A flexible following method for an autonomous vehicle in a multi-lane complex following scenario

Abstract

This invention discloses a flexible car-following method for autonomous vehicles in complex multi-lane car-following scenarios, comprising: 1. collecting traffic information of surrounding vehicles of interest in the multi-lane scenario; 2. calculating the longitudinal car-following intention for each target of interest; 3. constructing a longitudinal car-following intention conversion mechanism between multiple targets of interest; 4. calculating the acceleration, speed, and position of the autonomous vehicle; 5. repeating the above steps and updating in real time to achieve safe and stable car-following in complex multi-lane car-following scenarios. This invention can improve the response agility and decision-making accuracy of autonomous vehicles in complex car-following scenarios such as dynamic switching of vehicles ahead in multi-lane traffic and frequent lane changes, effectively avoiding sudden acceleration changes and incorrect target switching problems, and enhancing the continuity, smoothness, and safety of longitudinal car-following behavior.

Description

Technical Field

[0001] This invention belongs to the field of autonomous driving and intelligent transportation, specifically referring to a method for autonomous vehicles in complex multi-lane car-following scenarios that comprehensively considers driver attention shift, multi-target intent modeling and nonlinear function smooth switching mechanism. More specifically, it is a flexible car-following method for autonomous vehicles in complex multi-lane car-following scenarios. Background Technology

[0002] As autonomous driving technology continues to evolve, intelligent vehicles face increasingly complex traffic scenarios in real-world road environments, especially in multi-lane highway driving or urban congestion. Vehicle interactions not only involve longitudinal car-following within lanes but also frequent lateral lane changes and complex interactions in weaving zones. Achieving more intelligent, safe, and comfortable longitudinal car-following control under complex conditions has become an important direction for research in car-following models.

[0003] Most existing car-following models are based on a single-lane environment, assuming vehicles move in a convoy, and primarily focus on the longitudinal distance relationship and speed adjustment with the vehicle in front in the current lane. These models, including classic OVM, IDM, and FVD models, possess good intra-lane car-following stability and theoretical completeness, but generally neglect the problem of car-following target switching caused by lateral disturbances such as vehicles changing into or temporarily encroaching on adjacent lanes, or vehicles changing out of the current lane. Faced with these typical "cross-lane car-following" scenarios, these traditional models lack multi-target perception and dynamic attention mechanisms, often exhibiting problems such as sudden acceleration changes, control lag, or untimely target switching, making it difficult to meet the high requirements of smoothness and safety for autonomous driving systems.

[0004] Current research has attempted to incorporate delay effects, driver psychological characteristics, and perspective responses into car-following models to improve their ability to depict dynamic traffic flow and to consider the impact of adjacent vehicles in two-lane or even multi-lane traffic. However, these improvements mostly focus on enhancing overall traffic flow stability and rarely address the modeling of driver attention shifts and dynamic switching of car-following targets under "cross-car-following" scenarios. Especially in typical scenarios such as lateral vehicles attempting to merge into the current lane, the vehicle in front suddenly changing lanes, and both vehicles switching lanes simultaneously, existing models often fail to accurately simulate the driver's attention allocation process and lack a reasonable assessment mechanism for potential lateral interference targets, thus affecting the timeliness and rationality of car-following decisions.

[0005] Therefore, there is an urgent need for an intelligent car-following control method that can adapt to complex multi-lane car-following environments. This method would enable autonomous vehicles to accurately perceive the dynamic behavior of multiple targets ahead and around them when facing typical cross-interference scenarios such as frequent lateral lane changes, vehicles leaving the lane ahead, and vehicles merging from the lane next to them, and to achieve human-like attention shifting and response adjustment mechanisms. Such models are not only crucial for improving the safety and comfort of individual vehicles, but will also provide core support for building a more stable and efficient multi-lane intelligent transportation system. Therefore, developing flexible car-following mechanisms for multi-target dynamic interaction scenarios has become a key step for intelligent driving systems to achieve higher-level perception and decision-making capabilities, possessing significant engineering value and promising practical applications. Summary of the Invention

[0006] This invention addresses the shortcomings of existing car-following models in complex multi-lane car-following scenarios, such as insufficient response capability, rigid target switching, and inadequate perception of lateral interfering vehicles. It proposes a flexible car-following method for autonomous vehicles in complex multi-lane car-following scenarios, aiming to more accurately identify the dynamic behavior of surrounding vehicles when the vehicle performs longitudinal car-following tasks and reasonably simulate the attention shift process of human drivers under cross disturbances, thereby improving the safety, comfort, and driving stability of autonomous vehicles in cross-car-following situations.

[0007] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: The present invention provides a flexible car-following method for autonomous vehicles in complex multi-lane car-following scenarios. The method is characterized by its application in multi-lane scenarios with three or more lanes traveling in the same direction, and the autonomous vehicle is equipped with a real-time environmental perception and V2X communication module. The method includes the following steps: Step 1. Using an autonomous vehicle n The origin O is the centroid coordinate corresponding to the initial position, and the direction of travel along the lane is set as the longitudinal X-axis; the direction of travel perpendicular to the lane is set as the transverse Y-axis; thus constructing a local road coordinate system O-XY. Define autonomous vehicles n For the target vehicle, n The vehicle in front of the left adjacent lane l、n The vehicle in front of the right adjacent lane r、n The vehicle in front of me in the same lane n +1, the vehicle in front n +1 to the front car n Any one of the vehicles in +2 is marked as a vehicle to watch. j , D indicates the vehicle set to focus on; Step 2. Collect data on target vehicles in a multi-lane scenario at time t. n and vehicles to watch j Traffic information, and calculate the target vehicle at time t.n With regard to vehicles j lateral distance ; Step 3. Based on the traffic information at time t, construct the target vehicle at time t. n For vehicles under surveillance j The resulting longitudinal following intention ; Step 4. Select a unique primary vehicle of interest from D. i The remaining vehicles of interest are designated as potential vehicles of interest, and any potential vehicle of interest is recorded as... k ; Step 5. Based on the target vehicle at time t n For the main vehicle of concern i The dominant and driving intention and records of potential vehicles k Potential car-following intentions Construct a car-following intention conversion mechanism for lateral multi-vehicle interaction at time t, thereby obtaining autonomous vehicle... n acceleration at time t ; Step 6. Based on Calculate autonomous vehicles n velocity at time t+1 and location information ; Step 7. After assigning t+1 to t, return to step 1 and execute sequentially to achieve flexible car following for autonomous vehicles in complex multi-lane car following scenarios.

[0008] The characteristic of the flexible car-following method for autonomous vehicles in complex multi-lane car-following scenarios described in this invention is that the traffic information in step 2 includes: vehicle... n Position coordinates at time t and speed Pay attention to vehicles j Position coordinates at time t and speed Therefore, the target vehicle at time t can be calculated using equation (1). n With regard to vehicles j Lateral distance between the centers of gravity of the two vehicles ; (1) In equation (1), For the width of autonomous vehicles, For the width of the lane, For autonomous vehicles With regard to vehicles The lateral distance between the centers of gravity of the two vehicles.

[0009] Furthermore, in step 3, equation (2) is used to construct... :

[0010] (2) In equation (2), This is the maximum permitted speed for the current road segment. for tanh The shape parameter of the function, Indicates the target vehicle at time t n With regard to vehicles j The longitudinal distance between the centers of gravity of the two vehicles Indicates the target vehicle n With regard to vehicles j The expected following distance under safe driving conditions at time t, and we have: (3) In equation (3), For minimum safe distance, To determine the desired following distance, , These are the desired acceleration and desired deceleration parameters.

[0011] Furthermore, in step 4, the primary vehicle of interest is selected using equation (6). i , (6) In equation (6), express Time vehicle With the car in front The lateral distance between the centers of gravity of the two vehicles express Time vehicle With the car in front The lateral distance between the centers of gravity of the two vehicles; Let represent the Heaviside function, and we have: (7).

[0012] Furthermore, step 5 includes: Step 5.1. Determine from equation (8) and Flexible transfer function between ; (8) In equation (8), The parameter is an adjustable coefficient. Let represent the polarity factor, and we have: (9) Step 5.2. Obtain the two-dimensional optimized velocity function at time t from equation (10). ; (10) Step 5.3. Calculate the autonomous vehicle using equation (11). n acceleration at time t ; (11) In equation (11), express tanh The undetermined shape parameters of the function, This indicates the maximum acceleration.

[0013] Furthermore, step 6 includes: Step 6.1. Calculate the autonomous vehicle using equation (12). n velocity at time t+1 ; (12) In equation (12), This is a time step from time t to time t+1; Step 6.2. Calculate the autonomous vehicle using equation (13). n Location information at time t+1 ; (13) The present invention provides an electronic device, including a memory and a processor, characterized in that the memory is used to store a program supporting the processor in performing the method described therein, and the processor is configured to execute the program stored in the memory.

[0014] The present invention discloses a computer-readable storage medium storing a computer program, characterized in that the computer program is executed by a processor to perform the steps of the method described thereon.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Possesses adaptive sensing capability for complex lateral disturbances: This invention introduces a flexible transfer function and a dynamic adjustment mechanism for lateral spacing to construct a smooth switching model between dominant and potential car-following intentions. It can dynamically identify interference sources and reasonably adjust car-following strategies in complex traffic scenarios such as vehicles merging into adjacent lanes, short-term encroachment, and vehicles changing lanes in front, thereby improving the perception and response capabilities and robustness of autonomous vehicles to lateral interference behaviors.

[0016] 2. Depict the driver's attention shifting mechanism: This invention simulates the attention adjustment process of human drivers in cross-traffic scenarios, and integrates... sign Functions and tanh The function composition structure enables a natural transition of the car-following object at different stages, and can dynamically adjust the degree of attention to the car-following of the preceding vehicle, merging vehicle, or vehicle before that, effectively reducing sudden acceleration and deceleration and discontinuous acceleration control problems, thus improving the humanization of driving behavior and traffic comfort.

[0017] 3. Optimize the decision-making mechanism to balance speed response and safety: This invention selects the minimum value among multiple flexible following intentions as a two-dimensional optimized speed function to realize a priority response mechanism for high-risk, low-speed targets. This reflects the driver's tendency to be stable and conservative in a multi-target environment, helps to avoid drastic acceleration changes caused by rapid target switching or lateral insertion, and ensures traffic safety and the continuity of vehicle control.

[0018] 4. Possesses good model stability and engineering applicability: This invention extends the lateral perception capability of the traditional OVM model, forming a flexible car-following model framework for autonomous vehicles in complex multi-lane car-following scenarios, balancing structural simplicity and computational efficiency. The model theoretically possesses good linear stability. Validated on the NGSIM and Citysim datasets, the model significantly outperforms the traditional OVM and improved FVD models in terms of trajectory accuracy, fuel efficiency, and control smoothness, demonstrating feasibility and widespread application value for practical road deployment. Attached Figure Description

[0019] Figure 1 This is a simplified flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the overall process of the method of the present invention. Figure 3 This is a scene diagram of a multi-lane, multi-target moving scenario; Figure 4 A schematic diagram is constructed to illustrate the intention of longitudinal following. Figure 5 This is a schematic diagram of the working mechanism of the flexible transfer function. Detailed Implementation

[0020] The method of the present invention will now be described in detail with reference to the accompanying drawings.

[0021] In this embodiment, the "multi-lane" in a flexible following method for autonomous vehicles in complex multi-lane following scenarios refers to a road environment with three or more lanes traveling in the same direction and allowing vehicles to change lanes between adjacent lanes. The "complex following scenario" refers to situations where, during longitudinal following, the autonomous vehicle needs to interact with surrounding vehicles to achieve safe and stable longitudinal following behavior, especially when these vehicles have lateral lane-changing capabilities, such as vehicles changing lanes, changing lanes out, briefly encroaching on lanes, multiple vehicles merging into or simultaneously changing lanes out of the autonomous vehicle's lane. The autonomous vehicle travels along the lane centerline. Figure 1 and Figure 2 As shown, the flexible vehicle following method includes the following steps: Step 1. Using an autonomous vehicle n The initial position of the vehicle is defined by the centroid coordinates of the origin O, and the direction of travel along the lane is defined as the positive longitudinal X-axis; the direction of travel perpendicular to the lane and pointing to the left of the vehicle is defined as the positive transverse Y-axis; thus, a local road coordinate system O-XY is constructed. like Figure 3 As shown, an autonomous vehicle is defined. n For the target vehicle, n The vehicle in front of the left adjacent lane l、n The vehicle in front of the right adjacent lane r、n The vehicle in front of me in the same lane n +1, the vehicle in front n +1 to the front car n Any one of the vehicles in +2 is marked as a vehicle to watch. j , D indicates the focus on vehicle sets.

[0022] Step 2. Collect data on target vehicles in a multi-lane scenario at time t. n and vehicles to watch j Traffic information, including: vehicles n Position coordinates at time t and speed Pay attention to vehicles j Position coordinates at time t and speed Therefore, the target vehicle at time t can be calculated using equation (1). n With the vehicle j lateral distance ; (1) In equation (1), For the width of autonomous vehicles, For the width of the lane, For autonomous vehicles With regard to vehicles The lateral distance between the centers of gravity of the two vehicles.

[0023] Equation (1) draws on the idea of ​​the social force model to define an effective lateral distance calculation method based on physical boundary constraints. Its physical significance lies in: First, when the target vehicle... n With regard to vehicles j The actual lateral distance is less than or equal to the vehicle width. L At that time, the distance is forcibly assigned a value. L This not only reflects the property of vehicles as impenetrable physical entities, but also serves to identify target vehicles within this range as the primary vehicles of interest in subsequent processing; secondly, when the actual lateral distance between two vehicles is within the vehicle width... L With lane width W When the distance between two vehicles is greater than or equal to the lane width, the actual distance value is used directly to accurately depict the dynamic relative position of the two vehicles in the adjacent space, truthfully reflecting the real-time interaction impact in real traffic scenarios; finally, when the lateral distance between the two vehicles is greater than or equal to the lane width... W When that happens, its upper limit is truncated and fixed to . W This measure aims to define the effective boundary of lateral vehicle interaction, filtering out irrelevant vehicles separated by more than one lane. This setting fully aligns with real-world traffic patterns, namely, the interaction between vehicles in non-adjacent lanes and target vehicles. n The impact is extremely weak, thus effectively avoiding redundant horizontal interaction calculations.

[0024] Step 3. Based on the traffic information collected in Step 1 at time t, construct the target vehicle at time t using Equation (2). n For vehicles under surveillance j The resulting longitudinal following intention ; (2) In equation (2), This is the maximum permitted speed for the current road segment. for tanh The shape parameter of the function, Indicates the target vehicle n With regard to vehicles j The longitudinal distance between the centers of gravity of the two vehicles Indicates the target vehicle n With regard to vehicles j The expected following distance under safe driving conditions at time t, and we have: (3) In equation (3), The minimum safe distance is typically set between 4 and 10 meters (m). The expected following distance is expressed in seconds (s). , These are the desired acceleration and desired deceleration parameters, in meters (m). 2 / s. These parameters can be obtained from real traffic datasets such as NGSIM and Citysim using calibration algorithms such as genetic algorithms or Bayesian algorithms.

[0025] The parameters can be calibrated using equations (4) and (5). ,in, To approximate a constant to 1, consider tanh The asymptotic properties of the parameters are studied. The value range is defined as (0.8, 0.9999). Model stability analysis and numerical simulations have verified that this value range does not affect the model's stability or cause floating-point precision or numerical stability issues. When the longitudinal distance between the autonomous vehicle and the vehicle in front reaches... At this point, the longitudinal following intention obtained should be equal to the maximum permitted speed of the vehicle in the current road segment. . The maximum speed limit of the road segment can be obtained in advance through the V2X module of an autonomous vehicle, and It can be calibrated using real traffic datasets such as NGSIM and Citysim, and calibration algorithms such as genetic algorithms or Bayesian algorithms.

[0026] (4) (5) like Figure 4 As shown in the diagram, the process of constructing the longitudinal car-following intention, and the longitudinal car-following intention itself, are illustrated. by The process of change in impact. When starting from a standstill, only the minimum safe distance exists. s 0 At this time, the conditions for passage are not met, and the intention is to follow the car longitudinally. The value is 0; as the difference between the longitudinal distance and the desired distance increases... Gradually increasing in size, with a longitudinal following intention The value also gradually increases; when the difference (including in the case of a standstill start)... achieve At this time, the longitudinal following intention is obtained. It should be equal to the maximum speed allowed for vehicles on the current road segment. .

[0027] The implicit approximation used in the derivation of equation (4) aims to establish a physically meaningful finite boundary for the vehicle's "maximum following intention". This approximation is essentially derived from... tanh The asymptotic property of the function (i.e., approaching 1). Compared with traditional empirical parameter setting methods, it has the advantages of physical interpretability and improved calibration efficiency.

[0028] Furthermore, this approximation does not introduce structural errors into the model. Although this parameter is relative to... tanh It is an approximate representation of the function's extrema, but it does not prevent the vehicle from reaching the maximum desired speed defined by the system. The core reason is that this approximation only affects the dynamic transition interval ( s 0 arrive The model exhibits the "shape" of the S-curve (i.e., gradient change) without altering the fundamental dynamic structure of the model. From an asymptotic perspective, as the interaction distance approaches infinity (i.e., free flow state), the terminal velocity of the model remains highly consistent with the maximum velocity limits of classical OVM, FVD, and IDM models.

[0029] Step 4. Select a unique primary vehicle of interest from D. i The remaining vehicles of interest are designated as potential vehicles of interest, and any potential vehicle of interest is recorded as... k In step 4, the main vehicle of interest is selected using equation (6). i and potential vehicles of interest k： (6) In equation (6), express Time vehicle With the car in front lateral distance, express Time vehicle With the car in front The horizontal distance; Let represent the Heaviside function, and we have: (7) Step 5. Based on the target vehicle at time t n For the main vehicle of concern i The dominant and driving intention and records of potential vehicles k Potential car-following intentions Construct a car-following intention conversion mechanism for lateral multi-vehicle interaction at time t, thereby obtaining autonomous vehicle... n acceleration at time t .

[0030] Step 5.1. Determine from equation (8) and Flexible transfer function between : (8) In equation (8), The parameter is an adjustable coefficient, and its range of values ​​is defined in this study as follows: Let represent the polarity factor, and we have: (9) Combination Figure 5 This paper provides a detailed explanation of the longitudinal following intention flexible conversion mechanism and its technical effects in embodiments of the present invention. In multi-lane weaving scenarios, the autonomous vehicle (target vehicle) n The longitudinal control of the vehicle is highly susceptible to interference from the lateral movement of adjacent vehicles. To address the issues of sudden speed changes and control oscillations caused by instantaneous switching of the target vehicle in traditional models, this invention introduces a flexible transfer function. .like Figure 5 As shown, this function will target the vehicle as the primary vehicle of interest. i With potential vehicles of interest k The discrete longitudinal car-following intention switching is transformed into a process of continuous and smooth transition with lateral spacing. The specific mechanism is as follows: (1) For scenarios where the vehicle in front changes into the adjacent lane (e.g.) Figure 5 As shown in (a), for example, the vehicle in front in the left adjacent lane. l Replace the target vehicle n (Current lane) When potential attention vehicles in the left adjacent lane l Showing intent to change lanes and move toward the target vehicle n As the vehicles approached each other in their respective lanes, with the two vehicles moving laterally... The target vehicle gradually decreases. n The longitudinal following intention of the vehicle ahead follows an S-shaped curve, smoothly decreasing from a higher initial longitudinal following intention to a lower longitudinal following intention to adapt to the newly inserted target. This technical feature enables the autonomous driving system to intervene in advance based on lateral changes in space, achieving smooth deceleration and yielding, effectively avoiding emergency braking triggered by the sudden appearance of the following target, and improving driving safety and passenger comfort.

[0031] (2) Regarding the vehicle in front in the current lane n +1 Change the scene (e.g.) Figure 5 As shown in (b) above, for example, the vehicle in front in the current lane. n+ 1. Replace the target vehicle n Lane in question): The vehicle of primary concern in the current lane n +1 Drive out of the target vehicle n When in the lane, with the lateral spacing The gradual increase in the current lane, the vehicle in front n +1 target vehicle nThe longitudinal following constraint is gradually released, which is reflected in the smooth rise from the original lower longitudinal following intention to adapting to the new target (the preceding vehicle). n +1 to the front car n +2) Higher longitudinal following intent. This technical feature allows the target vehicle to smoothly reconstruct its following intent, naturally transitioning to following or free-cruising states towards targets further ahead.

[0032] (3) Shape parameters c Regulatory effect: Furthermore, Figure 5 The multiple evolution curves in the figure reflect the shape parameters c The moderating effect on oversensitivity. This is achieved by configuring different... c The system can flexibly adjust the smoothness of the speed transition by taking values ​​(e.g., from 0.8 to 0.9999). Smaller values... c A value close to 1 corresponds to a smoother, earlier transition response; while a value close to 1 corresponds to a more gradual, earlier transition response. c The value then manifests as a later and faster speed state switch occurring near the critical lateral distance.

[0033] Step 5.2. Obtain the two-dimensional optimized velocity function at time t from equation (10). ; (10) By selecting the minimum value among multiple flexible following intentions as the two-dimensional optimized speed function, a priority response mechanism for high-risk, low-speed targets is realized. This reflects the driver's tendency to be stable and conservative in a multi-target environment, which helps to avoid drastic acceleration changes caused by rapid target switching or lateral insertion, thus ensuring traffic safety and the continuity of vehicle control.

[0034] Step 5.3. Calculate the autonomous vehicle using equation (11). n acceleration at time t ; (11) In equation (11), express tanh The derivation of the undetermined shape parameters of the function is similar to that of Equation 4-5. Compared with traditional empirical parameter setting methods, this approach has the advantages of physical interpretability and improved calibration efficiency. Similarly, this approximation does not introduce structural errors into the model. The maximum acceleration can be selected within the range of (2,4). This parameter can be determined based on real traffic datasets such as NGSIM and CitySim, using parameter calibration methods such as genetic algorithms or Bayesian optimization, thereby improving the human-likeness of the model.

[0035] Step 6. Based on Calculate autonomous vehicles n velocity at time t+1 and location information ; Step 6.1. Calculate the autonomous vehicle using equation (12). n velocity at time t+1 ; (12) In equation (12), This is a time step from time t to time t+1.

[0036] Step 6.2. Calculate the autonomous vehicle using equation (13). n Location information at time t+1 ; (13) autonomous vehicles n The lateral position is always 0, which is determined by the characteristics of this algorithm. This algorithm belongs to lane following or adaptive cruise control algorithms, and is explained by establishing a coordinate system at the initial moment. Specifically, it is used for autonomous vehicles. n The origin O is the centroid coordinate corresponding to the initial position, and the direction of travel along the lane is the positive direction of the longitudinal X-axis; the positive direction of the lateral Y-axis is the direction perpendicular to the direction of travel along the lane and pointing to the left of the vehicle; thus, a local road coordinate system O-XY is constructed.

[0037] Step 7. After assigning t+1 to t, return to step 1 and execute sequentially to achieve flexible car following for autonomous vehicles in complex multi-lane car following scenarios.

[0038] In this embodiment, an electronic device includes a memory and a processor. The memory stores a program that supports the processor in executing the above-described method, and the processor is configured to execute the program stored in the memory.

[0039] In this embodiment, a computer-readable storage medium stores a computer program, which is executed by a processor to perform the steps of the above method.

Claims

1. A flexible car-following method for autonomous vehicles in complex multi-lane car-following scenarios, characterized in that, This method is applied to multi-lane scenarios with three or more lanes traveling in the same direction, and the autonomous vehicle is equipped with a real-time environmental perception and V2X communication module. The method includes the following steps: Step 1. Using an autonomous vehicle n The initial position of the vehicle is defined by the centroid coordinates of the origin O, and the direction of travel along the lane is defined as the positive longitudinal X-axis; the direction of travel perpendicular to the lane and pointing to the left of the vehicle is defined as the positive transverse Y-axis; thus, a local road coordinate system O-XY is constructed. Define autonomous vehicles n For the target vehicle, n The vehicle in front of the left adjacent lane l、n The vehicle in front of the right adjacent lane r、n The vehicle in front of me in the same lane n +1, the vehicle in front n +1 to the front car n Any one of the vehicles in +2 is marked as a vehicle to watch. j ,in D indicates the vehicle set to focus on; Step 2. Collect data on target vehicles in a multi-lane scenario at time t. n and vehicles to watch j Traffic information, and calculate the target vehicle at time t. n With regard to vehicles j lateral distance ; Step 3. Based on the traffic information at time t, construct the target vehicle at time t. n For vehicles under surveillance j The resulting longitudinal following intention ; Step 4. Select a unique primary vehicle of interest from D. i The remaining vehicles of interest are designated as potential vehicles of interest, and any potential vehicle of interest is recorded as... k ; Step 5. Based on the target vehicle at time t n For the main vehicle of concern i The dominant and driving intention and records of potential vehicles k Potential car-following intentions Construct a car-following intention conversion mechanism for lateral multi-vehicle interaction at time t, thereby obtaining autonomous vehicle... n acceleration at time t ; Step 6. Based on Calculate autonomous vehicles n velocity at time t+1 and location information ; Step 7. After assigning t+1 to t, return to step 1 and execute sequentially to achieve flexible car following for autonomous vehicles in complex multi-lane car following scenarios.

2. The flexible following method for autonomous vehicles in complex multi-lane following scenarios as described in claim 1, characterized in that, The traffic information in step 2 includes: vehicles n Position coordinates at time t and speed Pay attention to vehicles j Position coordinates at time t and speed Therefore, the target vehicle at time t can be calculated using equation (1). n With regard to vehicles j Lateral distance between the centers of gravity of the two vehicles ; (1) In equation (1), For the width of autonomous vehicles, For the width of the lane, For autonomous vehicles With regard to vehicles The lateral distance between the centers of gravity of the two vehicles.

3. A flexible following method for autonomous vehicles in complex multi-lane following scenarios as described in claim 2, characterized in that, In step 3, formula (2) is used to construct... : (2) In equation (2), This is the maximum permitted speed for the current road segment. for tanh The shape parameter of the function, For the target vehicle at time t n With regard to vehicles j The longitudinal distance between the centers of gravity of the two vehicles Indicates the target vehicle n With regard to vehicles j The expected following distance under safe driving conditions at time t, and we have: (3) In equation (3), For minimum safe distance, To determine the desired following distance, , These are the desired acceleration and desired deceleration parameters.

4. A flexible car-following method for autonomous vehicles in complex multi-lane car-following scenarios as described in claim 3, characterized in that, In step 4, the primary vehicle of interest is selected using equation (6). i： (6) In equation (6), express Time vehicle With the car in front The lateral distance between the centers of gravity of the two vehicles express Time vehicle With the car in front The lateral distance between the centers of gravity of the two vehicles; Let represent the Heaviside function, and we have: (7)。 5. A flexible car-following method for autonomous vehicles in complex multi-lane car-following scenarios as described in claim 4, characterized in that, Step 5 includes: Step 5.

1. Determine from equation (8) and Flexible transfer function between ; (8) In equation (8), The parameter is an adjustable coefficient. Let represent the polarity factor, and we have: (9) Step 5.

2. Obtain the two-dimensional optimized velocity function at time t from equation (10). ; (10) Step 5.

3. Calculate the autonomous vehicle using equation (11). n acceleration at time t ; (11) In equation (11), express tanh The undetermined shape parameters of the function, This indicates the maximum acceleration.

6. A flexible following method for autonomous vehicles in complex multi-lane following scenarios as described in claim 5, characterized in that, Step 6 includes: Step 6.

1. Calculate the autonomous vehicle using equation (12). n velocity at time t+1 ; (12) In equation (12), This is a time step from time t to time t+1; Step 6.

2. Calculate the autonomous vehicle using equation (13). n Location information at time t+1 ; (13)。 7. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store a program that supports a processor in executing the method of any one of claims 1-6, the processor being configured to execute the program stored in the memory.

8. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program is executed by a processor to perform the steps of the method according to any one of claims 1-6.

Images