses a similarity calculation method based on vehicle driving data. This method uses a trained model to input collected vehicle driving data and existing driving data to obtain the similarity between the two sets of data, which is then used to detect vehicle theft. However, this method only considers the vehicle's own state during driving, and does not consider the impact of the environment on the vehicle's state.

[0007] Currently, existing models for calculating vehicle driving behavior similarity lack comprehensive consideration of both the vehicle's own information and its surroundings, and fail to explore factors closely related to driving behavior, resulting in low confidence in the resulting similarity. Therefore, there is an urgent need to develop a comprehensive driving behavior similarity calculation model that incorporates key factors influencing a vehicle's execution of a specific driving behavior. This can improve the efficiency and accuracy of vehicle decision-making and control modules while also enhancing the safety of autonomous driving. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a driving similarity calculation method and device that comprehensively considers the motion information of a single vehicle and the topological information of multiple vehicles where the vehicle is located.

[0009] To achieve the above purpose, the present invention adopts the following technical solutions

[0010] A driving similarity calculation method comprises the following steps:

[0011] Step S1, obtaining reference vehicle data and target vehicle data;

[0012] Step S2, obtaining reference vehicle driving information based on the reference vehicle position information, multi-vehicle position topology information, and motion state information of a single vehicle in the reference vehicle data;

[0013] Step S3, obtaining target vehicle driving information based on the target vehicle location information, multi-vehicle location topology information, and motion state information of a single vehicle in the target vehicle data;

[0014] Step S4: obtaining the position similarity, multi-vehicle position topology similarity, topology attribute similarity, and motion state similarity of the reference vehicle and the target vehicle based on the reference vehicle driving information and the target vehicle driving information;

[0015] Step S5: Based on the position similarity, the topological structure similarity of the multiple vehicles, the topological attribute similarity and the motion state similarity, the similarity of the position topology information of the reference vehicle and the target vehicle is obtained.

[0016] Preferably, in step S1, based on the driving behavior, vehicle data that performs the behavior is extracted from the natural driving data set as the benchmark vehicle data.

[0017] Preferably, in step S1, based on the driving position and time of the reference vehicle, vehicle data with similar position and time are obtained from the non-reference vehicle data as target vehicle data.

[0018] Preferably, in step S5, the similarity calculation formula of the position topology information of the reference vehicle and the target vehicle combined with the single vehicle and multiple vehicles is:

[0019]

[0020] The present invention also provides a driving similarity calculation device, comprising the following steps:

[0021] An acquisition module, used to acquire reference vehicle data and target vehicle data;

[0022] A first processing module is configured to obtain reference vehicle driving information based on the reference vehicle position information, multi-vehicle position topology information, and motion state information of a single vehicle in the reference vehicle data;

[0023] The second processing module obtains the target vehicle driving information according to the target vehicle position information, the multi-vehicle position topology information, and the motion state information of the single vehicle in the target vehicle data;

[0024] A first calculation module is used to obtain the position similarity, topological structure similarity, topological attribute similarity and motion state similarity of the reference vehicle and the target vehicle based on the reference vehicle driving information and the target vehicle driving information;

[0025] The second calculation module is used to obtain the similarity of the comprehensive single vehicle and multi-vehicle position topology information of the reference vehicle and the target vehicle based on the position similarity, the multi-vehicle position topology structure similarity, the topological attribute similarity and the motion state similarity.

[0026] Preferably, in the acquisition module, vehicle data performing the driving behavior is extracted from the natural driving data set as the benchmark vehicle data based on the driving behavior.

[0027] Preferably, in the acquisition module, based on the driving position and time of the reference vehicle, vehicle data with similar position and time are acquired from the non-reference vehicle data as the target vehicle data.

[0028] Preferably, in the second calculation module, the similarity calculation formula of the reference vehicle and the target vehicle combined with the single vehicle and multiple vehicles position topology information is:

[0029]

[0030] The present invention's driving similarity calculation method, which considers both single-vehicle motion information and the surrounding multi-vehicle topology, comprehensively measures the vehicle's motion state, which is closely related to vehicle behavior, as well as the driving environment information that influences vehicle behavior. This method provides a powerful tool for identifying target vehicles with similar behavior to a benchmark vehicle in the field of autonomous driving. The method boasts clear analytical logic, strong generalization capabilities, and convenient data collection, paving the way for scalable application in multiple fields, including driving decision-making and vehicle safety control. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a flow chart of a method for calculating driving similarity according to an embodiment of the present invention;

[0032] Figure 2 This is a flow chart of a driving similarity calculation method that comprehensively considers the position topology information of a single vehicle and multiple vehicles according to an embodiment of the present invention;

[0033] Figure 3 This is the distribution map of forced lane change locations before highway exit ramps mined from the Next Generation Simulation dataset in the present invention;

[0034] Figure 4 A schematic diagram of a typical topological structure of a moving vehicle in the present invention;

[0035] Figure 5 This is a similarity distribution map between vehicles based on the Pearson correlation coefficient, obtained for data on vehicle pairs that performed forced lane changes at similar locations (within 200 meters of each other longitudinally) before a highway exit ramp and at similar times (within 15 minutes of each other).

[0036] Figure 6 This is a similarity distribution map between vehicles based on cosine similarity, obtained for data on vehicle pairs that performed forced lane changes at similar locations (within 200 meters of each other longitudinally) before a highway exit ramp and at similar times (within 15 minutes of each other).

[0037] Figure 7 This is a similarity distribution map between vehicles based on modified cosine similarity, obtained for pairs of vehicles that performed forced lane changes at similar locations (within 200 meters of each other longitudinally) before a highway exit ramp and at similar times (within 15 minutes of each other).

[0038] Figure 8 This is a similarity distribution map between vehicles based on Jaccard similarity, obtained for data on vehicle pairs that performed forced lane changes at similar locations (within 200 meters of each other longitudinally) before a highway exit ramp and at similar times (within 15 minutes of each other).

[0039] Figure 9 The similarity distribution map obtained by the present invention is used for data of vehicle pairs that perform forced lane changes at similar positions before the highway exit ramp (the longitudinal positions of the two vehicles are 200 meters apart) and at similar times (the two vehicles perform the behavior within 15 minutes). DETAILED DESCRIPTION

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0041] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] The definitions of the formula symbols involved in the embodiment of the present invention are shown in Table 1

[0043] Table 1

[0044]

[0045]

[0046] Example 1:

[0047] like Figure 1 As shown, an embodiment of the present invention provides a driving similarity calculation method, comprising the following steps:

[0048] Step S1, obtaining reference vehicle data and target vehicle data;

[0049] Step S2, obtaining reference vehicle driving information based on the reference vehicle position information, multi-vehicle position topology information, and motion state information of a single vehicle in the reference vehicle data;

[0050] Step S3, obtaining target vehicle driving information based on the target vehicle location information, multi-vehicle location topology information, and motion state information of a single vehicle in the target vehicle data;

[0051] Step S4: obtaining the position similarity, multi-vehicle position topology similarity, topology attribute similarity, and motion state similarity of the reference vehicle and the target vehicle based on the reference vehicle driving information and the target vehicle driving information;

[0052] Step S5: Based on the position similarity, the topological structure similarity of the multiple vehicles, the topological attribute similarity and the motion state similarity, the similarity of the position topology information of the reference vehicle and the target vehicle is obtained.

[0053] As an implementation of an embodiment of the present invention, in step S1, based on the driving behavior Act, vehicle data that performs the behavior is extracted from a natural driving dataset as benchmark vehicle data.

[0054] As an implementation of an embodiment of the present invention, in step S1, based on the driving position and time of the reference vehicle, vehicle data with similar positions and times are obtained from non-reference vehicle data as target vehicle data.

[0055] As an implementation method of an embodiment of the present invention, in step S5, the similarity calculation formula of the reference vehicle and the target vehicle's comprehensive single vehicle and multi-vehicle position topology information is:

[0056] Example 2:

[0057] like Figure 2 As shown, an embodiment of the present invention provides a driving similarity calculation method that comprehensively considers the motion information of a single vehicle and the topological information of multiple vehicles where the vehicle is located, including the following steps:

[0058] Step 1: Through the analysis of the natural driving dataset and the driving behavior of the vehicle, it can be seen that the driving environment, location topology information and the vehicle's own driving state information are directly related to the vehicle's driving behavior. Among them, the driving environment factors include the location of the vehicle when it is driving, such as different locations such as cities, rural areas, near highway entrances, near highway exits, and natural weather such as rain, snow, and fog; the location topology information is the topology composed of the target vehicle and the surrounding traffic vehicles closely related to the vehicle's location; the vehicle's own motion state includes speed and acceleration. Regarding the vehicle's driving environment information, the natural driving dataset Next Generation Simulation (NGSim, this data is a public dataset, and the collected road sections include US-101 and Lankershim Avenue in southbound Los Angeles, California, I-80 in eastbound Emeryville, California, and Peachtree Street in Atlanta, Georgia, https: / / catalog.data.gov / dataset / next-generation-simulation-ngsim-vehicle-trajectories-and-supporting-data), taking the forced lane change near the highway ramp exit as an example, such as Figure 3 As shown, it shows obvious location clustering. Regarding the topology of the vehicle's location, Figure 4 Represents the typical location topology related to the target vehicle's driving decision.

[0059] Step 2: Given that the most readily available environmental information in current traffic big data is location, which is directly related to driving behavior, the location information of vehicle vi is obtained as environmental information, which is expressed as loc vi .

[0060] Step 3: Extract the motion information of vehicle vi, expressed as:

[0061]

[0062] Step 4: V vi (G), E vi (G) and Ψ g，vi Represent the vertex set, edge set, and mapping relationship set of the topology where vehicle vi is located, and construct a graph model of vehicle vi at time t:

[0063] G B =(V vi (G), E vi (G),Ψ g，vi )

[0064] in,

[0065] V vi (G)={SV, FV, LVF, LVB, RVF, RVB}

[0066] E vi (G)={e1, e2, e3, e4, e5, e6}

[0067] Ψ g，vi (e1)=SV FV

[0068] Ψ g，vi (e2)=SV LVF

[0069] Ψ g，vi (e3)=SV LVB

[0070] Ψ g，vi (e4)=SV RVF

[0071] Ψ g，vi (e5)=SV RVB

[0072] Ψ g，vi (e6) = RVB RVF

[0073]

[0074] Among them, SV stands for Figure 2 The target vehicle in the discussion is the vehicle vi in ​​this paragraph, FV represents the front vehicle, LVF represents the left front vehicle, LVB represents the left rear vehicle, RVF represents the right front vehicle, and RVB represents the right rear vehicle.

[0075] Step 5: The traffic vehicle presence sign vector, the relative position vector, relative velocity vector, and relative acceleration vector of the topology where vehicle vi is located are respectively represented, and the complete topology information model of the vehicle is constructed:

[0076]

[0077] in,

[0078]

[0079] And define f i Value:

[0080]

[0081]

[0082] ∈((FV,vi),(vi,LVF),(vi,LVB),(vi,RVF),(vi,RVB),(RVB,RVF)})

[0083] Δloc vj-vk represents the relative position between vehicle vj and vehicle vk. For example, when (vj, vk) = (vi, LVF), Δloc vj-vk It is expressed as the relative position between vi and the vehicle in front of vi in ​​the left lane (i.e., LVF). The specific expression is as follows:

[0084] Δloc vj-vk =loc vj -loc vk

[0085]

[0086] ∈{(FV,vi), (vi,LVF), (vi,LVB), (vi,RVF), (vi,RVB), (RVB,RVF)})

[0087] Δvel vj-vk represents the relative speed between vehicle vj and vehicle vk. For example, when (vj, vk) = (vi, LVF), Δvel vj-vk It is expressed as the relative speed between vi and the vehicle in front of vi in ​​the left lane (i.e., LVF). The specific expression is as follows:

[0088] Δvel vj-vk =vel vj -vel vk

[0089]

[0090] ∈((FV,vi),(vi,LVF),(vi,LVB),(vi,RVF),(vi,RVB),(RVB,RVF)})

[0091] With Δacc vj-vk represents the relative acceleration between vehicle vj and vehicle vk. For example, when (vj, vk) = (vi, LVF), Δacc vj-vk It is expressed as the relative acceleration between vi and the vehicle in front of vi in ​​the left lane (i.e., LVF). The specific expression is as follows:

[0092] Δacc vj-v k=acc vj -acc vk

[0093] Step 6: Construct the driving information vector of vehicle vi:

[0094]

[0095] Step 7: Extract the benchmark vehicle from the natural driving dataset and construct its driving information:

[0096]

[0097] Step 8: Take the non-reference vehicles that are located at a similar location to the reference vehicle in a similar time period as the target vehicle set, and construct driving information about the target vehicles:

[0098]

[0099] Step 9: Based on the key factors in step 1, a driving similarity model between the target vehicle and the benchmark vehicle is constructed as follows:

[0100] sim=sim env ·sim GA

[0101] Among them, sim GA =sim topology ·sim motion ,sim topology =sim top_struc ·sim top_val ,sim env 、sim topology With sim motion They represent the similarity of the environment between the target vehicle and the benchmark vehicle, the similarity of the multi-vehicle topology between the target vehicle and the benchmark vehicle, and the similarity of the motion state between the target vehicle and the benchmark vehicle, respectively. top_struc With sim top_val They represent topological structure similarity and topological attribute similarity respectively.

[0102] Step 10: Based on the positions (loc) of the two vehicles (v1, v2) v1 ,loc v2 ), the environment similarity model is defined as follows:

[0103]

[0104] Here, σ represents the degree of centralization of the model’s definition of “closely located”.

[0105] Step 11: Based on the Jaccard similarity metric, define the modified Jaccard similarity metric and the topological similarity of the two vehicles (v1, v2) as follows:

[0106]

[0107]

[0108] in,

[0109]

[0110]

[0111] Among them, f v1,i With f v2,i The definition of f is the same as that of step 5 i , respectively indicating whether there is a traffic vehicle at position i in the multi-vehicle position topology where vehicles v1 and v2 are located.

[0112] Step 12: Based on the Pearson correlation coefficient, define the attribute similarity sim of the topology of the two vehicles (v1, v2) top_val and motion similarity sim motion :

[0113]

[0114]

[0115] Step 13: Combining Steps 9 to 12, the complete expression of the driving similarity calculation model of the two vehicles is as follows:

[0116]

[0117] from Figure 5-Figure 8 As can be seen, when only the basic similarity measurement method is used without any constraints, the distribution of driving similarities of the same type of driving behaviors in the dataset is very broad, without reflecting the clustering, and cannot be used to reflect the similar characteristics of the same driving behavior.

[0118] from Figure 9 As can be seen from the figure, the driving similarity distribution of the same type of driving behaviors in the dataset has obvious clustering, which can help effectively locate vehicles with highly similar driving behaviors.

[0119] Example 3:

[0120] An embodiment of the present invention further provides a driving similarity calculation device, comprising the following steps:

[0121] An acquisition module, used to acquire reference vehicle data and target vehicle data;

[0122] A first processing module is configured to obtain reference vehicle driving information based on the reference vehicle position information, multi-vehicle position topology information, and motion state information of a single vehicle in the reference vehicle data;

[0123] The second processing module obtains the target vehicle driving information according to the target vehicle position information, the multi-vehicle position topology information, and the motion state information of the single vehicle in the target vehicle data;

[0124] A first calculation module is used to obtain the position similarity, topological structure similarity, topological attribute similarity and motion state similarity of the reference vehicle and the target vehicle based on the reference vehicle driving information and the target vehicle driving information;

[0125] The second calculation module is used to obtain the similarity of the comprehensive single vehicle and multi-vehicle position topology information of the reference vehicle and the target vehicle based on the position similarity, the multi-vehicle position topology structure similarity, the topological attribute similarity and the motion state similarity.

[0126] As an implementation manner of the embodiment of the present invention, in the acquisition module, vehicle data performing the driving behavior is extracted from the natural driving data set according to the driving behavior as the benchmark vehicle data.

[0127] As an implementation of an embodiment of the present invention, in the acquisition module, based on the driving position and time of the reference vehicle, vehicle data with similar position and time are acquired from the non-reference vehicle data as target vehicle data.

[0128] As an implementation manner of an embodiment of the present invention, in the second calculation module, the similarity calculation formula of the reference vehicle and the target vehicle comprehensive single vehicle and multi-vehicle position topology information is:

[0129]

[0130] The above is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or replacements that can be easily thought of by any technician familiar with this technical field within the technical scope described in the present invention should be covered by the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A driving similarity calculation method, characterized in that: The following steps are involved: Step S1, obtaining reference vehicle data and target vehicle data; Step S2, obtaining reference vehicle driving information based on the reference vehicle position information, multi-vehicle position topology information, and motion state information of a single vehicle in the reference vehicle data; Step S3, obtaining target vehicle driving information based on the target vehicle location information, multi-vehicle location topology information, and motion state information of a single vehicle in the target vehicle data; Step S4: obtaining the position similarity, multi-vehicle position topology similarity, topology attribute similarity, and motion state similarity of the reference vehicle and the target vehicle based on the reference vehicle driving information and the target vehicle driving information; Step S5: obtaining the similarity of the position topology information of the reference vehicle and the target vehicle in the integrated single vehicle and multi-vehicle position topology information based on the position similarity, the multi-vehicle position topology similarity, the topology attribute similarity, and the motion state similarity; The driving similarity model between the target vehicle and the benchmark vehicle is constructed as follows: yes=yes env ·Yes GA Among them, sim GA =sim topology ·sim motion ,sim topology =sim top_struc ·sim top_val ,sim env 、sim topology With sim motion They represent the similarity of the environment between the target vehicle and the benchmark vehicle, the similarity of the multi-vehicle topology between the target vehicle and the benchmark vehicle, and the similarity of the motion state between the target vehicle and the benchmark vehicle, respectively. top_struc With sim top_val Represent the topological structure similarity and topological attribute similarity respectively; Based on the positions (loc v1 ,loc v2 ), the environment similarity model is defined as follows: Among them, σ represents the degree of concentration of the model's definition of "close location"; Based on the Jaccard similarity metric, the modified Jaccard similarity metric and the topological similarity of the two vehicles (v1, v2) are defined as follows: in, Among them, f v1,i With f v2,i Respectively indicate whether there is a traffic vehicle at position i in the multi-vehicle position topology where vehicles v1 and v2 are located; Based on the Pearson correlation coefficient, the attribute similarity sim of the topology of the two vehicles (v1, v2) is defined respectively top_val and motion similarity sim motion :

2. The driving similarity calculation method according to claim 1, wherein: In step S1 , based on the driving behavior, vehicle data that performs the behavior is extracted from the natural driving data set as the benchmark vehicle data.

3. The driving similarity calculation method according to claim 2, wherein: In step S1 , based on the driving position and time of the reference vehicle, vehicle data with similar positions and times are obtained from the non-reference vehicle data as target vehicle data.

4. The driving similarity calculation method according to claim 3, wherein: In step S5, the similarity calculation formula of the reference vehicle and the target vehicle is as follows:

5. A driving similarity calculation device for implementing the driving similarity calculation method according to claim 1, characterized in that: The following steps are involved: An acquisition module, used to acquire reference vehicle data and target vehicle data; A first processing module is configured to obtain reference vehicle driving information based on the reference vehicle position information, multi-vehicle position topology information, and motion state information of a single vehicle in the reference vehicle data; The second processing module obtains the target vehicle driving information according to the target vehicle position information, the multi-vehicle position topology information, and the motion state information of the single vehicle in the target vehicle data; A first calculation module is used to obtain the position similarity, topological structure similarity, topological attribute similarity and motion state similarity of the reference vehicle and the target vehicle based on the reference vehicle driving information and the target vehicle driving information; The second calculation module is used to obtain the similarity of the comprehensive single vehicle and multi-vehicle position topology information of the reference vehicle and the target vehicle based on the position similarity, the multi-vehicle position topology structure similarity, the topological attribute similarity and the motion state similarity.

6. The driving similarity calculation device according to claim 5, wherein: In the acquisition module, vehicle data performing the driving behavior is extracted from the natural driving data set according to the driving behavior as the benchmark vehicle data.

7. The driving similarity calculation device according to claim 6, wherein: In the acquisition module, based on the driving position and time of the reference vehicle, vehicle data with similar positions and times are acquired from the non-reference vehicle data as target vehicle data.

8. The driving similarity calculation device according to claim 7, wherein: In the second calculation module, the similarity calculation formula of the reference vehicle and the target vehicle combined with the single vehicle and multiple vehicle position topology information is:

Images