An automatic driving behavior decision system and method based on driving preference

Abstract

The present invention discloses an automatic driving behavior decision system and method based on driving preference, including a vehicle history data collection module, which collects and processes the historical data of vehicle driving; a driving preference identification module, which obtains the driving preference identification result according to the historical data and preset label data; a road environment information processing module, which collects and processes the road environment information of the vehicle driving and obtains the scene recognition result; and a behavior decision module, which outputs the vehicle action instruction according to the scene recognition result and the recognition result of the driving preference. The present invention can better meet the driver's personalized needs while ensuring safety by identifying the driver's driving habits and automatically matching the driver's driving preferences.

Description

Technical Field

[0001] The present invention belongs to the technical field of automobile automatic driving, and more specifically to an automatic driving behavior decision system and method based on driving preference. Background Art

[0002] According to statistics, 90% of traffic accidents are caused by driver errors. The main goal of developing self-driving cars is to improve driving safety and driver acceptance. Self-driving behavior decision-making is one of the difficult problems that need to be overcome in the development of self-driving systems, and it is also a key technology to effectively reduce the incidence of collision traffic accidents. Self-driving cars in future intelligent transportation systems should be equipped with advanced decision-making systems to make safe and driver-acceptable driving decisions. However, factors such as the diversity of road scenes, sudden changes in light, and weather conditions have brought challenges to the development of self-driving behavior decision-making technology.

[0003] At present, there are two main types of decision-making technologies: rule-based and learning-based. The rule-based method is to determine the judgment conditions by hard coding for specific scenarios, and make decisions based on the current environmental information, such as decision trees and finite state machines. This method is highly debuggable and easy to implement, and is currently the most commonly used method in engineering. However, since it is only for specific scenarios, it has poor adaptability to the driving environment and is not universal. In view of the shortcomings of rule-based decision-making technology, a learning-based method has been proposed. For example, my country's patent CN111845741A discloses an automatic driving decision control method and system based on hierarchical reinforcement learning. The scheme includes an action layer agent and a decision layer agent. The action layer agent trains the correct action network by interacting with road environment information, and the decision layer agent trains the correct decision network by interacting with traffic status information, thereby avoiding the occurrence of sudden acceleration and deceleration and improving the comfort of the whole vehicle. The advantage of the learning-based decision method is that due to the diversity of network structure and output, the complexity of the end-to-end system can be reduced, the flexibility of the system can be improved, and the network can produce real behavior like a human driver. However, since data on dangerous scenarios is difficult to obtain and cannot be used to train learning models, learning-based methods cannot handle dangerous collision scenarios well. In addition, the controllability of the learning model is poor, making it difficult to perform targeted debugging for special scenarios during engineering development.

[0004] Whether it is rule-based or learning-based behavior decision-making technology, neither takes the driver's driving preferences (driving style) into consideration. In real driving, each driver has unique driving preferences, such as aggressive driving, normal driving, and steady driving. On the premise of ensuring safety, how autonomous driving behavior decisions meet driver preferences is directly related to the experience and acceptability of autonomous driving functions, and is also a key issue that needs to be considered during the development of autonomous driving systems. Summary of the invention

[0005] To solve the above problems, the present invention provides an automatic driving behavior decision system and method based on driving preference, which can better meet the driver's personalized needs while ensuring safety by identifying the driver's driving habits and automatically matching the driver's driving preferences.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows: an automatic driving behavior decision system based on driving preference, comprising:

[0007] Vehicle historical data collection module, collects and processes vehicle driving historical data;

[0008] A driving preference recognition module, which obtains a driving preference recognition result according to the historical data and preset label data;

[0009] The road environment information processing module collects and processes the road environment information of the vehicle and obtains the scene recognition result;

[0010] The behavior decision module outputs a vehicle action instruction according to the scene recognition result and the driving preference recognition result.

[0011] Based on the above system, the present invention also provides an automatic driving behavior decision method based on driving preference, comprising the following steps:

[0012] S1. Obtain preset label data;

[0013] S2, collect and process historical data of vehicle travel;

[0014] S3, collecting and processing the road environment information of the vehicle, and obtaining the scene recognition result;

[0015] S4, obtaining a driving preference recognition result according to the historical data and the label data;

[0016] S5. Outputting vehicle action instructions according to the driving preference recognition results and the scene recognition results.

[0017] As an optimization, the preset label data collects the driver's driving preference data through a questionnaire and stores it as corresponding label data.

[0018] As an optimization, the content of the questionnaire includes the driver's fuel consumption statistics per 100 kilometers during normal driving and / or self-evaluation and / or 100 kilometers acceleration time test and / or vehicle speed and throttle opening changes, and the driver's driving preference data for aggressive driving, normal driving, and conservative driving are determined based on the content of the questionnaire.

[0019] As an optimization, the historical data includes a vehicle speed signal and a throttle opening change rate of the vehicle.

[0020] As an optimization, the road environment information includes obstacle information, vehicle information and map information.

[0021] As an optimization, S4 includes,

[0022] S301, establishing a probability statistics model using a kernel density function, wherein the kernel function includes a Gaussian kernel function, a polynomial kernel function, or a linear kernel function;

[0023] S302, offline calculation of the probability of conservative driving and aggressive driving based on the Bayesian principle;

[0024] S303, dividing the probability difference between conservative driving and aggressive driving into three levels according to the normal distribution principle, using the three levels as recognition criteria, and determining a recognition model;

[0025] S304: Input the historical data into a recognition model to obtain a recognition result of the driving preference.

[0026] As an optimization, S5 includes,

[0027] S501, determining the lane ID where the vehicle is located and the lane ID where the obstacle is located according to the map information;

[0028] S502, identifying the current driving scenario based on the vehicle information and obstacle information, including normal meeting, oncoming vehicle, extreme insertion, and static detour;

[0029] S503, calculating the threat level of surrounding obstacles based on the conditional random field, and screening out obstacles with the highest threat level;

[0030] S504, based on the current headway between the obstacle and the vehicle, the collision time, and the stopping time, a probability-based risk assessment model is established using a Gaussian model to calculate the collision probability in the current driving scenario;

[0031] S505, classifying the collision risk level into three levels according to the collision probability: emergency braking, normal braking and mild braking, and matching them with three driving preferences;

[0032] S506: Output corresponding vehicle action instructions according to the driving preference recognition result.

[0033] Compared with the prior art, the present invention has the following advantages:

[0034] The present invention incorporates the driver's driving preference factors into the autonomous driving behavior decision-making process. By collecting vehicle historical data, the current driver's driving preferences, such as aggressive, normal, and conservative, are automatically identified based on Bayesian theory. According to the identification of the current driving scene, autonomous driving behavior decisions are made based on risk probability assessment, and finally a behavioral instruction set that can meet safety requirements and the driver's driving preferences is output. This solution can better meet the driver's personalized needs while ensuring safety by identifying the driver's driving habits and automatically matching the driver's driving preferences. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The overall system block diagram for behavioral decision making of the present invention;

[0036] Figure 2 The flowchart of the automatic identification of driving preference based on Bayesian method of the present invention is as follows;

[0037] Figure 3 This is a flowchart of the autonomous driving behavior decision-making based on risk probability assessment of the present invention. DETAILED DESCRIPTION

[0038] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0039] Example: See Figure 1-Figure 3 ,

[0040] An automatic driving behavior decision system based on driving preference, comprising:

[0041] Vehicle historical data collection module, collects and processes vehicle driving historical data;

[0042] A driving preference recognition module, which obtains a driving preference recognition result according to the historical data and preset label data;

[0043] The road environment information processing module collects and processes the road environment information of the vehicle and obtains the scene recognition result;

[0044] The behavior decision module outputs a vehicle action instruction according to the scene recognition result and the driving preference recognition result.

[0045] Specifically, Figure 1As shown, the present invention includes the following modules: vehicle history data acquisition module, driving preference recognition module, road environment information processing module, scene recognition module and behavior decision module. The vehicle history data acquisition module collects vehicle speed signals and throttle opening change rate, and after noise reduction and screening, transmits stable and effective data to the driving preference recognition module. The driving preference recognition module automatically identifies the preference type based on the Bayesian scheme and transmits the identified result to the behavior decision module. The road environment information processing module transmits the position, type, speed, acceleration of obstacles, the position, speed, acceleration of the vehicle, and map information to the scene recognition module. The behavior decision module determines the vehicle action instruction based on the risk probability assessment method according to the results of the scene recognition, obstacle information and driving preference.

[0046] Based on the above system, the present invention also provides an automatic driving behavior decision method based on driving preference, comprising the following steps:

[0047] S1. Obtain preset label data; the preset label data collects the driver's driving preference data through a questionnaire and stores it as corresponding label data. The content of the questionnaire includes the driver's fuel consumption statistics per 100 kilometers during normal driving and / or self-evaluation and / or 100 kilometers acceleration time test and / or vehicle speed and throttle opening changes, and the driver's driving preference data for aggressive driving, normal driving, and conservative driving are determined based on the content of the questionnaire. Specifically, the questionnaire questions can truly reflect the driver's driving habits and driving style, including but not limited to: fuel consumption statistics per 100 kilometers during normal driving, self-evaluation, 100 kilometers acceleration time test and other methods. Collect the driver's aggressive driving, normal driving, and conservative driving driving data in the questionnaire: including vehicle speed and throttle opening changes, and determine the label data.

[0048] S2. Collect and process historical data of vehicle travel; the historical data includes the vehicle speed signal and throttle opening change rate.

[0049] S3. Collect and process the road environment information of the vehicle to obtain scene recognition results; the road environment information includes obstacle information, vehicle information and map information.

[0050] S4, obtaining a driving preference recognition result according to the historical data and the label data; specifically, Figure 2 As shown, the driving preference automatic identification module uses the Bayesian principle to identify the current driver's driving preference based on the label data and the data collected by the vehicle history data collection module in the current period.

[0051] S301, establishing a probability statistics model using a kernel density function, wherein the kernel function includes a Gaussian kernel function, a polynomial kernel function, or a linear kernel function;

[0052] S302, offline calculation of the probability of conservative driving and aggressive driving based on the Bayesian principle;

[0053] S303, dividing the probability difference between conservative driving and aggressive driving into three levels according to the normal distribution principle, using the three levels as recognition criteria, and determining a recognition model;

[0054] S304: Input the vehicle speed and throttle opening change rate collected by the vehicle history data collection module into the recognition model to obtain the recognition result of the driving preference.

[0055] S5. Outputting vehicle action instructions according to the driving preference recognition results and the scene recognition results. Figure 3 As shown, the function of the autonomous driving behavior decision module based on risk probability assessment is to output control instructions for specific behavioral actions of the vehicle, such as cruising, following the vehicle, keeping lane, detouring, stopping and waiting, etc.

[0056] S501, determining the basic conditions for dividing the driving scene: including the lane ID of the vehicle and the lane ID of the obstacle vehicle;

[0057] S502, determining the current driving scene according to the moving direction, speed, and position of the vehicle and the moving direction, speed, and position of the obstacle; formulating scene recognition rules to recognize the current driving scene, including: normal meeting, oncoming vehicle, extreme insertion, static detour, etc.;

[0058] S503, calculating the threat level of surrounding obstacles based on the conditional random field, and screening out obstacles with the highest threat level;

[0059] S504, based on the current headway between the obstacle and the vehicle, the collision time, and the stopping time, a probability-based risk assessment model is established using a Gaussian model to calculate the collision probability in the current driving scenario;

[0060] S505, classifying the collision risk level into three levels according to the collision probability: emergency braking, normal braking and mild braking, and matching them with three driving preferences;

[0061] S506. Outputting corresponding vehicle action instructions according to the driving preference identification result, that is, calibrating the braking deceleration of the actual vehicle for each driving preference.

[0062] The present invention incorporates the driver's driving preference factors into the autonomous driving behavior decision-making process. By collecting vehicle historical data, the current driver's driving preferences, such as aggressive, normal, and conservative, are automatically identified based on Bayesian theory. According to the identification of the current driving scene, autonomous driving behavior decisions are made based on risk probability assessment, and finally a behavioral instruction set that can meet safety requirements and the driver's driving preferences is output. This solution can better meet the driver's personalized needs while ensuring safety by identifying the driver's driving habits and automatically matching the driver's driving preferences.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the technical solution. Those skilled in the art should understand that those modifications or equivalent substitutions of the technical solution of the present invention that do not depart from the purpose and scope of the technical solution should be included in the scope of the claims of the present invention.

Claims

1. A method for autonomous driving behavior decision-making based on driving preference, characterized in that: The following steps are included: S1. Obtain preset label data; S2, collect and process historical data of vehicle travel; S3, collecting and processing the road environment information of the vehicle, and obtaining the scene recognition result; the road environment information includes obstacle information, vehicle information and map information; S4, obtaining a driving preference recognition result according to the historical data and the label data; S401, establishing a probability statistics model using a kernel density function, wherein the kernel density function includes a Gaussian kernel function, a polynomial kernel function, or a linear kernel function; S402, offline calculation of the probability of conservative driving and aggressive driving based on the Bayesian principle; S403, dividing the probability difference between conservative driving and aggressive driving into three levels according to the normal distribution principle, using the three levels as recognition criteria, and determining a recognition model; S404, inputting the historical data into a recognition model to obtain a recognition result of driving preference; S5. Outputting vehicle action instructions according to the driving preference recognition results and the scene recognition results; S501, determining the lane ID where the vehicle is located and the lane ID where the obstacle is located according to the map information; S502, identifying the current driving scenario based on the vehicle information and obstacle information, including normal meeting, oncoming vehicle, extreme insertion, and static detour; S503, calculating the threat level of surrounding obstacles based on the conditional random field, and screening out obstacles with the highest threat level; S504, based on the current headway between the obstacle and the vehicle, the collision time, and the stopping time, a probability-based risk assessment model is established using a Gaussian model to calculate the collision probability in the current driving scenario; S505, classifying the collision risk level into three levels according to the collision probability: emergency braking, normal braking and mild braking, and matching them with three driving preferences; S506: Output corresponding vehicle action instructions according to the driving preference recognition result.

2. The automatic driving behavior decision method based on driving preference according to claim 1, characterized in that: The preset label data collects the driver's driving preference data through a questionnaire and stores it as corresponding label data.

3. The automatic driving behavior decision method based on driving preference according to claim 2, characterized in that: The content of the questionnaire includes the driver's fuel consumption statistics per 100 kilometers during normal driving and / or self-evaluation and / or 100 kilometers acceleration time test and / or vehicle speed and throttle opening changes, and the driver's driving preference data for aggressive driving, normal driving, and conservative driving are determined based on the content of the questionnaire.

4. The method for autonomous driving behavior decision-making based on driving preference according to any one of claims 1 to 3, characterized in that: The historical data includes a vehicle speed signal and a throttle opening change rate of the vehicle.

Images