A high-risk scene identification method and device based on vehicle-vehicle cooperation and a medium

Abstract

The application relates to a high-risk scene recognition method and device based on vehicle-vehicle cooperation and a medium; the method comprises the following steps: mapping collected environment information to a BEV perspective; using a backbone network to encode environment information collected by a sensor to obtain intermediate feature information; encrypting and compressing self-perception information; receiving other intelligent networked vehicle perception information through a V2X sensor; spatially aligning the obtained perception information, and converting the feature information to a self-vehicle coordinate system; analyzing the discrete degree of the perception features in the same area; fusing multi-source perception information of the self-vehicle to obtain panoramic perception result output and input the panoramic perception result output to a decision planning module; and analyzing the uncertainty of the cooperative perception result of the self-vehicle; the application recognizes a high-risk scene in driving from a new dimension by comprehensively analyzing the perception understanding ability of an autonomous vehicle and the complexity of a scene, reminds a driver to intervene or switch a driving strategy, and improves the driving safety of the autonomous vehicle.

Description

Technical Field

[0001] This invention belongs to the field of autonomous driving technology and relates to a method, device and medium for identifying high-risk scenarios based on vehicle-to-vehicle cooperation. Background Technology

[0002] Artificial intelligence technology is driving the automotive industry towards rapid intelligent development, bringing profound changes to the production and lifestyles of modern society. Safe driving in any scenario is a fundamental requirement for autonomous vehicles, which are influenced by complex factors such as the human-vehicle-road relationship. Real-world driving scenarios are highly dynamic and partially observable, with complex interaction topologies among traffic participants, posing significant challenges to the scenario understanding of autonomous vehicles. The degree of risk in a driving scenario is one of the important factors affecting the safe operation of autonomous vehicles.

[0003] Scene perception technology for autonomous vehicles is the channel through which intelligent cars perceive the external world and is one of the core technologies in autonomous driving. In recent years, with the continuous development of high-performance sensors and artificial intelligence technologies, the ability of autonomous vehicles to collect and understand scenes has been greatly improved. However, when vehicles encounter scenes with many obstructions or long distances, the effect is still significantly insufficient. To address the limitations of single-vehicle perception, researchers have introduced V2X communication technology into the perception module of autonomous driving, and perception technology is gradually moving from traditional "single-vehicle perception" to collaborative perception based on "vehicle-road-cloud". By fusing multi-source, multi-modal, and multi-view scene features from multiple intelligent connected vehicles and roadside equipment in a scene, richer panoramic perception results are obtained, greatly improving the scene understanding ability of autonomous vehicles. Deep learning algorithms are the mainstream algorithms, enabling autonomous driving to make significant progress in perception accuracy, but their "black box" model's poor interpretability poses a significant threat to driving safety. Therefore, we should clearly define the capability boundaries of the model, identify high-risk scenarios that the model cannot handle, and avoid generating "overconfident" erroneous results that could lead to serious traffic accidents.

[0004] In summary, assessing the risk level of a scenario depends not only on the complexity of the scenario itself but also on the algorithmic capabilities of the autonomous vehicle. If the perception system exhibits high uncertainty in its understanding of the scenario, it indicates that the model's ability to understand the scenario is limited, and the scenario can be considered high-risk. By leveraging collaborative perception technology and analyzing the dispersion of scenario understanding from different perception perspectives by multiple sensing devices and the uncertainty of collaborative perception results, a novel perspective for assessing the risk level of a scenario can be provided, demonstrating strong theoretical and technical feasibility. Summary of the Invention

[0005] This invention incorporates the environmental perception capabilities of intelligent connected vehicles into a scenario risk assessment system, measuring the risk level of a scenario through the uncertainty of the intelligent connected vehicle's perception of the scenario. Specifically, this invention proposes a high-risk scenario identification method, device, and medium based on vehicle-to-vehicle collaboration. The vehicle receives perception information from other intelligent connected vehicles in the environment via V2X devices, analyzes the dispersion of perception results for the same area and the uncertainty of panoramic perception results, constructs the capability boundary of the perception model, thereby achieving effective identification of high-risk scenarios and providing assurance for the safe driving of intelligent connected vehicles.

[0006] To solve the above-mentioned technical problems, the present invention is implemented using the following technical solution, which is described below in conjunction with the accompanying drawings:

[0007] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0008] A method for identifying high-risk scenarios based on vehicle-to-vehicle cooperation includes the following steps:

[0009] Step 1: In this scenario, intelligent connected vehicles use their own sensors to collect information about the surrounding environment.

[0010] Step 2: Each vehicle maps the collected environmental information onto the BEV perspective.

[0011] Step 3: Each intelligent connected vehicle uses the backbone network to encode the environmental information collected by the sensors to obtain intermediate feature information of the surrounding scene.

[0012] Step 4: Each intelligent connected vehicle encrypts and compresses its own perceived information and broadcasts it to other intelligent connected vehicles in the scene via V2X devices.

[0013] Step 5: The vehicle receives perception information broadcast by intelligent connected vehicles in the environment via V2X sensors, and then decodes and decompresses it.

[0014] Step Six: The vehicle spatially aligns the acquired perception information and transforms the feature information into the vehicle's coordinate system.

[0015] Step 7: For the overlapping areas of multi-source sensing data, analyze the dispersion of sensing features in the same area. If it is greater than the threshold, then the current driving scenario is judged to be a high-risk scenario.

[0016] Step 8: The vehicle integrates multi-source perception information to obtain a panoramic perception result output, which is then output to the decision planning module.

[0017] Step 9: Analyze the uncertainty of the vehicle's collaborative perception results. If it exceeds the threshold, the current driving scenario is considered a high-risk scenario.

[0018] Furthermore, in step one, the intelligent connected vehicle uses its own sensors to collect information about the surrounding environment. The specific details are as follows:

[0019] Intelligent connected vehicles V exist in the environment i (i∈[0,1,...,n]), each vehicle is equipped with the same perception sensors and a collaborative perception model, which can collect raw sensor data of the surrounding driving environment at a certain time frequency, denoted as...

[0020] X i i∈[0,1,…,]

[0021] Where V0 represents the vehicle, X0 represents the vehicle's perception information, and V i (i∈[1,2,...,n]) represents intelligent connected vehicles other than the vehicle itself in the scenario, X i (i∈[1,2,...,n]) represents the perception information of intelligent connected vehicles other than the vehicle itself in the scene.

[0022] Meanwhile, in the scenario, intelligent connected vehicles V i (i∈[0,1,...,n]) are all equipped with V2X communication devices, which can realize information sharing between vehicles.

[0023] V2X devices refer to Vehicle-to-Everything devices, which are devices used by intelligent connected vehicles to communicate with the outside world.

[0024] Furthermore, in step two, each vehicle maps the collected environmental information to the BEV perspective, as detailed below:

[0025] Each vehicle will collect surrounding environmental information X from its sensors. i Transforming from the sensor coordinate system and mapping to the BEV viewpoint using a transformation function, it can be expressed as:

[0026]

[0027] in, This represents the vehicle Vi's perception of its surroundings from a BEV perspective, fbev (·) represents the BEV perspective mapping model.

[0028] BEV stands for Bird's-eye-view, referring to a bird's-eye view perspective.

[0029] Furthermore, in step three, each intelligent connected vehicle uses the backbone network to encode the environmental information collected by the sensors to obtain intermediate feature information of the surrounding scene, as detailed below:

[0030] Each intelligent connected vehicle uses the backbone network to process the information collected by the sensors. Encode the surrounding scene to obtain intermediate features F. i F i It is a high-dimensional tensor, which can be represented as

[0031]

[0032] Among them, F i f represents the feature information of the scene surrounding vehicle Vi. enc (·) represents the backbone network model. The f between different vehicles enc (·) Shared parameters.

[0033] Furthermore, in step four, each intelligent connected vehicle encrypts and compresses its own perceived information and broadcasts it to other intelligent connected vehicles in the scene via V2X devices. The specific content is as follows:

[0034] Each intelligent connected vehicle obtains perception information D from vehicle-side computation. i The information is encrypted and compressed, and then broadcast to the scene via V2X devices. The perceived information D... i Including feature information F of the surrounding scene i Pose information P i and time information T i .

[0035] D i ={F i P i T i}i∈{0,1,...,n}

[0036] in,

[0037] P i ={x i y i , z i lon i lat i alt i ,heading i}i∈{0,1,...,n}

[0038] Among them, (x i y i , z i For intelligent connected vehicles V i Global coordinates in the geodetic coordinate system, (lon) i lat i alt i ) are respectively intelligent connected vehicles V i Longitude, latitude, and altitude, heading i For intelligent connected vehicles V i The heading angle.

[0039] Furthermore, in step six, the vehicle spatially aligns the acquired perception information, transforming the feature information into the vehicle's coordinate system. The specific details are as follows:

[0040] The vehicle acquires perception information of intelligent connected vehicles in the environment through V2X devices. i Then, based on the pose information of the two vehicles, a coordinate transformation model is constructed to transform the acquired D... i Environmental characteristics F i (i∈[1,2,...,n]) undergo coordinate transformation to its own coordinate system, and "spatial position-feature information" mapping is added to obtain the intelligent connected vehicle V in the vehicle's coordinate system. i Feature information

[0041]

[0042] Among them, Γ i→j (·) indicates intelligent connected vehicle V i Coordinate system to intelligent connected vehicle V j Coordinate system transformation model, Γ i→0 (·) indicates intelligent connected vehicle V i The transformation model from the coordinate system to the vehicle. This indicates the location information within the scene.

[0043] Furthermore, as described in step seven, for the overlapping areas of multi-source sensing data, the dispersion of sensing features in the same area is analyzed. If it exceeds a threshold, the current driving scenario is considered a high-risk scenario. The specific details are as follows:

[0044] For the overlapping perception area of ​​multiple vehicles in the scenario, the dispersion of multi-source data is analyzed. If the dispersion index of the overlapping perception area of ​​multiple vehicles is greater than the threshold δ1, it indicates that the vehicle perception model has a large uncertainty in the perception of the current driving scenario, or even does not have the ability to cope with the current driving scenario. The current driving scenario should be judged as a high-risk area, and the vehicle should issue a takeover warning to the driver or adjust the decision planning strategy.

[0045] Assume that the received environmental features F of each intelligent connected vehicle i→0 Follows probability distribution p i (x) For the overlapping region R, the vehicle analysis examines the dispersion of feature information in the perceived overlapping region. The relative entropy (KL divergence) measures the dispersion of the vehicle's features from the perceived features of other vehicles in the environment.

[0046]

[0047] in, This represents the expectation of the distribution.

[0048] The discrete index of the entire scene is defined as

[0049]

[0050] If the dispersion index D > δ1, then the current scenario is judged as a high-risk scenario.

[0051] Furthermore, as described in step eight, the vehicle integrates multi-source perception information to obtain a panoramic perception result output, which is then output to the decision-making and planning module. The specific details are as follows:

[0052] The vehicle integrates multi-source perception features and performs feature extraction and scene understanding through a collaborative perception information fusion network and a detection head to obtain a panoramic perception result of the scene.

[0053] Y = head(f fusion (cconcat(F 1→0 (x),F 2→0 (x),...,F n→0 (x)))

[0054] Where Y represents the current perception result, head(·) is the detection head model, and f fusion (·) represents the multi-source information fusion model; concat(·) is the feature concatenation function, F i→0 (x) represents the intelligent connected vehicle V i Features in the vehicle's coordinate system.

[0055] Regarding the perceived result Y

[0056] Y = {obs} j}j∈[0,n o ]

[0057]

[0058] Where, n o To perceive the number of obstacles in the output, (x j y j , zj ) is an obstacle (obs) j Location information, bbox j For obstacles (obs) j The boundingbox parameter, c j For obstacles (obs) j type, , where is the confidence level.

[0059] Furthermore, in step nine, the uncertainty of the vehicle's collaborative perception results is analyzed. If the uncertainty exceeds a threshold, the current driving scenario is determined to be a high-risk scenario. The specific details are as follows:

[0060] Based on the confidence level of the perception results, the vehicle measures the degree of uncertainty by calculating the information entropy of each traffic participant (obstacle) in the perception results.

[0061]

[0062] Among them, H j This indicates the detection of obstacles (obs). j The information entropy. For the entire scene, calculate the degree of uncertainty of the perception results to obtain the uncertainty index of the entire scene.

[0063]

[0064] Here, H represents the uncertainty index of the entire scenario. If H > δ2, the current driving scenario is considered a high-risk scenario.

[0065] A high-risk scene identification device based on vehicle-to-vehicle cooperation includes:

[0066] The information acquisition module is used by intelligent connected vehicles in various scenarios to collect information about their surrounding environment using their own sensors.

[0067] The mapping module is used by each vehicle to map the collected environmental information to the BEV perspective;

[0068] The encoding module is used by each intelligent connected vehicle to encode the environmental information collected by the sensors using the backbone network to obtain intermediate feature information of the surrounding scene.

[0069] The encryption and compression module allows each intelligent connected vehicle to encrypt and compress its own sensed information and broadcast it to other intelligent connected vehicles in the scene via V2X devices.

[0070] The decompression module is used to receive and decompress information from other intelligent connected vehicles in the environment via V2X sensors.

[0071] The conversion module is used to spatially align the perception information acquired by the vehicle and convert the feature information to the vehicle's coordinate system.

[0072] The first analysis module is used to analyze the dispersion of the sensing features in the overlapping areas of multi-source sensing data. If it is greater than the threshold, the current driving scenario is considered a high-risk scenario.

[0073] The result output module is used to fuse multi-source perception information of the vehicle to obtain panoramic perception results and output them to the decision planning module.

[0074] The second analysis module is used to analyze the uncertainty of the vehicle's collaborative perception results. If the uncertainty exceeds the threshold, the current driving scenario is considered a high-risk scenario.

[0075] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described above.

[0076] Compared with the prior art, the beneficial effects of the present invention are:

[0077] 1. This invention proposes a high-risk scenario identification method based on vehicle-to-vehicle collaboration, which incorporates the self-perception capabilities of intelligent vehicles into the scenario risk evaluation system. By comprehensively analyzing the perception and understanding capabilities of autonomous vehicles and the complexity of scenarios, high-risk scenarios in driving are identified from a new dimension to remind the driver to intervene or switch driving strategies, thereby improving the driving safety of autonomous vehicles.

[0078] 2. This invention proposes a novel uncertainty calculation method using collaborative perception technology, enabling autonomous vehicles to receive perception information from multiple sensing devices in a scene, analyze the degree of dispersion of scene perception from multiple sources and the uncertainty of panoramic perception output results, and quantitatively analyze the autonomous vehicle's ability to understand the scene.

[0079] 3. Based on a collaborative perception framework, this invention enables vehicles to acquire and fuse multi-source, multi-modal, and multi-view information from the environment, overcoming the occlusion and long-distance perception problems inherent in single-vehicle perception. This improves the accuracy of perception and scene understanding, providing data support for driving safety. Attached Figure Description

[0080] The invention will now be further described with reference to the accompanying drawings:

[0081] Figure 1 This is a flowchart of a high-risk scenario identification method based on vehicle-to-vehicle collaboration as described in this invention. Detailed Implementation

[0082] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some, but not all, embodiments of this invention. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this invention, and should not be construed as limiting the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. The embodiments of this invention will be described in detail below with reference to the accompanying drawings.

[0083] In the description of this invention, it should be understood that the terms "center", "longitudinal", "lateral", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting the scope of protection of this invention.

[0084] The present invention will now be described in detail with reference to the accompanying drawings:

[0085] See appendix Figure 1 The present invention proposes a high-risk scenario identification method based on vehicle-to-vehicle cooperation, which includes the following explanatory process:

[0086] like Figure 1 As shown, this invention proposes a high-risk scenario identification method based on vehicle-to-vehicle collaboration. It incorporates the environmental perception capabilities of intelligent connected vehicles into the scenario risk assessment system, providing a new perspective for measuring the identification of high-risk scenarios for autonomous vehicles. This invention measures the risk level of a scenario by assessing the uncertainty of the intelligent connected vehicle's perception of the scenario. It utilizes V2X devices to receive perception information from other intelligent connected vehicles in the environment, and fuses multi-source and multi-view perception information through a collaborative perception model. It analyzes the dispersion of perception results for the same area and the uncertainty of panoramic perception results. If the dispersion index or uncertainty index exceeds a threshold, the perception system is considered unable to accurately understand the scenario, thus achieving effective identification of high-risk scenarios and ensuring the safe driving of autonomous vehicles.

[0087] Specifically, it includes:

[0088] Step 1: In this scenario, intelligent connected vehicles use their own sensors to collect information about the surrounding environment.

[0089] Intelligent connected vehicles V exist in the environment i (∈[0,1,...,n]), each vehicle is equipped with the same sensing sensor, which can collect raw sensor data of the surrounding driving environment at a certain time frequency, represented as...

[0090] X i i∈[0,1,…,n]

[0091] Where V0 represents the vehicle, X0 represents the vehicle's perception information, and V i (i∈[1,2,...,n]) represents intelligent connected vehicles other than the vehicle itself in the scenario, X i (i∈[1,2,...,n]) represents the perception information of intelligent connected vehicles other than the vehicle itself in the scene. X i The information format and dimensional information are related to the sensor type and model.

[0092] Meanwhile, in the scenario, intelligent connected vehicles V i (i∈[0,1,...,n]) are all equipped with V2X communication devices, which can adapt to the transmission requirements of the information format in this invention and realize information sharing between vehicles.

[0093] Step 2: Each vehicle maps the collected environmental information to the BEV perspective:

[0094] Each vehicle will collect surrounding environmental information X from its sensors. i Transforming from the sensor coordinate system and mapping it to the BEV perspective using a transformation function, it can be expressed as:

[0095]

[0096] in, This represents the vehicle Vi's perception of its surroundings from a BEV perspective, f bev (•) represents the BEV perspective mapping model.

[0097] Step 3: Each intelligent connected vehicle uses the backbone network to encode the environmental information collected by the sensors, obtaining intermediate feature information of the surrounding scene:

[0098] Each intelligent connected vehicle uses the backbone network to process the information collected by the sensors. Encode the surrounding scene to obtain intermediate features F. i F i It is a high-dimensional tensor, which can be represented as

[0099]

[0100] Among them, F i Indicates vehicle V iFeature information of the surrounding scene, f enc (·) represents the backbone network model. The f between different vehicles enc (·) Shared parameters.

[0101] Step 4: Each intelligent connected vehicle encrypts and compresses its own perceived information and broadcasts it to other intelligent connected vehicles in the scene via V2X devices:

[0102] Each intelligent connected vehicle obtains perception information D from vehicle-side computation. i The information is encrypted and compressed, and then broadcast to the scene via V2X devices. The perceived information D... i Including feature information F of the surrounding scene i Pose information P i and time information T i .

[0103] D i ={F i P i T i}i∈{0,1,...,n}

[0104] The pose information includes,

[0105] P i ={x i y i , z i lon i lat i alt i ,heading i}i∈{0,1,...,n}

[0106] Among them, (x i y i , z i (lon) represents the global coordinates of the intelligent connected vehicle Vi in the geodetic coordinate system. i lat i alt i ) are respectively intelligent connected vehicles V i Longitude, latitude, and altitude, heading i For intelligent connected vehicles V i The heading angle.

[0107] Step 5: The vehicle receives perception information from other connected vehicles via V2X sensors and decompresses it.

[0108] Step Six: The vehicle spatially aligns the acquired perception information, transforming the feature information into the vehicle's coordinate system:

[0109] The vehicle obtains information about intelligent connected vehicles in the environment via V2X devices. i Perceptual information D of (i∈[1,2,...,n]) i Then, based on the pose information of the two vehicles, a coordinate transformation model is constructed to transform the acquired D... i Environmental characteristics F i (i∈[1,2,...,n]) undergo coordinate transformation to its own coordinate system, and "spatial position-feature information" mapping is added to obtain the intelligent connected vehicle V in the vehicle's own coordinate system. i Feature information

[0110]

[0111] Among them, Γ i→j (·) indicates the coordinate system from the Vi coordinate system of the intelligent connected vehicle to the V coordinate system of the intelligent connected vehicle. j Coordinate system transformation model, Γ i→0 (·) represents the transformation model from the i-coordinate system of the intelligent connected vehicle to the vehicle itself. This indicates the location information within the scene.

[0112] Step 7: For overlapping areas of multi-source sensing data, analyze the dispersion of sensing features in the same area. If it exceeds a threshold, the current driving scenario is considered a high-risk scenario.

[0113] For the overlapping perception area of ​​multiple vehicles in the scenario, the dispersion of multi-source data is analyzed. If the dispersion index of the overlapping perception area of ​​multiple vehicles is greater than the threshold δ1, it indicates that the vehicle perception model has a large uncertainty in the perception of the current driving scenario, or even does not have the ability to cope with the current driving scenario. The current driving scenario should be judged as a high-risk area, and the vehicle should issue a takeover warning to the driver or adjust the decision planning strategy.

[0114] Assume that the received environmental features F of each intelligent connected vehicle i→0 Follows probability distribution p i (x) For the overlapping region R, the vehicle analysis examines the dispersion of feature information in the perceived overlapping region. The relative entropy (KL divergence) measures the dispersion of the vehicle's features from the perceived features of other vehicles in the environment.

[0115]

[0116] in, This represents the expectation of the distribution.

[0117] The discrete index of the entire scene is defined as

[0118]

[0119] If the dispersion index D > δ1, then the current scenario is judged as a high-risk scenario.

[0120] Step 8: The vehicle integrates multi-source perception information to obtain a panoramic perception result, which is then output to the decision-making and planning module.

[0121] The vehicle integrates multi-source perception features and performs feature extraction and scene understanding through a collaborative perception information fusion network and a detection head to obtain a panoramic perception result of the scene.

[0122] Y = head(f fusion (concat(F 1→0 (x), F 2→0 (x), ..., F n→0 (x)))

[0123] Where Y represents the current perception result, head(·) is the detection head model, and f fusion (·) represents the multi-source information fusion model; concat(·) is the feature concatenation function, F i→0 (x) represents the intelligent connected vehicle V i Features in the vehicle's V0 coordinate system.

[0124] Regarding the perceived result Y

[0125] Y = {obs} j}j∈[0,n o ]

[0126]

[0127] Where, n o To perceive the number of obstacles in the output, (x j y j , z j ) is an obstacle (obs) j Location information, bbox j For obstacles (obs) j The boundingbox parameter, c j For obstacles (obs) j type, , where is the confidence level.

[0128] Step Nine: Analyze the uncertainty of the vehicle's collaborative perception results. If the uncertainty exceeds the threshold, the current driving scenario is considered a high-risk scenario.

[0129] Based on the confidence level of the perception results, the vehicle measures the degree of uncertainty by calculating the information entropy of each obstacle in the perception results.

[0130]

[0131] Among them, H j This indicates the detection of obstacles (obs). j The information entropy. For the entire scene, calculate the degree of uncertainty of the perception results to obtain the uncertainty index of the entire scene.

[0132]

[0133] Here, H represents the uncertainty index of the entire scenario. If H > δ2, the current driving scenario is considered a high-risk scenario.

[0134] The invention provides a high-risk scene identification device based on vehicle-to-vehicle cooperation, comprising:

[0135] The information acquisition module is used by intelligent connected vehicles in various scenarios to collect information about their surrounding environment using their own sensors.

[0136] The mapping module is used by each vehicle to map the collected environmental information to the BEV perspective;

[0137] The encoding module is used by each intelligent connected vehicle to encode the environmental information collected by the sensors using the backbone network to obtain intermediate feature information of the surrounding scene.

[0138] The encryption and compression module allows each intelligent connected vehicle to encrypt and compress its own sensed information and broadcast it to other intelligent connected vehicles in the scene via V2X devices.

[0139] The decompression module is used to receive and decompress information from other intelligent connected vehicles in the environment via V2X sensors.

[0140] The conversion module is used to spatially align the perception information acquired by the vehicle and convert the feature information to the vehicle's coordinate system.

[0141] The first analysis module is used to analyze the dispersion of the sensing features in the overlapping areas of multi-source sensing data. If it is greater than the threshold, the current driving scenario is considered a high-risk scenario.

[0142] The result output module is used to fuse multi-source perception information of the vehicle to obtain panoramic perception results, which are then input into the decision planning module.

[0143] The second analysis module is used to analyze the uncertainty of the vehicle's collaborative perception results. If the uncertainty exceeds the threshold, the current driving scenario is considered a high-risk scenario.

[0144] The present invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a high-risk scene identification method based on vehicle-to-vehicle cooperation. This high-risk scene identification method based on vehicle-to-vehicle cooperation includes the following steps:

[0145] Step 1: In this scenario, intelligent connected vehicles use their own sensors to collect information about the surrounding environment.

[0146] Step 2: Each vehicle maps the collected environmental information to the BEV perspective;

[0147] Step 3: Each intelligent connected vehicle uses the backbone network to encode the environmental information collected by the sensors to obtain intermediate feature information of the surrounding scene;

[0148] Step 4: Each intelligent connected vehicle encrypts and compresses its own perceived information and broadcasts it to other intelligent connected vehicles in the scene via V2X devices;

[0149] Step 5: The vehicle receives perception information broadcast by intelligent connected vehicles in the environment through V2X sensors, and decodes and decompresses it;

[0150] Step Six: The vehicle spatially aligns the acquired perception information and transforms the feature information into the vehicle's coordinate system;

[0151] Step 7: For the overlapping areas of multi-source sensing data, analyze the dispersion of sensing features in the same area. If it is greater than the threshold, then the current driving scenario is judged to be a high-risk scenario.

[0152] Step 8: The vehicle integrates multi-source perception information to obtain a panoramic perception result output, which is then output to the decision planning module;

[0153] Step 9: Analyze the uncertainty of the vehicle's collaborative perception results. If it exceeds the threshold, the current driving scenario is considered a high-risk scenario.

[0154] The computer-readable storage medium provided by the present invention has computer-executable instructions that are not limited to the method operations described above, but can also execute related operations in the high-risk scene identification method based on vehicle-to-vehicle cooperation provided in any embodiment of the present invention.

[0155] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0156] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line DSL) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk, SSD), etc.

[0157] In the above embodiments, the various units and modules are divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0158] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be included within the scope of protection of the present invention. Furthermore, all content not described in detail in this specification is prior art known to those skilled in the art.

Claims

1. A method for identifying high-risk scenarios based on vehicle-to-vehicle cooperation, characterized in that, Includes the following steps: Step 1: In this scenario, intelligent connected vehicles use their own sensors to collect information about the surrounding environment. Step 2: Each vehicle maps the collected environmental information to the BEV perspective; Step 3: Each intelligent connected vehicle uses the backbone network to encode the environmental information collected by the sensors to obtain intermediate feature information of the surrounding scene; Step 4: Each intelligent connected vehicle encrypts and compresses its own perceived information and broadcasts it to other intelligent connected vehicles in the scene via V2X devices; Step 5: The vehicle receives perception information broadcast by intelligent connected vehicles in the environment through V2X sensors, and decodes and decompresses it; Step Six: The vehicle spatially aligns the acquired perception information and transforms the feature information into the vehicle's coordinate system; Step 7: For the overlapping areas of multi-source sensing data, analyze the dispersion of sensing features in the same area. If it is greater than the threshold, then the current driving scenario is judged to be a high-risk scenario. Step 8: The vehicle integrates multi-source perception information to obtain a panoramic perception result output, which is then output to the decision planning module; Step 9: Analyze the uncertainty of the vehicle's collaborative perception results. If it exceeds the threshold, the current driving scenario is considered a high-risk scenario. Step seven describes analyzing the dispersion of sensor features in overlapping areas of multi-source sensing data. If the dispersion exceeds a threshold, the current driving scenario is considered a high-risk scenario. The specific details are as follows: For overlapping sensing areas of multiple vehicles in a scenario, the dispersion of multi-source data is analyzed. If the dispersion index of the overlapping sensing area of ​​multiple vehicles is greater than a threshold... This indicates that the vehicle perception model has significant uncertainty in perceiving the current driving scenario, and may even lack the ability to perceive the current driving scenario. The current driving scenario should be judged as a high-risk area, and the vehicle should issue a takeover warning to the driver or adjust the decision-making and planning strategy. Assuming the received environmental characteristics of each intelligent connected vehicle Follows probability distribution For overlapping areas The dispersion of feature information in the overlapping areas of perception of self-vehicle analysis is measured by relative entropy, which measures the dispersion of the distribution of self-vehicle features and the perception features of other vehicles in the environment. in, Represents the expected value of the distribution; The discrete index of the entire scene is defined as If the dispersion index If so, the current scenario will be judged as a high-risk scenario.

2. The high-risk scene identification method based on vehicle-to-vehicle cooperation according to claim 1, characterized in that, In step one, the intelligent connected vehicle uses its own sensors to collect information about the surrounding environment. The specific details are as follows: Intelligent connected vehicles exist in the environment Each vehicle is equipped with the same perception sensors and a collaborative perception model, enabling it to collect raw sensor data about the surrounding driving environment at a certain time frequency, represented as... in, Indicates a vehicle, This represents the vehicle's sensory information. This refers to intelligent connected vehicles other than the vehicle itself in the scenario. This represents the perception information of intelligent connected vehicles other than the vehicle itself in the scene; Meanwhile, intelligent connected vehicles in the scenario All are equipped with V2X communication devices, enabling information sharing between vehicles; In step two, each vehicle maps the collected environmental information to the BEV perspective, as detailed below: Each vehicle will collect information about its surrounding environment from its sensors. The transformation is performed from the sensor coordinate system and mapped to the BEV viewpoint using a transformation function, as follows: in, Indicates vehicle Perception of the surrounding environment from the perspective of a BEV This represents the BEV perspective mapping model.

3. The high-risk scenario identification method based on vehicle-to-vehicle cooperation according to claim 1, characterized in that, In step three, each intelligent connected vehicle uses the backbone network to encode the environmental information collected by the sensors to obtain intermediate feature information of the surrounding scene, as detailed below: Each intelligent connected vehicle uses the backbone network to process the information collected by the sensors. Encode the intermediate features of the surrounding scene. , It is a high-dimensional tensor, represented as in, Indicates vehicle Feature information of the surrounding scene, Represents the backbone network model; between different vehicles Shared parameters.

4. The method for identifying high-risk scenarios based on vehicle-to-vehicle cooperation according to claim 1, characterized in that, In step four, each intelligent connected vehicle encrypts and compresses its own perceived information and broadcasts it to other intelligent connected vehicles in the scene via V2X devices. The specific content is as follows: The perception information obtained by each intelligent connected vehicle through on-board computing The information is encrypted and compressed, and then broadcast to the scene via V2X devices, including the sensing information. Including feature information of the surrounding scene Pose information and time information ; in, in, For intelligent connected vehicles Global coordinates in the geodetic coordinate system Intelligent connected vehicles Longitude, latitude and altitude For intelligent connected vehicles The heading angle.

5. The high-risk scene identification method based on vehicle-to-vehicle cooperation according to claim 1, characterized in that, In step six, the vehicle spatially aligns the acquired perception information and transforms the feature information into the vehicle's coordinate system. The specific details are as follows: The vehicle obtains information about intelligent connected vehicles in the environment through V2X devices. Perceived information Then, based on the pose information of the two vehicles, a coordinate transformation model is constructed to obtain the coordinate transformation data. Medium environmental characteristics A coordinate transformation is performed to convert the vehicle to its own coordinate system, and a "spatial position-feature information" mapping is added to obtain the intelligent connected vehicle in its own coordinate system. Feature information in, Indicating intelligent connected vehicles Coordinate system to intelligent connected vehicles Coordinate system transformation model, Indicating intelligent connected vehicles The transformation model from the coordinate system to the vehicle. This indicates the location information within the scene.

6. The high-risk scene identification method based on vehicle-to-vehicle cooperation according to claim 1, characterized in that, Step eight describes the process of fusing multi-source perception information to obtain a panoramic perception result, which is then output to the decision-making and planning module. The specific details are as follows: The vehicle integrates multi-source perception features and performs feature extraction and scene understanding through a collaborative perception information fusion network and a detection head to obtain a panoramic perception result of the scene. in, This indicates the current perception result. For the detection head model, A multi-source information fusion model; For feature concatenation function, Indicating intelligent connected vehicles In the car Features in a coordinate system; Regarding the perception results in, To sense the number of obstacles in the output, Obstacles Location information, Obstacles of parameter, For types of obstacles, , where is the confidence level.

7. A high-risk scene identification method based on vehicle-to-vehicle cooperation according to claim 6, characterized in that, Step nine involves analyzing the uncertainty of the vehicle's collaborative perception results. If the uncertainty exceeds a threshold, the current driving scenario is considered a high-risk scenario. The specific details are as follows: Based on the confidence level of the perception results, the autonomous vehicle measures the degree of uncertainty by calculating the information entropy of each traffic participant in the perception results; in, Indicates obstacle detection Information entropy; For the entire scene, calculate the degree of uncertainty of the perception results for the entire scene, and obtain the uncertainty index of the entire scene. in, This represents the uncertainty index of the entire scenario. If so, the current driving scenario is considered a high-risk scenario.

8. The identification device for a high-risk scene identification method based on vehicle-to-vehicle cooperation as described in any one of claims 1 to 7, characterized in that, include: The information acquisition module is used by intelligent connected vehicles in various scenarios to collect information about their surrounding environment using their own sensors. The mapping module is used by each vehicle to map the collected environmental information to the BEV perspective; The encoding module is used by each intelligent connected vehicle to encode the environmental information collected by the sensors using the backbone network to obtain intermediate feature information of the surrounding scene. The encryption and compression module allows each intelligent connected vehicle to encrypt and compress its own sensed information and broadcast it to other intelligent connected vehicles in the scene via V2X devices. The decompression module is used to receive and decompress information from other intelligent connected vehicles in the environment via V2X sensors. The conversion module is used to spatially align the perception information acquired by the vehicle and convert the feature information to the vehicle's coordinate system. The first analysis module is used to analyze the dispersion of the sensing features in the overlapping areas of multi-source sensing data. If it is greater than the threshold, the current driving scenario is considered a high-risk scenario. The result output module is used to fuse multi-source perception information of the vehicle to obtain panoramic perception results and output them to the decision planning module. The second analysis module is used to analyze the uncertainty of the vehicle's collaborative perception results. If the uncertainty exceeds the threshold, the current driving scenario is considered a high-risk scenario.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images