Intelligent driving collision avoidance system and method based on vehicle networking

Abstract

The application discloses an intelligent driving collision avoidance system and method based on Internet of Vehicles, and belongs to the technical field of automatic driving. The system comprises a vehicle-mounted unit, a roadside unit and a service platform. The vehicle-mounted unit is used for collecting vehicle state information and surrounding environment information and sending the information to the roadside unit and the service platform. The roadside unit is used for collecting road condition information, performing target detection based on the surrounding environment information and the road condition information, and obtaining obstacle information. The service platform is used for evaluating collision risks according to the vehicle state information and the obstacle information, predicting the best collision avoidance strategy based on the collision risks, and sending instructions to the vehicle-mounted unit for collision avoidance control. The system can analyze potential collision risks based on real-time shared vehicle state information and road condition information, and improve the safety of road traffic. The system solves the problem that traditional collision avoidance systems mainly rely on the perception and decision of a single vehicle and lack cooperation with other vehicles and road infrastructure.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, in particular to a smart driving collision avoidance system and method based on vehicle networking. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute the prior art.

[0003] With the rapid development of the automotive industry and the increasing perfection of intelligent transportation systems, automatic driving technology has become an important direction for future transportation development. However, in complex traffic environments, how to effectively avoid collisions and ensure driving safety is still a major challenge for automatic driving technology.

[0004] Traditional collision avoidance systems mainly rely on the perception and decision-making of a single vehicle, lacking coordination with other vehicles and road infrastructure, and have the following problems:

[0005] 1. Limited perception range: Due to the physical limitations of sensors, vehicles can only detect obstacles or other vehicles within their direct line of sight. This means that in some cases, such as curves, adverse weather conditions, or under obstructions, the effectiveness and reliability of the system will be greatly reduced.

[0006] 2. Information silos: Each vehicle is an independent information processing unit and cannot share real-time data with other vehicles or road infrastructure. This leads to incomplete information, making it difficult for the system to make optimal collision avoidance decisions.

[0007] 3. Delayed response: From perception to decision-making to execution, it takes a certain amount of time, especially at high speeds, and even a small time difference can lead to accidents. SUMMARY

[0008] To solve the problems of the prior art, the present application provides a smart driving collision avoidance system and method based on vehicle networking, which analyzes potential collision risks based on real-time shared vehicle state information and road condition information, improving road traffic safety.

[0009] In a first aspect, the present application provides a smart driving collision avoidance system based on vehicle networking, which comprises a vehicle-mounted unit, a roadside unit and a service platform.

[0010] A smart driving collision avoidance system based on vehicle networking, comprising a vehicle-mounted unit, a roadside unit and a service platform.

[0011] The vehicle-mounted unit is configured to collect vehicle state information and surrounding environment information and send the information to the road-side unit and the service platform; the road-side unit is configured to collect road condition information, perform target detection based on the surrounding environment information and the road condition information, and obtain obstacle information; and the service platform is configured to evaluate a collision risk based on the vehicle state information and the obstacle information, predict an optimal collision avoidance strategy based on the collision risk, and send an instruction to the vehicle-mounted unit to perform collision avoidance control.

[0012] In some embodiments, the target detection based on the surrounding environment information and the road condition information and the obtaining of the obstacle information specifically include: processing the environment image and the road traffic image through a trained target detection model to obtain an obstacle detection result; and processing the obstacle detection results in consecutive time through a trained trajectory tracking model to obtain a motion trajectory of a dynamic obstacle.

[0013] In some embodiments, the target detection model is a YOLOv8s model, and the trajectory tracking model is a ReID model.

[0014] In some embodiments, the evaluation of the collision risk based on the vehicle state information and the obstacle information includes:

[0015] determining a longitudinal risk time based on the longitudinal position and the longitudinal speed of the vehicle and the dynamic obstacle and the length of the vehicle;

[0016] determining a lateral risk time based on the lateral position and the lateral speed of the vehicle and the dynamic obstacle and the width of the vehicle;

[0017] determining a risk index based on the longitudinal risk time and the lateral risk time; and regarding the existence of a potential collision risk if the risk index is greater than a preset risk threshold.

[0018] In some embodiments, the determination of the longitudinal risk time based on the longitudinal position and the longitudinal speed of the vehicle and the dynamic obstacle and the length of the vehicle specifically includes: calculating a longitudinal collision time and a longitudinal headway time of the vehicle and the dynamic obstacle; and determining the lateral risk time based on the longitudinal collision time and the longitudinal headway time.

[0019] In some embodiments, the prediction of the optimal collision avoidance strategy based on the collision risk specifically includes: processing real-time traffic data through a behavior prediction model to obtain a behavior prediction result of the dynamic obstacle; and screening the behavior prediction result in a preset collision avoidance strategy library to determine the optimal collision avoidance strategy.

[0020] In some embodiments, the behavior prediction model is an improved BiLSTM network including an input layer, a BiLSTM layer, an Attention layer, and an output layer.

[0021] In some embodiments, the behavior prediction result is screened in a preset collision avoidance strategy library to determine the optimal collision avoidance strategy, specifically by calculating the Pearson correlation coefficient between the behavior prediction result and the collision avoidance strategy, and determining the optimal collision avoidance strategy based on the size of the Pearson correlation coefficient.

[0022] In some embodiments, the on-board unit comprises a first sensor module, a first communication module and a control module, the road-side unit comprises a second sensor module, a second communication module and a data processing module, and the service platform is deployed on an edge computing device supported by a 5G network.

[0023] In a second aspect, the present application provides an intelligent driving collision avoidance method based on vehicle networking.

[0024] An intelligent driving collision avoidance method based on vehicle networking, based on the intelligent driving collision avoidance system based on vehicle networking, comprising:

[0025] The on-board unit collects vehicle state information and surrounding environment information and sends them to the road-side unit and the service platform.

[0026] The road-side unit collects road condition information, performs target detection based on the surrounding environment information and the road condition information, and obtains obstacle information.

[0027] The service platform evaluates the collision risk based on the vehicle state information and the obstacle information, predicts the optimal collision avoidance strategy based on the collision risk, and sends instructions to the on-board unit for collision avoidance control.

[0028] Compared with the prior art, the present application has the following advantages:

[0029] 1. The technical scheme provided by the present application realizes real-time and comprehensive monitoring of road conditions and vehicle driving conditions through real-time communication and data sharing between vehicles, road-sides and service platforms, expands the perception range of vehicles, provides comprehensive data support for collision avoidance decision-making, and improves the effectiveness and reliability of automatic driving.

[0030] 2. The technical scheme provided by the present application realizes pre-evaluation and collaborative collision avoidance of potential collision risks, and improves the safety of road traffic; after evaluating the potential collision risk, the service platform communicates with other vehicles or traffic infrastructure to share information for early warning of other vehicles.

[0031] 3. The technical scheme provided by the present application ensures that information can be immediately communicated to all relevant parties through 5G communication, reducing the delay of data transmission. BRIEF DESCRIPTION OF DRAWINGS

[0032] The accompanying drawings, which form a part of this specification, are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification. The embodiments of the application, and their

[0033] Figure 1 The system framework schematic diagram provided for the embodiments of the application. DETAILED DESCRIPTION

[0034] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0035] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments in accordance with the application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0036] The embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0037] Embodiment one

[0038] The automatic driving of existing vehicles relies on the perception and decision of a single vehicle, and lacks coordination with other vehicles and road infrastructure; therefore, the application provides a vehicle networking-based intelligent driving cooperative collision avoidance system, which analyzes potential collision risks based on real-time shared vehicle state information and road condition information, and improves the safety of road traffic.

[0039] Next, combined with Figure 1 A vehicle networking-based intelligent driving cooperative collision avoidance system disclosed in the embodiment will be described in detail. The vehicle networking-based intelligent driving cooperative collision avoidance system comprises a vehicle-mounted unit, a roadside unit and a service platform, the vehicle-mounted unit comprises a first sensor module, a first communication module and a control module, is deployed on a vehicle participating in cooperative collision avoidance and is connected to the service platform; the roadside unit comprises a second sensor module, a second communication module and a data processing module, is deployed on a road requiring cooperative collision avoidance; and the service platform is deployed on an edge computing device supported by a 5G network, can quickly analyze data from the vehicle-mounted unit and the roadside unit, reduces the time of data transmission to the service platform and reduces the time delay.

[0040] The first sensor module is used to collect vehicle state information and surrounding environment information, the vehicle state information includes the vehicle's own speed, position, acceleration, driving direction, etc., and the surrounding environment information is an environmental image; the first communication module is used to communicate with other vehicle-mounted units and roadside units in real time, and is realized by using vehicle-to-everything (V2X) technology; the control module is used to receive information from the communication module, and generate collision avoidance instructions according to the cooperative collision avoidance algorithm, and control the vehicle to perform collision avoidance operations. The first sensor module includes a vehicle-mounted GPS, an inertial measurement unit, a wheel speed sensor, a steering angle sensor, a camera and the like.

[0041] The second sensor module is used to collect road condition information and traffic signal information, such as road slipperiness, traffic congestion, traffic light status, road traffic image, etc.; the second communication module is used to send the collected information to the vehicle-mounted unit; the data processing module is used to receive information from the vehicle-mounted unit and combine local information for comprehensive analysis to support cooperative collision avoidance. Specifically, target detection is performed on the environmental image and the road traffic image to detect the surrounding obstacles, and the detected obstacles are continuously tracked to obtain their motion trajectories and speed change information. The second sensor module includes a temperature sensor, a humidity sensor, a temperature sensor, and a camera.

[0042] The service platform is used to receive and process information from vehicle-mounted units and roadside units, analyze potential collision risks, and send collision avoidance instructions to related vehicles; the service platform needs to have strong data processing capability and intelligent decision-making capability to support the operation of the cooperative collision avoidance algorithm.

[0043] Specifically, the collision risk assessment model processes the vehicle's own speed, acceleration, driving direction, and the position, speed, and acceleration of the obstacles, and calculates the potential collision risk between the vehicle and the obstacles. After evaluating the potential collision risk, communicate with other vehicle-mounted units or roadside units, share information, and a big data model analyzes a large amount of real-time traffic data, including vehicle position, speed, direction, environment, etc., to predict the behavior of other vehicles and pedestrians, and predict the best collision avoidance strategy.

[0044] As an implementation, target detection is performed on the environmental image, road condition image and traffic signal image to detect the surrounding obstacles, and the detected obstacles are continuously tracked to obtain their motion trajectories and speed change information. Specifically:

[0045] S101, input the environmental image and the road traffic image into the trained YOLOv8s model for processing to obtain obstacle detection results, the obstacle detection results being the image of the framed obstacle and the coordinate information of the obstacle.

[0046] The YOLO series model is known for its efficient real-time detection capability and high accuracy, and is very suitable for applications such as autonomous driving that require fast decision-making. Therefore, in this embodiment, the existing YOLOv8s model is directly applied for obstacle detection, and before use, environmental images and road traffic images under various environmental conditions are collected and labeled for the position and category of obstacles to construct a training set. The YOLOv8s model is trained through the training set until the effect is good. Here, the detected obstacles have a total of four categories, including cars, bicycles, pedestrians, and roadblocks.

[0047] S102, through the trained ReID model, the image sequence of the obstacle is tracked, and the motion trajectory of the dynamic obstacle is output; according to the motion trajectory of the dynamic obstacle, the position, speed and acceleration of the dynamic obstacle can be calculated.

[0048] As an implementation, the specific process of calculating the potential collision risk between the vehicle and the dynamic obstacle through the collision risk assessment model for the speed, driving direction of the vehicle and the position, speed of the dynamic obstacle is as follows:

[0049] S201, according to the longitudinal position, longitudinal speed and length of the vehicle and the longitudinal position and longitudinal speed of the dynamic obstacle, the longitudinal collision time and longitudinal time between the vehicle and the dynamic obstacle are calculated; according to the longitudinal collision time and longitudinal time, the longitudinal risk time is determined. Wherein, the longitudinal collision time is represented as:

[0050]

[0051] The longitudinal time is represented as:

[0052]

[0053] The longitudinal risk time is represented as:

[0054]

[0055] In the formula, y i represents the longitudinal position of the i-th dynamic obstacle, y0 represents the longitudinal position of the vehicle, l represents the length of the vehicle, v y0 represents the longitudinal speed of the vehicle, v yi represents the longitudinal speed of the i-th dynamic obstacle, ξ a , ξ b represents a threshold value.

[0056] TTC represents the time required before the vehicle collides with the dynamic obstacle, the smaller the value, the greater the risk of collision; THW represents the time required for the vehicle to travel to the relative distance of the dynamic obstacle, the smaller the value, the greater the risk of collision. When the relative distance between the vehicle and the dynamic obstacle is short and the relative speed is small, if the dynamic obstacle brakes urgently, the current distance is not enough to brake for safety, but due to the small relative speed, TTC cannot represent the risk, and THW has a small value, which can represent a greater risk, thus the longitudinal risk time is defined according to the threshold value.

[0057] S202, determining a lateral risk time according to the lateral position, lateral speed and vehicle width of the vehicle and the lateral position and lateral speed of the dynamic obstacle. The lateral risk time is represented as:

[0058]

[0059] In the formula, x0 represents the lateral position of the vehicle, xi represents the lateral position of the i-th dynamic obstacle, b represents the vehicle width, vi represents the lateral speed of the i-th dynamic obstacle, and v represents the lateral speed of the vehicle. i xi x0

[0060] S203, determining a risk index according to the longitudinal risk time and the lateral risk time; if the risk index is greater than a preset risk threshold, considering that there is a potential risk of collision. The risk index is represented as:

[0061]

[0062] In the formula, η is a very small number to prevent calculation errors.

[0063] As an embodiment, the big data model predicts the behavior of other vehicles and pedestrians by analyzing a large amount of real-time traffic data, including the positions, speeds, directions, and environments of other vehicles, and predicts the specific best collision avoidance strategy, which includes:

[0064] S301, inputting the position, speed, direction and environment information of other vehicles and pedestrians at the previous moment into the trained behavior prediction model to obtain a behavior prediction result.

[0065] In this embodiment, an improved BiLSTM network is used as a behavior prediction model, the input is position, speed, direction and environment information, and the output is a behavior prediction result, such as lane changing, crossing the road, turning, etc.

[0066] ​​​Specifically, different from the existing BiLSTM network, it includes an input layer, a BiLSTM layer, an Attention layer, and an output layer. The Attention layer is the output of the learning function F added on the basis of the BiLSTM layer, and is expressed as:

[0067]

[0068] ξ t =F(h′ t );

[0069]

[0070] S302, according to the behavior prediction result of other vehicles and pedestrians, screening in the preset collision avoidance strategy library, determining the best collision avoidance strategy through correlation analysis and sending to the vehicle side unit.

[0071] For example, the vehicle is suggested to slow down, change lanes or stop, while coordinating traffic lights and other infrastructure, optimizing traffic flow and reducing collision risk. Through 5G communication, information can be instantly conveyed to all relevant parties, including vehicles, traffic lights and other infrastructure. After receiving the collision avoidance instructions, the vehicle performs local fast execution through edge computing, such as adjusting the driving route and speed to achieve collaborative collision avoidance.

[0072] Here, the correlation analysis is to calculate the Pearson correlation coefficient between the behavior prediction result and the collision avoidance strategy, and to determine the best collision avoidance strategy based on the size of the Pearson correlation coefficient.

[0073] Embodiment two

[0074] Based on the intelligent driving collision avoidance system based on vehicle networking in embodiment one, the embodiment discloses an intelligent driving collision avoidance method based on vehicle networking, comprising:

[0075] The vehicle unit collects vehicle state information and surrounding environment information and sends them to the road side unit and the service platform;

[0076] The road side unit collects road condition information, performs target detection based on the surrounding environment information and the road condition information, and obtains obstacle information;

[0077] The service platform evaluates the collision risk according to the vehicle state information and the obstacle information, predicts the best collision avoidance strategy based on the collision risk, and sends instructions to the vehicle unit for collision avoidance control.

[0078] The description of each embodiment in the above embodiments has its own emphasis, and the parts not described in detail in a certain embodiment can refer to the relevant description of other embodiments.

[0079] The above merely provides the preferred embodiments of the present application, and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the principles and technical scope of the present application shall fall into the scope of the present application.

Claims

1. A collision avoidance system based on intelligent driving of Internet of Vehicles, characterized in that, The vehicle-mounted unit, the roadside unit and the service platform are included. The vehicle-mounted unit is used for collecting vehicle state information and surrounding environment information and sending to the roadside unit and the service platform; the roadside unit is used for collecting road condition information, target detection based on surrounding environment information and road condition information, and obtaining obstacle information; The service platform is used for evaluating collision risk according to vehicle state information and obstacle information, predicting the best collision avoidance strategy based on collision risk, and sending instructions to the vehicle-mounted unit for collision avoidance control; According to the longitudinal position, longitudinal speed and length of the vehicle and the longitudinal position and longitudinal speed of the dynamic obstacle, the longitudinal risk time is determined. The longitudinal risk time is determined according to the longitudinal position, longitudinal speed and length of the vehicle and the longitudinal position and longitudinal speed of the dynamic obstacle, and specifically, the longitudinal collision time and the longitudinal vehicle-to-vehicle time of the vehicle and the dynamic obstacle are calculated. The longitudinal collision time is represented as: ； The longitudinal vehicle-to-vehicle time is represented as: ； The longitudinal risk time is represented as: ； wherein denotes the longitudinal position of the i-th dynamic obstacle, denotes the longitudinal position of the vehicle, denotes the vehicle length, denotes the longitudinal vehicle speed, denotes the longitudinal speed of the i-th dynamic obstacle, , denotes a threshold value. 2.The intelligent driving collision avoidance system based on the Internet of Vehicles according to claim 1, wherein, The target detection model is YOLOv8s model, and the trajectory tracking model is ReID model. 3.The V2X-based intelligent driving collision avoidance system of claim 2, wherein, The collision risk is evaluated according to the vehicle state information and the obstacle information, which includes: 4.The V2X-based intelligent driving collision avoidance system of claim 1, wherein, According to the lateral position, lateral speed and width of the vehicle and the lateral position and lateral speed of the dynamic obstacle, the lateral risk time is determined. According to the longitudinal risk time and the lateral risk time, the risk index is determined; if the risk index is greater than the preset risk threshold, it is considered that there is potential collision risk. The longitudinal risk time is determined according to the longitudinal position, longitudinal speed and length of the vehicle and the longitudinal position and longitudinal speed of the dynamic obstacle, and specifically, the longitudinal collision time and the longitudinal vehicle-to-vehicle time of the vehicle and the dynamic obstacle are calculated. 5.The V2X-based intelligent driving collision avoidance system of claim 4, wherein, The best collision avoidance strategy is predicted based on the collision risk, which specifically includes: using a behavior prediction model to process real-time traffic data to obtain a behavior prediction result of the dynamic obstacle; based on the behavior prediction result, the best collision avoidance strategy is determined by screening in a preset collision avoidance strategy library. 6.The V2X-based intelligent driving collision avoidance system of claim 1, wherein, The behavior prediction model is an improved BiLSTM network, including an input layer, a BiLSTM layer, an Attention layer and an output layer. 7.The V2X-based intelligent driving collision avoidance system of claim 6, wherein, The best collision avoidance strategy is determined by screening in the preset collision avoidance strategy library based on the behavior prediction result, which specifically includes: calculating the Pearson correlation coefficient between the behavior prediction result and the collision avoidance strategy, and determining the best collision avoidance strategy based on the size of the Pearson correlation coefficient. 8.The V2X-based intelligent driving collision avoidance system of claim 6, wherein, The vehicle-mounted unit includes a first sensor module, a first communication module and a control module, the roadside unit includes a second sensor module, a second communication module and a data processing module, and the service platform is deployed on an edge computing device supported by a 5G network. 9.The intelligent driving collision avoidance system based on the Internet of Vehicles according to claim 1, wherein, It includes:

10. The method for avoiding collision in intelligent driving based on vehicle networking, applied to the system for avoiding collision in intelligent driving based on vehicle networking as claimed in any one of claims 1-9, characterized in that, ​ The vehicle-mounted unit collects vehicle state information and surrounding environment information and sends them to the roadside unit and the service platform; The roadside unit collects road condition information, conducts target detection based on the surrounding environment information and the road condition information, and obtains obstacle information; The service platform evaluates collision risk based on the vehicle state information and the obstacle information, predicts the best collision avoidance strategy based on the collision risk, and sends instructions to the vehicle-mounted unit for collision avoidance control.

Images