Vehicle safe lane changing method based on scene dependence causal influence analysis

Abstract

The invention discloses a vehicle safe lane changing method based on scene dependent causal influence analysis, and belongs to the technical field of automatic driving vehicle auxiliary lane changing, and the method comprises the steps: 1, collecting data based on an electronic rearview mirror of a vehicle, defining input parameters, calculating a basic safe distance, and calculating a dynamic safe distance; 2, generating a safe lane changing model based on situation-dependent causal influence, and realizing safe and efficient vehicle lane changing decision by the safe lane changing model through integrating situation awareness, an attention mechanism and a causal reasoning technology; and step 3, optimizing based on a particle swarm optimization lane changing strategy, constructing an objective function, and modeling a lane changing decision into a multi-objective optimization problem in order to optimize a decision-making vehicle lane changing process. The invention aims to construct a situation-dependent causal relationship model based on the data collected by the electronic rearview mirror, which is used for analyzing and predicting the safety of the lane changing behavior and providing auxiliary lane changing decision support for the automatic driving vehicle.

Description

Technical Field

[0001] This invention relates to the technical field of lane changing assistance for autonomous vehicles, and in particular to a safe lane changing method for vehicles based on context-dependent causal influence analysis. Background Technology

[0002] Lane changing is a complex maneuver during driving, requiring drivers to consider multiple factors. This involves not only assessing the current vehicle's status (such as speed and distance) but also evaluating the dynamic information of surrounding vehicles in real time (such as the speed and position of other vehicles in adjacent lanes). Furthermore, lane changing can significantly impact other drivers in both the original and target lanes, increasing the risk of traffic conflicts or accidents. Therefore, to ensure lane changing safety, advanced technologies are needed to achieve real-time perception and analysis of the operational status of multiple vehicles.

[0003] As an advanced driver assistance device, electronic rearview mirrors can optimize collected visual data through image processing algorithms and transmit this information to the in-vehicle display screen, providing drivers with a clear and stable visual experience. Combining the real-time status of the target vehicle with information about the surrounding environment, electronic rearview mirrors can accurately perceive the movement of surrounding vehicles, thus providing drivers or autonomous driving systems with a scientific basis for lane-changing decisions.

[0004] Against this backdrop, this study aims to construct a context-dependent causal model based on data collected from electronic rearview mirrors to analyze and predict the safety of lane-changing behavior and provide assisted lane-changing decision support for autonomous vehicles. In real-world driving scenarios, the interaction between the target vehicle and other vehicles (especially the vehicles in front and behind in adjacent lanes) is dynamic and complex. To design a complete safe driving model, it is necessary to comprehensively consider factors such as the relative motion state of the target vehicle with the vehicles in front and behind, the road environment, and actual operational constraints. Summary of the Invention

[0005] This invention addresses the shortcomings of existing technologies by providing a vehicle lane-changing safety method based on scenario-dependent causal influence analysis. By using cameras and sensors installed on the vehicle's rearview mirror, the method monitors the surrounding environment in real time, helping drivers make more informed lane-changing decisions.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0007] A method for safe lane changing of vehicles based on context-dependent causal influence analysis.

[0008] Step 1: Collect data based on the vehicle's electronic rearview mirrors, define input parameters, calculate the basic safety distance and the dynamic safety distance;

[0009] The input parameters are defined as: target vehicle: speed acceleration reaction time Maximum acceleration Maximum braking acceleration ;

[0010] Vehicle ahead in adjacent lane: distance to the vehicle in front ,speed acceleration ;

[0011] Following vehicles in adjacent lanes: distance between vehicles ,speed acceleration ;

[0012] Road restrictions: Maximum speed limit ;

[0013] Step 2: Based on context-dependent causal influence, a safe lane-changing model is generated. The safe lane-changing model integrates context awareness, attention mechanisms, and causal reasoning techniques to achieve safe and efficient vehicle lane-changing decisions.

[0014] The overall framework can be divided into the following three main parts:

[0015] Context-aware layer: Collects real-time dynamic features related to lane changes;

[0016] Attention mechanism layer: dynamically adjusts the attention level of each feature, highlighting the factors that have the greatest impact on lane-changing decisions;

[0017] Counterfactual causal reasoning layer: Enhances risk prediction capabilities and assesses causal impacts under different scenarios;

[0018] Step 3: Based on particle swarm optimization, optimize the lane-changing strategy and construct the objective function. The lane-changing process of the vehicle considers safety factors, driving efficiency and control cost factors. In order to optimize the decision-making process of the vehicle lane-changing, the lane-changing decision is modeled as a multi-objective optimization problem.

[0019] Furthermore, in step 1, the basic safe distance is calculated as the distance the vehicle needs to travel from when the driver notices a problem to when it safely changes lanes. This distance is related to the operating status of the vehicle ahead in the adjacent lane, and its mathematical formula is described as follows: The basic safety distance was calculated, among which It is defined as an infinitesimal.

[0020] Furthermore, in step 1, the calculation of the dynamic safety distance takes into account the dynamic interaction between the target vehicle's safety distance, the speed of vehicles in adjacent lanes, and their acceleration / deceleration states during lane changing. To further quantify and analyze the dynamic safety distance, a dynamic adjustment coefficient is introduced as follows:

[0021]

[0022] in, Let represent the dynamic adjustment coefficients for the vehicle in front and the vehicle behind, respectively. Therefore, the dynamic safe distance between vehicles in adjacent lanes is determined as follows: The dynamic safety distance is calculated.

[0023] Furthermore, in step 2, the goal of the emotion perception layer is to collect real-time dynamic features directly related to lane changes. To eliminate dimensional differences and improve model convergence efficiency, the input state vector of this layer is defined according to standardization requirements as follows: .

[0024] Furthermore, in step 2, the goal of the attention mechanism layer is to dynamically adjust the level of attention given to each feature;

[0025] The linear transformation yields the query vector, key vector, and value vector as follows:

[0026]

[0027] in It is a learnable parameter matrix with 7 dimensions;

[0028] The attention score matrix is ​​calculated as follows: ,in, It is the feature dimension;

[0029] Solving the context vector By dynamically weighting and fusing input features, the system extracts the most critical feature combination for the current decision. Its function is analogous to the cognitive process of a human driver actively focusing on key risk points in complex road conditions. ,in, It includes the weighted key features: .

[0030] Furthermore, in step 2, the counterfactual causal reasoning layer is first constructed using a structural causal model (SCM), which formally expresses the causal relationships between variables through a causal graph, providing an interpretable logical framework for risk assessment. Its key path is described below:

[0031] Target vehicle main causal path:

[0032] Path of impact from the vehicle in front:

[0033] The path of the following vehicle:

[0034] Counterfactual intervention path:

[0035] in These represent the old and new dynamic safety distances, respectively. To quantify risk, These represent the acceleration adjustment amount, adjusted acceleration, and adjusted speed of the target vehicle, respectively, in response to counterfactual intervention.

[0036] The counterfactual intervention construction process involves simulating safety states under different decision-making conditions through virtual intervention. In this paper, the target vehicle acceleration is selected as the intervention object.

[0037] Original safety distance calculation: ,in This refers to the length of the vehicle body;

[0038] Counterfactual intervention: This part involves virtually modifying the acceleration of the target vehicle. By simulating the safety status under different driving strategies, the updated speed of the target vehicle can be obtained. ,in The time spent changing lanes;

[0039] Then, based on the formulas for basic safety distance and dynamic safety distance, the updated basic safety distance is obtained. Dynamic adjustment coefficients of front and rear vehicles Dynamic safe distance between vehicles ;

[0040] Therefore, based on the original safety distance calculation formula, the safety distance after intervention can be obtained as follows:

[0041] ;

[0042] Lane change risk quantitative assessment: Taking into account factors such as distance, speed, and acceleration, and combining normalization processing, a lane change risk quantitative assessment model is constructed as follows: ;

[0043] Among them, the definition These represent the differences between the distance between the vehicles in front and behind and the dynamic safe distance between the vehicles in front and behind; (Definition) The speed of the target vehicle is the speed difference between the speeds of the vehicles in front and behind it in the adjacent lane. These are weighting coefficients, obtained based on historical data and simulation tests.

[0044] Furthermore, in step 3, the objective function can be defined as:

[0045] ;

[0046] in, This represents the weighting coefficient of each factor. Adjustment amount for the target vehicle speed. Adjustment amount for the target vehicle's acceleration. The target vehicle's speed of movement;

[0047] The objective function shows that the objectives of lane changing for vehicles mainly include four aspects: overall risk, speed adjustment, acceleration adjustment, and abruptness.

[0048] In addition to the factors mentioned above, vehicles must also meet safety constraints during lane changes, as follows:

[0049] Safety distance constraints:

[0050] Speed ​​constraints: .

[0051] Furthermore, in step 3, considering the need for both rapid decision-making and safety-first optimization, the particle swarm optimization (PSO) algorithm is selected as the solution algorithm, as detailed below:

[0052] Particle encoding and initialization: Define each particle as representing a set of candidate solutions. The particle velocity is initialized to a random value, and the feasible region is defined as follows: ;

[0053] Fitness function design: The objective function is shown in the formula, and the optimization goal is to obtain... A penalty is introduced into it, with a penalty coefficient of 1. The formula is: ;

[0054] Particle update rules: This section mainly includes two parts, namely velocity update and position update;

[0055]

[0056] in, The inertia coefficient, For individual and group learning factors, It is a random number.

[0057] In summary, compared with the prior art, the beneficial effects of the above technical solution are:

[0058] 1. The present invention aims to construct a context-dependent causal relationship model based on data collected by electronic rearview mirrors, to analyze and predict the safety of lane-changing behavior, and to provide auxiliary lane-changing decision support for autonomous vehicles.

[0059] 2. The method of this invention employs multi-sensor fusion technology, including radar and cameras, to achieve comprehensive monitoring of the surrounding environment during lane changes. The patented algorithm processes the data acquired by the sensors, analyzing environmental information around the vehicle, including vehicles, pedestrians, and traffic signs. Sensors, data processing software, and control systems are integrated into the rearview mirror system to ensure accurate and timely feedback. Multiple operating interfaces, such as touch and voice control, are designed to allow drivers to easily view monitoring information. When the system detects a potential hazard, it alerts the driver through lights and sounds, providing real-time monitoring of the vehicle's surroundings and helping drivers make more informed lane-changing decisions.

[0060] 3. The method of the present invention is an intelligent assisted driving technology. By using cameras and sensors installed on the rearview mirror of the vehicle, it monitors the environmental conditions around the vehicle in real time, helping the driver to change lanes more safely and conveniently. This technology can detect the lane changing environment, blind spots, vehicles to the side and rear, pedestrians, etc., and provide visual and audio prompts to help the driver make more informed driving decisions.

[0061] 4. The method of the present invention can enhance driving safety. By monitoring the vehicle’s surrounding environment in real time, especially when changing lanes, blind spots or potential dangers can be detected in advance, reducing the occurrence of traffic accidents.

[0062] 5. The method of the present invention can improve driving convenience, making it easier for the driver to observe the situation behind the vehicle, especially in narrow spaces or complex traffic environments, which helps to improve driving efficiency and comfort.

[0063] 6. The method of the present invention can effectively prevent malicious lane cutting. When a vehicle changes lanes, the environmental monitoring system can detect illegal behaviors of other vehicles in a timely manner, such as malicious lane cutting, and help the driver to react quickly. Attached Figure Description

[0064] Figure 1 This is a schematic diagram of lane changing in an embodiment of the present invention;

[0065] Figure 2 This is a lane-changing cause-effect diagram in an embodiment of the present invention. Detailed Implementation

[0066] The principles and features of the present invention are described below with reference to all the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0067] This invention discloses a method for safe lane changing of vehicles based on scenario-dependent causal influence analysis.

[0068] Reference Figure 1 , Figure 2As shown, a method for safe lane changing of vehicles based on context-dependent causal influence analysis is presented.

[0069] Step 1: Collect data based on the vehicle's electronic rearview mirrors, define input parameters, calculate the basic safety distance and the dynamic safety distance;

[0070] The input parameters are defined as: target vehicle: speed acceleration reaction time Maximum acceleration Maximum braking acceleration ;

[0071] Vehicle ahead in adjacent lane: distance to the vehicle in front ,speed acceleration ;

[0072] Following vehicles in adjacent lanes: distance between vehicles ,speed acceleration ;

[0073] Road restrictions: Maximum speed limit .

[0074] The basic safe distance is defined as the distance the vehicle needs to travel from when the driver notices a problem to when it can safely change lanes. This distance is related to the driving status of the vehicle ahead in the adjacent lane, and its mathematical formula is described as follows: The basic safety distance was calculated, among which It is defined as an infinitesimal.

[0075] To calculate the dynamic safety distance, considering the dynamic interaction between the target vehicle's safety distance, the speeds of vehicles in adjacent lanes, and their acceleration / deceleration states during lane changes, a dynamic adjustment coefficient is introduced as follows to further quantify and analyze the dynamic safety distance:

[0076]

[0077] in, Let represent the dynamic adjustment coefficients for the vehicle in front and the vehicle behind, respectively. Therefore, the dynamic safe distance between vehicles in adjacent lanes is determined as follows: The dynamic safety distance is calculated.

[0078] Step 2: Based on context-dependent causal influence, a safe lane-changing model is generated. The safe lane-changing model integrates context awareness, attention mechanisms, and causal reasoning techniques to achieve safe and efficient vehicle lane-changing decisions.

[0079] The overall framework can be divided into the following three main parts:

[0080] Context-aware layer: Collects real-time dynamic features related to lane changes;

[0081] The goal of the emotion perception layer is to collect real-time dynamic features directly related to lane changes. To eliminate dimensional differences and improve model convergence efficiency, the input state vector of this layer is defined according to standardization requirements as follows: .

[0082] Attention mechanism layer: dynamically adjusts the attention level of each feature, highlighting the factors that have the greatest impact on lane-changing decisions;

[0083] The goal of the attention mechanism layer is to dynamically adjust the level of attention given to each feature;

[0084] The linear transformation yields the query vector, key vector, and value vector as follows:

[0085]

[0086] in It is a learnable parameter matrix with 7 dimensions;

[0087] The attention score matrix is ​​calculated as follows: ,in, It is the feature dimension;

[0088] Solving the context vector By dynamically weighting and fusing input features, the system extracts the most critical feature combination for the current decision. Its function is analogous to the cognitive process of a human driver actively focusing on key risk points in complex road conditions. ,in, It includes the weighted key features: The specific meanings are shown in the table below:

[0089]

[0090] Counterfactual causal reasoning layer: Enhances risk prediction capabilities and assesses causal impacts under different scenarios;

[0091] The counterfactual causal reasoning layer is first constructed using a structural causal model (SCM), which formally expresses the causal relationships between variables through a causal diagram, providing an interpretable logical framework for risk assessment. Its key path is described below:

[0092] Target vehicle main causal path:

[0093] Path of impact from the vehicle in front:

[0094] The path of the following vehicle:

[0095] Counterfactual intervention path:

[0096] in These represent the old and new dynamic safety distances, respectively. To quantify risk, These represent the acceleration adjustment amount, adjusted acceleration, and adjusted speed of the target vehicle, respectively, in response to counterfactual intervention.

[0097] The counterfactual intervention construction process involves simulating safety states under different decision-making conditions through virtual intervention. In this paper, the target vehicle acceleration is selected as the intervention object.

[0098] Original safety distance calculation: ,in This refers to the length of the vehicle body;

[0099] Counterfactual intervention: This part involves virtually modifying the acceleration of the target vehicle. This simulates the safety status under different driving strategies. Furthermore, it allows obtaining the updated speed of the target vehicle. ,in The time spent changing lanes;

[0100] Then, based on the formulas for basic safety distance and dynamic safety distance, the updated basic safety distance is obtained. Dynamic adjustment coefficients of front and rear vehicles Dynamic safe distance between vehicles ;

[0101] Therefore, based on the original safety distance calculation formula, the safety distance after intervention can be obtained as follows:

[0102] ;

[0103] Lane change risk quantitative assessment: Taking into account factors such as distance, speed, and acceleration, and combining normalization processing, a lane change risk quantitative assessment model is constructed as follows: ;

[0104] Among them, the definition These represent the differences between the distance between the vehicles in front and behind and the dynamic safe distance between the vehicles in front and behind; (Definition) The speed of the target vehicle is the speed difference between the speeds of the vehicles in front and behind it in the adjacent lane. These are weighting coefficients, obtained based on historical data and simulation tests.

[0105] Step 3: Based on particle swarm optimization, optimize the lane-changing strategy and construct the objective function. The lane-changing process of the vehicle considers safety factors, driving efficiency and control cost factors. In order to optimize the decision-making process of the vehicle lane-changing, the lane-changing decision is modeled as a multi-objective optimization problem.

[0106] The objective function can be defined as:

[0107] ;

[0108] in, This represents the weighting coefficient of each factor. Adjustment amount for the target vehicle speed. Adjustment amount for the target vehicle's acceleration. The target vehicle's speed of movement;

[0109] The objective function shows that the objectives of lane changing for vehicles mainly include four aspects: overall risk, speed adjustment, acceleration adjustment, and abruptness.

[0110] In addition to the factors mentioned above, vehicles must also meet safety constraints during lane changes, as follows:

[0111] Safety distance constraints:

[0112] Speed ​​constraints: .

[0113] To address the optimization requirements of both rapid decision-making and paramount safety, Particle Swarm Optimization (PSO) was selected as the solution algorithm. The details are as follows:

[0114] Particle encoding and initialization: Define each particle as representing a set of candidate solutions. The particle velocity is initialized to a random value, and the feasible region is defined as follows: ;

[0115] Fitness function design: The objective function is shown in the formula, and the optimization goal is to obtain... A penalty is introduced into it, with a penalty coefficient of 1. The formula is: ;

[0116] Particle update rules: This section mainly includes two parts, namely velocity update and position update;

[0117]

[0118] in, The inertia coefficient, For individual and group learning factors, These are random numbers, and their specific meanings are shown in the table below:

[0119]

[0120] The implementation principle of a vehicle safe lane-changing method based on scenario-dependent causal influence analysis in an embodiment of the present invention is as follows:

[0121] 1. Define input parameters:

[0122] Target vehicle: Speed acceleration reaction time Maximum acceleration Maximum braking acceleration ;

[0123] Vehicle ahead in adjacent lane: distance to the vehicle in front ,speed acceleration ;

[0124] Following vehicles in adjacent lanes: distance between vehicles ,speed acceleration ;

[0125] Road restrictions: Maximum speed limit .

[0126] 1.1 Calculate the basic safety distance

[0127] The basic safe distance is defined as the distance a driver needs to travel from noticing a problem to safely changing lanes. This distance is related to the driving status of the vehicle ahead in the adjacent lane, and its mathematical description is as follows:

[0128] (1)

[0129] 1.2 Calculate the dynamic safety distance

[0130] Considering the dynamic interaction between the target vehicle's safe distance and the speed and acceleration / deceleration states of vehicles in adjacent lanes during lane changing, a dynamic adjustment coefficient is first introduced to further quantify and analyze the dynamic safe distance:

[0131] (2)

[0132] in, These represent the dynamic adjustment coefficients for the preceding and following vehicles, respectively.

[0133] Therefore, the dynamic safe distance between vehicles in adjacent lanes can be determined as follows:

[0134] (3)

[0135] 2. A safe lane-changing model based on context-dependent causal effects

[0136] This model aims to achieve safe and efficient lane-changing decisions by integrating context awareness, attention mechanisms, and causal reasoning techniques. Its overall framework can be divided into three main parts: Context Awareness Layer: Collects real-time dynamic features related to lane changing. Attention Mechanism: Dynamically adjusts the level of attention given to each feature, highlighting the factors that have the greatest impact on lane-changing decisions. Counterfactual Causal Reasoning Layer: Enhances risk prediction capabilities and assesses the causal impact under different scenarios.

[0137] 2.1 Emotional Perception Layer

[0138] The goal of this layer is to collect real-time dynamic features directly related to lane changes. Based on the aforementioned introduction, to eliminate dimensional differences and improve model convergence efficiency, the input state vector of this layer is defined according to standardization requirements as follows:

[0139] (4)

[0140] 2.2 Attention Mechanism Layer

[0141] The goal of this layer is to dynamically adjust the level of attention given to each feature.

[0142] Linear transformation yields the query vector, key vector, and value vector.

[0143] (5)

[0144] in It is a learnable parameter matrix with a dimension of 7.

[0145] In this invention, the parameter matrix of the attention mechanism layer All are learnable linear transformation weight matrices, used to generate the query vector, key vector, and value vector, respectively; where each parameter... Indicates the input feature number 1 Dimension to output feature The mapping weights of the dimension are dynamically updated through backpropagation during model training to achieve adaptive adjustment of the degree of attention to input features, thereby improving the accuracy of feature representation and decision-making in the electronic rearview mirror-assisted lane changing process.

[0146] Calculate the attention score matrix

[0147] (6)

[0148] in, It is the feature dimension.

[0149] Solving the context vector

[0150] By dynamically weighting and fusing input features, the most critical feature combination for the current decision is extracted. Its function is analogous to the cognitive process of a human driver proactively focusing on key risk points in complex road conditions.

[0151] (7)

[0152] in, Includes weighted key features ( ),

[0153] 2.3 Counterfactual Causal Reasoning Layer

[0154] Structural Causal Model (SCM) Construction

[0155] like Figure 2 As shown, a cause-effect graph is used to formally express the causal relationships between variables, providing an interpretable logical framework for risk assessment. The key path is described below:

[0156] Target vehicle main causal path:

[0157] (8)

[0158] Path of impact from the vehicle in front:

[0159] (9)

[0160] The path of the following vehicle:

[0161] (10)

[0162] Counterfactual intervention path:

[0163] (11)

[0164] in These represent the old and new dynamic safety distances, respectively. To quantify risk, These represent the acceleration adjustment amount, the adjusted acceleration, and the adjusted speed of the target vehicle, respectively, in response to counterfactual intervention.

[0165] Counterfactual intervention construction

[0166] This step simulates the safety state under different decision-making conditions through virtual intervention. In this paper, the target vehicle acceleration is selected as the intervention object.

[0167] Original safety distance calculation:

[0168] (12)

[0169] in This refers to the vehicle's length.

[0170] Counterfactual intervention:

[0171] This part modifies the target vehicle's acceleration virtually. This simulates the safety status under different driving strategies. Furthermore, it allows obtaining the updated speed of the target vehicle. ,in The time spent changing lanes.

[0172] Then, the updated basic safety distance is obtained according to formulas (1), (2), and (3). Dynamic adjustment coefficients of front and rear vehicles Dynamic safe distance between vehicles .

[0173] Therefore, according to formula (12), the safe distance after intervention can be obtained as follows:

[0174] (13)

[0175] Lane change risk quantitative assessment:

[0176] Taking into account factors such as distance, speed, and acceleration, and combining them with normalization processing, a quantitative assessment model for lane-changing risk is constructed as follows:

[0177] (14)

[0178] Among them, the definition These represent the differences between the distance between the vehicles in front and behind and the dynamic safe distance between the vehicles in front and behind; (Definition) The speed of the target vehicle is the speed difference between the speeds of the vehicles in front and behind it in the adjacent lane. These are weighting coefficients, which can be obtained based on historical data and simulation tests.

[0179] 3. Lane-changing strategy optimization based on particle swarm optimization (PSO)

[0180] 3.1 Construction of the objective function

[0181] In addition to safety factors, the lane-changing process also needs to consider driving efficiency, control costs, and other factors. Therefore, to optimize the lane-changing decision-making process, the lane-changing decision is modeled as a multi-objective optimization problem. Thus, the objective function can be defined as follows:

[0182] (15)

[0183] in, This represents the weighting coefficient of each factor. Adjustment amount for the target vehicle speed. Adjustment amount for the target vehicle's acceleration. The target vehicle's speed.

[0184] Formula (15) shows that the objectives of vehicle lane changing mainly include four aspects: comprehensive risk, speed adjustment amount, acceleration adjustment amount, and swiftness.

[0185] In addition to the factors mentioned above, vehicles must also meet safety constraints during lane changes, as follows:

[0186] Safety distance constraints

[0187] (16)

[0188] Speed ​​constraints

[0189] (17)

[0190] 3.2 Particle Swarm Optimization Algorithm for Optimization

[0191] Considering the need for rapid decision-making and the paramount importance of safety in optimization, this paper selects particle swarm optimization (PSO) as the solution algorithm, as detailed below:

[0192] Particle encoding and initialization: Define each particle as representing a set of candidate solutions. The particle velocity is initialized to a random value, and the feasible region is defined as follows: .

[0193] Fitness function design: The objective function is shown in formula (15). The optimization goal is to obtain... A penalty is introduced into it, with a penalty coefficient of 1. :

[0194] (18)

[0195] Particle update rules: This part mainly includes two parts, namely velocity update and position update.

[0196] (19)

[0197] in, The inertia coefficient, For individual and group learning factors, It is a random number.

[0198] This invention aims to construct a context-dependent causal relationship model based on data collected by electronic rearview mirrors, to analyze and predict the safety of lane-changing behavior, and to provide auxiliary lane-changing decision support for autonomous vehicles.

[0199] This invention employs multi-sensor fusion technology, including radar and cameras, to achieve comprehensive monitoring of the surrounding environment during lane changes. The patented algorithm processes the data acquired by the sensors, analyzing environmental information around the vehicle, including vehicles, pedestrians, and traffic signs. Sensors, data processing software, and a control system are integrated into the rearview mirror system to ensure accurate and timely feedback. Multiple user interfaces, including touch and voice controls, are designed to allow drivers to easily view monitoring information. When the system detects a potential hazard, it alerts the driver through lights and sounds.

[0200] The method of this invention is also an intelligent driver assistance technology. By using cameras and sensors installed on the vehicle's rearview mirror, it monitors the surrounding environment in real time, making lane changing safer and more convenient for the driver. This technology can detect lane changing environment, blind spots, vehicles to the side and rear, pedestrians, etc., and provide visual and audible prompts to help the driver make more informed driving decisions.

[0201] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for safe lane changing of vehicles based on context-dependent causal influence analysis, characterized in that: Step 1: Collect data based on the vehicle's electronic rearview mirrors, define input parameters, calculate the basic safety distance and the dynamic safety distance; The input parameters are defined as: target vehicle: speed acceleration reaction time Maximum acceleration Maximum braking acceleration ; Vehicle ahead in adjacent lane: distance to the vehicle in front ,speed acceleration ; Following vehicles in adjacent lanes: distance between vehicles ,speed acceleration ; Road restrictions: Maximum speed limit ; Step 2: Based on context-dependent causal influence, a safe lane-changing model is generated. The safe lane-changing model integrates context awareness, attention mechanisms, and causal reasoning techniques to achieve safe and efficient vehicle lane-changing decisions. The overall framework can be divided into the following three main parts: Context-aware layer: Collects real-time dynamic features related to lane changes; Attention mechanism layer: dynamically adjusts the attention level of each feature, highlighting the factors that have the greatest impact on lane-changing decisions; Counterfactual causal reasoning layer: Enhances risk prediction capabilities and assesses causal impacts under different scenarios; Step 3: Based on particle swarm optimization, optimize the lane-changing strategy and construct the objective function. The lane-changing process of the vehicle considers safety factors, driving efficiency and control cost factors. In order to optimize the decision-making process of the vehicle lane-changing, the lane-changing decision is modeled as a multi-objective optimization problem.

2. The vehicle safe lane-changing method based on scenario-dependent causal influence analysis according to claim 1, characterized in that: In step 1, the basic safe distance is calculated as the distance the vehicle needs to travel from when the driver notices a problem to when it safely changes lanes. This distance is related to the driving status of the vehicle ahead in the adjacent lane, and its mathematical formula is described as follows: The basic safety distance was calculated, among which It is defined as an infinitesimal.

3. The vehicle safe lane-changing method based on scenario-dependent causal influence analysis according to claim 2, characterized in that: In step 1, the dynamic safety distance is calculated by considering the dynamic interaction between the target vehicle's safety distance, the speeds of vehicles in adjacent lanes, and their acceleration / deceleration states during lane changing. To further quantify and analyze the dynamic safety distance, a dynamic adjustment coefficient is introduced as follows: in, Let represent the dynamic adjustment coefficients for the vehicle in front and the vehicle behind, respectively. Therefore, the dynamic safe distance between vehicles in adjacent lanes is determined as follows: The dynamic safety distance is calculated.

4. The vehicle safe lane-changing method based on scenario-dependent causal influence analysis according to claim 1, characterized in that: In step 2, the goal of the emotion perception layer is to collect real-time dynamic features directly related to lane changes. To eliminate dimensional differences and improve model convergence efficiency, the input state vector of this layer is defined according to standardization requirements as follows: .

5. A vehicle safe lane-changing method based on scenario-dependent causal influence analysis according to claim 4, characterized in that: In step 2, the goal of the attention mechanism layer is to dynamically adjust the level of attention given to each feature; The linear transformation yields the query vector, key vector, and value vector as follows: in It is a learnable parameter matrix with 7 dimensions; The attention score matrix is ​​calculated as follows: ,in, It is the feature dimension; Solving the context vector By dynamically weighting and fusing input features, the system extracts the most critical feature combination for the current decision. Its function is analogous to the cognitive process of a human driver actively focusing on key risk points in complex road conditions. ,in, It includes the weighted key features: .

6. The vehicle safe lane-changing method based on scenario-dependent causal influence analysis according to claim 3, characterized in that: In step 2, the counterfactual causal reasoning layer is first constructed using a structural causal model (SCM). This model formally expresses the causal relationships between variables through a causal graph, providing an interpretable logical framework for risk assessment. Its key path is described below: Target vehicle main causal path: Path of impact from the vehicle in front: Path of impact from following vehicles: Counterfactual intervention path: in These represent the old and new dynamic safety distances, respectively. To quantify risk, These represent the acceleration adjustment amount, adjusted acceleration, and adjusted speed of the target vehicle, respectively, in response to counterfactual intervention. The counterfactual intervention construction process involves simulating safety states under different decision-making conditions through virtual intervention. In this paper, the target vehicle acceleration is selected as the intervention object. Original safety distance calculation: ,in This refers to the length of the vehicle body; Counterfactual intervention: This part involves virtually modifying the acceleration of the target vehicle. By simulating the safety status under different driving strategies, the updated speed of the target vehicle can be obtained. ,in The time spent changing lanes; Then, based on the formulas for basic safety distance and dynamic safety distance, the updated basic safety distance is obtained. Dynamic adjustment coefficients of front and rear vehicles Dynamic safe distance between vehicles ; Therefore, based on the original safety distance calculation formula, the safety distance after intervention can be obtained as follows: ； Lane change risk quantitative assessment: Taking into account factors such as distance, speed, and acceleration, and combining normalization processing, a lane change risk quantitative assessment model is constructed as follows: ; Among them, the definition These represent the differences between the distance between the vehicles in front and behind and the dynamic safe distance between the vehicles in front and behind; (Definition) The speed of the target vehicle is the speed difference between the speeds of the vehicles in front and behind it in the adjacent lane. These are weighting coefficients, obtained based on historical data and simulation tests.

7. A vehicle safe lane-changing method based on scenario-dependent causal influence analysis according to claim 6, characterized in that: In step 3, the objective function can be defined as: ； in, This represents the weighting coefficient of each factor. Adjustment amount for the target vehicle speed. Adjustment amount for the target vehicle's acceleration. The target vehicle's speed of movement; The objective function shows that the objectives of lane changing for vehicles mainly include four aspects: overall risk, speed adjustment, acceleration adjustment, and abruptness. In addition to the factors mentioned above, vehicles must also meet safety constraints during lane changes, as follows: Safety distance constraints: Speed ​​constraints: .

8. A vehicle safe lane-changing method based on scenario-dependent causal influence analysis according to claim 7, characterized in that: In step 3, considering the need for fast decision-making and safety-first optimization, the particle swarm optimization (PSO) algorithm is selected as the solution algorithm. The specific details are as follows: Particle encoding and initialization: Define each particle as representing a set of candidate solutions. The particle velocity is initialized to a random value, and the feasible region is defined as follows: ; Fitness function design: The objective function is shown in the formula, and the optimization goal is to obtain... A penalty is introduced into it, with a penalty coefficient of 1. The formula is: ; Particle update rules: This section mainly includes two parts, namely velocity update and position update; in, The inertia coefficient, For individual and group learning factors, It is a random number.

Images