Intelligent lane changing decision-making method for lane changing of automatic driving vehicle based on multi-dimensional evaluation

Abstract

The invention provides an automatic driving vehicle lane changing intelligent decision-making method based on multi-dimensional evaluation, and the method comprises the steps: obtaining the environment information of a passing road section of an automatic driving vehicle queue through advanced sensing equipment and technology, constructing an optimization decision-making space in an automatic driving lane changing vehicle target queue, and carrying out the optimization decision-making space; and determining a candidate set where the lane changing vehicle is inserted into the target queue. And then, selecting a sine curve and model prediction control (MPC) to carry out track planning and control on the lane-changing vehicle, and traversing in sequence to obtain tracks and control information of the lane-changing vehicle inserted into each candidate position. Then, calculating the influence of the lane changing vehicle at each candidate position on four dimension indexes of operation safety, stability, efficiency and smoothness of the original queue and the target queue of the lane changing vehicle, dynamically updating the weights of the four dimension indexes according to road environment information, and carrying out weighted calculation to obtain a comprehensive score; finally, after traversal is finished, the comprehensive scores are compared, the insertion position with the highest score and the trajectory planning and control decision of the insertion position are selected as the optimal lane changing strategy to complete lane changing, the queue safety is guaranteed, and the passing efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the fields of connected autonomous driving and vehicle platooning, and specifically to an intelligent lane-changing decision method for autonomous driving vehicles based on multi-dimensional evaluation. Background Art

[0002] As digitalization, Internet of Things (IoT), and intelligent technologies deeply penetrate the transportation sector, intelligent vehicle-road collaborative systems have become a key technology path for improving modern transportation efficiency. Based on the V2X communication technology architecture, this study addresses the issue of intelligent connected vehicle platooning control and proposes a platooning lane-changing decision model for collaborative driving. This model aims to achieve optimal allocation of spatiotemporal road resources by constructing a decision-making algorithm for vehicle lane-changing insertion positions. This research has important theoretical and practical significance for improving the operational stability and traffic efficiency of multi-agent transportation systems.

[0003] Regarding the selection of vehicle lane change insertion positions, Simon et al. considered the position and speed of the target queue vehicle to determine the lane change insertion position for a single vehicle. When the correct distance between the two vehicles in front and behind the lane change insertion position was reached, the single vehicle merged with the target queue. Zhao et al. considered the single vehicle and the target queue, using average time and latency requirements as the basis for vehicle insertion position. Heinovski et al. determined the lane change insertion position based on the vehicle's expected speed and position on the road, and initially chose to insert from the back of the queue. Sarker et al. selected the position in the queue closest to the single vehicle at the time of the lane change to quickly restore the queue. Li Pengfei et al. conducted experiments on designated lane change insertion positions under different emergency conditions, studying the switching control between different conditions and the impact of vehicle insertion after the vehicle has been inserted into the queue. Current research has mostly focused on vehicle insertion control methods and single-vehicle insertion, ignoring the selection of vehicle lane change insertion positions and lane changes between two queues. The indicators considered in position selection are relatively simple. This study provides an intelligent lane-changing decision-making method for autonomous vehicles based on multi-dimensional evaluation, and studies the optimal lane-changing insertion position selection and lane-changing strategy for vehicles between two queues to achieve safe and efficient lane-changing for two queues. Summary of the Invention

[0004] To address the shortcomings of existing lane-changing strategies, the present invention proposes an intelligent lane-changing decision-making method for autonomous driving vehicles based on multi-dimensional evaluation. This method considers the impact of the lane-changing vehicle on the original and target queues when inserting into the target queue in multiple dimensions and selects the optimal lane-changing insertion position. On the basis of ensuring safe lane changing and comfortable driving, the lane-changing vehicle can be inserted into the existing vehicle formation and form a new formation, thereby improving traffic efficiency.

[0005] The specific implementation steps of this method are as follows:

[0006] Step 1: Use sensing equipment and technology to obtain the road environment information of the original lane-changing vehicle queue (referred to as queue N) and the target lane-changing vehicle queue (referred to as queue M). The specific information is as follows:

[0007] Obtain data such as the number of vehicles in the two queues, the distance between vehicles, the speed of vehicles, and the motion status of vehicles requesting lane change. Determine the number of vehicles in the current queue M, the coordinate x of the kth vehicle's forward direction along the lane line. M,k and speed V M,k , the distance d between the kth car and the k+1th car M,k , the number of vehicles in the team N is n, and the coordinates of the k-th vehicle moving along the lane line are x N,k With speed V N,k , the distance d between the kth car and the k+1th car N,k , the lane-changing vehicle X is at position x in the convoy N. Assume that the length of all vehicles is L and the minimum safe distance between adjacent vehicles is d global_safe_min =2L. Set the absolute value of the speed fluctuation of a single vehicle to no more than v global_max =5km / h.

[0008] Step 2: Based on the data from step 1 regarding the environmental information of the two queues' traffic sections, and considering the impact of the speed difference and desired spacing between vehicles in the target queue on the lane change insertion position, an optimized decision space vector is constructed to determine the candidate set of positions for the lane-changing vehicle to insert into the convoy. The specific implementation steps for step 2 are as follows:

[0009] Step 21: Based on the data of the environmental information of the two queues passing through the road section in step 1, an optimized decision space vector Θ is constructed. The specific expression is as follows:

[0010] Θ={(k,Δv,d safe )}

[0011]

[0012] Where k is the position number of the vehicle X that is about to be inserted into the convoy M (not considering the position before or after the front vehicle). The number of the vehicle before position k is set to k, the number of the vehicle after position k is set to k+1, and so on. Δv is the speed adjustment value, which is used to simulate the maximum speed change of vehicle X during the lane change process. v Mmin 、v Mmax They represent the minimum and maximum speeds of the M fleet, v X represents the speed of vehicle X at the time of lane change, and Δv is limited to ensure the speed v of the lane-changing vehicle after the lane change. X +Δv in [v Mmin ,v Mmax ] range; d safe is the safe distance that the lane-changing vehicle X and the vehicles in the convoy M should meet before and after the insertion position during the lane change, d safe_max Set by road conditions, dsafe_min is the minimum safe distance between the lane-changing vehicle X and the vehicles in front and behind the platoon M during the lane change and μ is the road friction coefficient.

[0013] Step 22: Monte Carlo sampling is used to generate h random points for each insertion gap k and put them into the initial candidate strategy set, as follows:

[0014]

[0015] Among them, p is the location point of lane change insertion, k n is a randomly generated point for each insertion gap k, h is the total number of random points generated for the insertion gap k, To insert k in lane-changing vehicle X n The maximum speed adjustment of lane-changing vehicle X when inserting k n The maximum value of the continuous speed change of the lane-changing vehicle X during the process, is the insertion point k of the team M during the lane change process n The safety distance that should be maintained between the front and rear vehicles and the lane-changing vehicle X;

[0016] Step 23: For lane change insertion point k n , if Δv and d are safe If certain conditions are met, the vehicle will be included in the candidate set for insertion into the formation. The conditions are as follows:

[0017]

[0018] Among them, x M,k+1 、x M,k They represent the lane change insertion point k n The absolute positions of the car behind k+1 and the car in front k, Indicates the insertion point k n The absolute position, v M,k+1 、v M,k 、v X They represent the speeds of the following vehicle, the leading vehicle, and the lane-changing vehicle at the time of lane change, and L represents the length of the vehicle.

[0019] Step 3: Select an appropriate lane-changing trajectory curve and trajectory tracking control to achieve vehicle lane changing. The present invention uses a sinusoidal lane-changing trajectory curve and Model Prediction Control (MPC). The specific implementation steps of Step 3 are as follows:

[0020] Step 31, a lane-changing trajectory planning function based on a sine curve, the function expression is as follows:

[0021]

[0022] Among them, y r (x r ) is the reference lane-changing trajectory function, the subscript r represents the reference trajectory related quantity, x r is the vertical absolute position, D is the lane width, is the longitudinal displacement of the lane-changing vehicle at the initial moment of lane change, is the lateral displacement of the lane-changing vehicle at the initial moment of lane change, Δx x (t) is the planned longitudinal lane-changing displacement at time t, which is expressed as follows:

[0023]

[0024] Among them, v x (tT) is the speed of the lane-changing vehicle X at time tT, D is the lane width, T is the sampling period, t0,t end are the initial and end time of lane change, respectively, a x (t) is the acceleration of the lane-changing vehicle at time t, a y is the lateral acceleration of the lane-changing vehicle.

[0025] Step 32: Perform lane change trajectory tracking control based on MPC

[0026] First, the curvature of the trajectory curve is calculated to obtain the expected heading angle and expected front wheel turning angle of the lane-changing vehicle, and the error state vector of the lane-changing vehicle is calculated from this. and error control vector

[0027] Secondly, during the trajectory tracking process, in order to ensure that the error between the actual trajectory and the desired trajectory and the control error are as small as possible, the objective function is designed to ensure that the vehicle can quickly and smoothly track the reference trajectory. The minimum cost function and constraints of MPC are as follows:

[0028]

[0029] The constraints are as follows:

[0030] U min ≤U(t+w|t)≤U max ,w=0,1,…,N c -1

[0031] Among them, Q and R are weight matrices, N u and N x are the number of control variables and state variables, I is the unit matrix, U(t+w|t) is the control variable at the prediction time step t+w, N p and N cThey are prediction time domain and control time domain, U min and U max are the minimum and maximum values ​​of the control variable U.

[0032] Step 4 adjusts the vehicle parameters of the two platoons after the lane change, including speed, absolute position, and distance between the front and rear vehicles. Based on the candidate set of vehicle positions for the lane-changing vehicle determined in Step 2, a lane-change impact assessment module is constructed, including a safety assessment unit, a platoon stability assessment unit, a traffic efficiency assessment unit, and a transition smoothness assessment unit. The calculation results of each unit are normalized. The specific implementation steps of Step 4 are as follows:

[0033] Step 41: Modify the parameters of the vehicles in the M fleet after the lane change:

[0034] Assume that the lane-changing vehicle changes lanes between the i-th vehicle and the i+1-th vehicle in the convoy M. After the lane change, adjust the speed and absolute position of the vehicles in the convoy M. Adjust the speed of all vehicles before the lane-changing vehicle to Within the range, and adjust the absolute position to: x M,k +d global_safe_min / 2(k≤i); adjust the speed of all vehicles after the lane-changing vehicle to Absolute position adjustment: x M,k -d global_safe_min / 2(k>i); the speed of the lane-changing vehicle is adjusted to The absolute vertical position is adjusted to: (x M,i +x M,i+1 ) / 2. Among them, v X 、v M,k is the speed of the lane-changing vehicle and the corresponding vehicle in the team M at the lane-changing time, x M,k 、x M,i 、x M,i+1 is the absolute position of the corresponding vehicle in the fleet M at the time of lane change.

[0035] Step 42: Modify the parameters of the vehicles in the N-vehicle fleet after the lane change:

[0036] After changing lanes, adjust the speed and absolute position of all vehicles in convoy N. The speed and absolute position of all vehicles before the lane-changing vehicle relative to the position of convoy N remain unchanged; adjust the speed of all vehicles after the lane-changing vehicle relative to the position of convoy N to [v X-1 ,v N,k +v global_max ] range, the absolute position is adjusted to: x N,k +L+d global_safe_min (k>x). Among them, v X-1It represents the speed of the preceding vehicle X-1 of the platoon N of the lane-changing vehicle X at the time of lane change, and L represents the length of the vehicle.

[0037] By limiting the speed and absolute distance between the vehicles in convoys M and N, and ensuring that the distance between two adjacent vehicles during the lane change is not less than d global_safe_min , which can realize the safe and stable operation of dual fleets.

[0038] Step 43: Calculate the safety lane change impact index of the dual queue:

[0039] a. First, we traverse all vehicles in fleets M and N and calculate the TTC (time to collision) between each pair of leading and trailing vehicles after the lane change:

[0040]

[0041] Where L is the vehicle length, v k is the speed of the following vehicle, v k-1 is the speed of the preceding vehicle, x k is the coordinate of the following vehicle’s forward direction along the lane line, x k-1 is the coordinate of the front vehicle’s forward direction along the lane line. First, compare the speeds of the front and rear vehicles. When the rear vehicle’s speed v k Less than or equal to the speed of the vehicle in front v k-1 When , the safe collision time is set to infinity (no collision), and traverse to the next step, with car k as the front car and car k+1 as the back car, and then compare the speeds of the two cars.

[0042] b. The normalized scoring mechanism for collision time is as follows:

[0043]

[0044] Among them, TTC threshold It is the value recommended by the U.S. National Highway Traffic Safety Administration, generally between 2.5 and 3 seconds. MAX It is the upper limit of the actual observation safety time interval, generally 10s, TTC min is the minimum TTC of all adjacent vehicles in the two platoons k .

[0045] Step 44, calculate the stability lane change impact index of the dual queue:

[0046] a. Introducing the coefficient of variation of the spacing between the M and N queues of the fleet:

[0047]

[0048] Where u represents the fleet M or N, σ gaps (u) is the standard deviation of the distance between vehicles in this platoon, μ gaps(u) is the average distance between vehicles in this platoon. This formula reflects the degree of dispersion of the distance between vehicles in this platoon.

[0049] b. Introducing velocity gradient factor Calculate the speed gradient difference of the front and rear vehicles at the lane change insertion position:

[0050]

[0051] Among them, V front is the speed of the vehicle in front of the lane-changing position, V rear is the speed of the vehicle behind the lane-changing position, V insert is the lane-changing vehicle speed, Δx front is the distance between the lane-changing vehicle and the preceding vehicle, Δx rear is the distance between the lane-changing vehicle and the following vehicle. This formula measures the magnitude of the speed difference between the vehicles before and after the insertion point.

[0052] c. Stability evaluation formula:

[0053]

[0054] Among them, ζ and η are the weights of the coefficient of variation of the spacing between the M and N queues of the fleet, and are generally taken as ζ = 0.5 and η = 0.5. ω1 and ω2 are the weight coefficients of the total spacing coefficient of variation and the speed gradient difference, and are generally taken as ω1 = 0.7 and ω2 = 0.3.

[0055] Step 45, calculate the efficiency lane change impact index of the dual queue:

[0056] ratio=v new / v orig

[0057]

[0058] Among them, v orig is the initial head vehicle speed of the fleet M, v new is the speed of the first vehicle in the platoon M after the inserted vehicle, if v new ≥v orig , the team accelerates, using the hyperbolic tangent function and normalizing the efficiency index; if v new <v orig , then a Gaussian function is used to penalize the speed drop. τ is the speed fluctuation tolerance, which is generally set to 20% to control the speed drop trend.

[0059] Step 46, calculate the smooth lane change impact index of the dual queue:

[0060]

[0061] Where Δv Xis the speed difference of vehicle X before and after the lane change, Indicates that the absolute value of the speed adjustment is normalized to the [0,1] interval.

[0062] Step 5: Based on the environmental information data of the road section where the vehicle queue passes in step 1, such as weather visibility, wind speed, road friction coefficient, etc., the weight of each dimensional evaluation unit in step 4 is dynamically adjusted. The specific implementation steps of step 5 are as follows:

[0063] Step 51, define the original weight W base =[ω safety ,ω stability ,ω efficiency ,ω transition ]=[1,1,1,1], define the environmental factor F env =[f safety ,f stability ,f efficiency ,f transition ].

[0064] Step 52: Calculate the adjustment factor of each weight. The expression is as follows:

[0065]

[0066] in, φ and ψ are the setting environment parameters, represents the road friction coefficient, φ represents the weather visibility, and ψ represents the wind speed.

[0067] Step 53: Set the adjusted weights. The expression is as follows:

[0068] W adjusted =softmax(α·W base +β·F env )

[0069]

[0070] s i =α·W base +β·F env

[0071] Among them, α and β are hyperparameters that control the balance between the original weights and the environmental influence (default α = 1.0, β = 0.5), ensuring that the sum of the adjusted weights in the softmax function is 1.

[0072] Step 54: (α·W base +β·F env ) is substituted into the softmax function to obtain the adjusted weight W adjusted .

[0073] Step 6: traverse each lane change insertion position in the candidate set and perform a weighted average of the lane change impact index obtained in step 4 and the corresponding weight in step 5 to obtain a comprehensive index for each lane change insertion position. Compare the comprehensive scores at each expected lane change position and select the insertion position with the highest score and its lane change decision as the optimal lane change strategy. The specific implementation steps of step 6 are as follows:

[0074] Step 61: Determine the comprehensive score calculation formula, which is as follows:

[0075] S total =W adjusted ·(S safety ,S stability ,S efficiency ,S transition )

[0076] Among them, S total Give an overall score for this lane change position.

[0077] In step 62, each lane change insertion position in the candidate set is traversed, and the weight obtained in step 5 is added to the various indicators calculated in step 4 to obtain a comprehensive score for the lane change insertion position. The comprehensive scores of each insertion position in the candidate set are compared, and the insertion position with the highest score and its lane change decision are selected as the optimal lane change strategy.

[0078] The advantages of the present invention compared with the prior art are:

[0079] (1) Different from other strategies that use dynamic analysis or formation control algorithms to form a formation, the present invention provides corresponding control strategies for vehicle insertion and formation reconstruction based on the scenario of lane-changing vehicles joining the vehicle formation, ensuring the safety of the formation. On the basis of meeting the requirements of safe lane changing and comfortable driving, the lane-changing vehicles can be inserted into the existing vehicle formation and form a new formation, thereby improving traffic efficiency.

[0080] (2) Compared with traditional control methods that use a single indicator (such as shortest lane change time or minimum energy consumption) as the optimization target, this invention innovatively proposes a multi-objective dynamic weight allocation mechanism. By introducing the road surface-weather influencing factor and the softmax function, it achieves autonomous weight adjustment of indicators such as safety, efficiency, and comfort, solving the problem of insufficient scenario generalization caused by manually setting fixed weights in existing technologies.

[0081] (3) Different from vehicle cooperative control algorithms that rely on a preset formation structure, the present invention proposes a joint optimization method for formation reconstruction and lane change insertion. Existing technologies often assume that the formation shape is fixed, resulting in a global reorganization of vehicles for insertion and high communication overhead. The present invention uses local position prediction and gap elasticity calculation to complete insertion by only adjusting the spacing between 3-4 adjacent vehicles, significantly shortening the formation reconstruction time.

[0082] (4) To address the step-like speed regulation flaw of traditional methods in transition smoothness control, the present invention develops a speed gradient model based on a cosine function. By combining the maximum speed adjustment constraint with a smoothing factor, the sudden acceleration / deceleration phenomenon during vehicle insertion is eliminated. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 Flowchart for realizing the method of the present invention;

[0084] Figure 2 A schematic diagram of the initial state of the vehicle planned for the present invention;

[0085] Figure 3 Insert a vehicle platoon diagram for lane-changing vehicles. DETAILED DESCRIPTION

[0086] The present invention will be described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the examples are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0087] like Figure 1 As shown, the specific implementation method of the method of the present invention includes the following steps:

[0088] Step 1: Use sensing equipment and technology to obtain the road environment information of the original lane-changing vehicle queue (referred to as queue N) and the target lane-changing vehicle queue (referred to as queue M). The specific information is as follows:

[0089] Obtain data such as the number of vehicles in the two queues, the distance between vehicles, the speed of vehicles, and the movement status of vehicles requesting lane change, such as Figure 2 As shown. Determine the number of vehicles m in the current fleet M and the coordinate x of the kth vehicle's forward direction along the lane line M,k and speed V M,k , the distance d between the kth car and the k+1th car M,k , the number of vehicles in the team N is n, and the coordinates of the k-th vehicle moving along the lane line are x N,k With speed V N,k , the distance d between the kth car and the k+1th car N,k , the lane-changing vehicle X is at position x in the convoy N. Assume that the length of all vehicles is L and the minimum safe distance between adjacent vehicles is d global_safe_min =2L. Set the absolute value of the speed fluctuation of a single vehicle to no more than v global_max =5km / h.

[0090] Step 2: Based on the data from step 1 regarding the environmental information of the two queues' traffic sections, and considering the impact of the speed difference and desired spacing between vehicles in the target queue on the lane change insertion position, an optimized decision space vector is constructed to determine the candidate set of positions for the lane-changing vehicle to insert into the convoy. The specific implementation steps for step 2 are as follows:

[0091] Step 21: Based on the data of the environmental information of the two queues passing through the road section in step 1, an optimized decision space vector Θ is constructed. The specific expression is as follows:

[0092] Θ={(k,Δv,d safe )}

[0093]

[0094] Where k is the position number of the vehicle X that is about to be inserted into the convoy M (not considering the position before or after the front vehicle). The number of the vehicle before position k is set to k, the number of the vehicle after position k is set to k+1, and so on. Δv is the speed adjustment value, which is used to simulate the maximum speed change of vehicle X during the lane change process. v Mmin 、v Mmax They represent the minimum and maximum speeds of the M fleet, v X represents the speed of vehicle X at the time of lane change, and Δv is limited to ensure the speed v of the lane-changing vehicle after the lane change. X +Δ in [v Mmin ,v Mmax ] range; d safe is the safe distance that the lane-changing vehicle X and the vehicles in the convoy M should meet before and after the insertion position during the lane change, d safe_max Set by road conditions, d safe_min is the minimum safe distance between the lane-changing vehicle X and the vehicles in front and behind the platoon M. μ is the road friction coefficient.

[0095] Step 22: Monte Carlo sampling is used to generate several random points for each insertion gap k and put them into the initial candidate strategy set, as follows:

[0096]

[0097] Among them, p is the location point of lane change insertion, k n is a randomly generated point for each insertion gap k, h is the total number of random points generated for the insertion gap k, To insert k in lane-changing vehicle X n The maximum speed adjustment of lane-changing vehicle X when inserting k n The maximum value of the continuous speed change of the lane-changing vehicle X during the process, is the insertion point k of the team M during the lane change process n The safety distance that should be maintained between the front and rear vehicles and the lane-changing vehicle X;

[0098] Step 23: For lane change insertion point k n , if Δv and d are safeIf all of them meet certain conditions, they will be included in the candidate set for vehicle insertion into the formation. The conditions are as follows:

[0099]

[0100] Among them, x M,k+1 、x M,k They represent the lane change insertion point k n The absolute positions of the following car k+1 and the preceding car k, x kn Indicates the insertion point k n The absolute position, v M,k+1 、v M,k 、v X They represent the speeds of the following vehicle, the leading vehicle, and the lane-changing vehicle at the time of lane change, and L represents the length of the vehicle.

[0101] Step 3: Select an appropriate lane-changing trajectory curve and trajectory tracking control to achieve vehicle lane changing. The present invention uses a sinusoidal lane-changing trajectory curve and Model Prediction Control (MPC). The specific implementation steps of Step 3 are as follows:

[0102] Step 31, a lane-changing trajectory planning function based on a sine curve, the function expression is as follows:

[0103]

[0104] Among them, y r (x r ) is the reference lane-changing trajectory function, the subscript r represents the reference trajectory related quantity, x r is the vertical absolute position, D is the lane width, is the longitudinal displacement of the lane-changing vehicle at the initial moment of lane change, is the lateral displacement of the lane-changing vehicle at the initial moment of lane change, Δx x (t) is the planned longitudinal lane-changing displacement at time t, which is expressed as follows:

[0105]

[0106] Among them, v x (tT) is the speed of the lane-changing vehicle X at time tT, D is the lane width, T is the sampling period, t0,t end are the initial and end time of lane change, respectively, a x (t) is the acceleration of the lane-changing vehicle at time t, a y is the lateral acceleration of the lane-changing vehicle.

[0107] Step 32: Perform lane change trajectory tracking control based on MPC

[0108] First, the curvature of the trajectory curve is calculated to obtain the expected heading angle and expected front wheel turning angle of the lane-changing vehicle, and the error state vector of the lane-changing vehicle is calculated from this. and error control vector

[0109] Secondly, during the trajectory tracking process, in order to ensure that the error between the actual trajectory and the desired trajectory and the control error are as small as possible, the objective function is designed to ensure that the vehicle can quickly and smoothly track the reference trajectory. The minimum cost function and constraints of MPC are as follows:

[0110]

[0111] The constraints are as follows:

[0112] U min ≤U(t+w|t)≤U max ,w=0,1,…,N c -1

[0113] Among them, Q and R are weight matrices, N u and N x are the number of control variables and state variables, I is the unit matrix, U(t+w|t) is the control variable at the prediction time step t+w, N p and N c They are prediction time domain and control time domain, U min and U max are the minimum and maximum values ​​of the control variable U.

[0114] The core idea is to update the lane-changing vehicle's speed in real time through a longitudinal control algorithm, then use the dynamically updated speed to generate the lane-changing longitudinal displacement in real time, and then transfer it to a sine function to achieve dynamic trajectory planning.

[0115] Step 4 adjusts the vehicle parameters of the two platoons after the lane change, including speed, absolute position, and distance between the front and rear vehicles. Based on the candidate set of vehicle positions for the lane-changing vehicle determined in Step 2, a lane-change impact assessment module is constructed, including a safety assessment unit, a platoon stability assessment unit, a traffic efficiency assessment unit, and a transition smoothness assessment unit. The calculation results of each unit are normalized. The specific implementation steps of Step 4 are as follows:

[0116] Step 41: Modify the parameters of the vehicles in the M fleet after the lane change:

[0117] Assume that the lane-changing vehicle changes lanes between the i-th vehicle and the i+1-th vehicle. After the lane change, adjust the speed and absolute position of the vehicles in the convoy M. Adjust the speed of all vehicles before the lane-changing vehicle to Within the range, and adjust the absolute position to: x M,k +d global_safe_min / 2(k≤i); the speed of all vehicles after the lane-changing vehicle is adjusted to Absolute position adjustment: x M,k -d global_safe_min / 2(k>i); the speed of the lane-changing vehicle is adjusted to The absolute vertical position is adjusted to: (x M,i +x M,i+1 ) / 2. Among them, v X 、v M,k is the speed of the lane-changing vehicle and the corresponding vehicle in the team M at the lane-changing time, x M,k 、x M,i 、x M,i+1 is the absolute position of the corresponding vehicle in the fleet M at the time of lane change.

[0118] Step 42: Modify the parameters of the vehicles in the N-vehicle fleet after the lane change:

[0119] After changing lanes, adjust the speed and absolute position of all vehicles in convoy N. The speed and absolute position of all vehicles before the lane-changing vehicle relative to the position of convoy N remain unchanged; adjust the speed of all vehicles after the lane-changing vehicle relative to the position of convoy N to [v X-1 ,v N,k +v global_max ] range, the absolute position is adjusted to: x N,k +L+d global_safe_min (k>x). Among them, v X-1 It represents the speed of the preceding vehicle X-1 of the platoon N of the lane-changing vehicle X at the time of lane change, and L represents the length of the vehicle.

[0120] By limiting the speed and absolute distance between the vehicles in convoys M and N, and ensuring that the distance between two adjacent vehicles during the lane change is not less than d global_safe_min , which can realize the safe and stable operation of dual fleets.

[0121] Step 43: Calculate the safety lane change impact index of the dual queue:

[0122] Incorporating safety assessment into evaluation metrics is key to addressing the limitations of traditional lane-changing strategies. Traditional methods rely on a single safety threshold and fail to quantify the cascading risks caused by vehicles entering a queue. These include the surge in collision probability due to speed differences, the propagation of disturbances caused by the loss of queue spacing consistency, and the psychological safety risks associated with mixed traffic. This approach quantifies safety risks by incorporating piecewise linear normalization of the time-to-collision (TTC) value.

[0123] a. First, traverse all the cars in the fleet M and N and calculate the TTC (collision) between each pair of front and rear cars after the lane change.

[0124] Collision time):

[0125]

[0126] Where L is the vehicle length, v k is the speed of the following vehicle, v k-1 is the speed of the preceding vehicle, x k is the coordinate of the following vehicle’s forward direction along the lane line, x k-1 is the coordinate of the front vehicle’s forward direction along the lane line. First, compare the speeds of the front and rear vehicles. When the rear vehicle’s speed v k Less than or equal to the speed of the vehicle in front v k-1 When , the safe collision time is set to infinity (no collision), and traverse to the next step, car k is the front car, car k+1 is the back car, and then compare the speed of the two cars.

[0127] b. The normalized scoring mechanism for collision time is as follows:

[0128]

[0129] Among them, TTC threshold It is the value recommended by the U.S. National Highway Traffic Safety Administration, generally between 2.5 and 3 seconds. MAX It is the upper limit of the actual observation safety time interval, generally 10s, TTC min is the minimum TTC of all adjacent vehicles in the two platoons k .

[0130] Step 44, calculate the stability lane change impact index of the dual queue:

[0131] Introducing stability assessment into evaluation metrics is a key measure to address the dynamic disturbance propagation and queue imbalance risks caused by lane-changing vehicles cutting into the queue. Traditional methods often focus on instantaneous safe spacing, but ignore the long-tail impact of insertion on the queue structure. When lane-changing vehicles cut in at non-equilibrium speeds, they disrupt the coordinated following relationship between vehicles, triggering the transmission of speed fluctuations and the amplification of spacing oscillations, forming disturbance waves that propagate bidirectionally along the queue (coupled forward compression waves and backward rarefaction waves). This solution proposes a dual-modal stability assessment model that quantifies speed fluctuations through the following innovations.

[0132] a. Introducing the coefficient of variation of the spacing between the M and N queues of the fleet:

[0133]

[0134] Where u represents the fleet M or N, σ gaps (u) is the standard deviation of the distance between vehicles in this platoon, μ gaps(u) is the average distance between vehicles in this platoon. This formula reflects the degree of dispersion of the distance between vehicles in this platoon.

[0135] b. Introducing velocity gradient factor Calculate the speed gradient difference of the front and rear vehicles at the lane change insertion position:

[0136]

[0137] Among them, V front is the speed of the vehicle in front of the lane-changing position, V rear is the speed of the vehicle behind the lane-changing position, V insert is the lane-changing vehicle speed, Δx front is the distance between the lane-changing vehicle and the preceding vehicle, Δx rear is the distance between the lane-changing vehicle and the following vehicle. This formula measures the magnitude of the speed difference between the vehicles before and after the insertion point.

[0138] c. Stability evaluation formula:

[0139]

[0140] Among them, ζ and η are the weights of the coefficient of variation of the spacing between the M and N queues of the fleet, and are generally taken as ζ = 0.5 and η = 0.5. ω1 and ω2 are the weight coefficients of the total spacing coefficient of variation and the speed gradient difference, and are generally taken as ω1 = 0.7 and ω2 = 0.3.

[0141] Step 45, calculate the efficiency lane change impact index of the dual queue:

[0142] The introduction of a traffic efficiency evaluation unit within the evaluation metrics aims to address the core contradiction between local efficiency gains and global flow loss caused by lane-changing behaviors. Traditional lane-changing strategies often prioritize the shortest path for individual vehicles, but overlook the "mobile bottleneck effect" of insertions on overall traffic flow. This unit, when a lane-changing vehicle enters a queue at an uncoordinated speed, triggers a chain of speed adjustments for the following vehicles. This evaluation unit quantifies speed adjustments using an asymmetric Gaussian-hyperbolic function, achieving a Pareto optimal balance of safety and efficiency.

[0143] ratio=v new / v orig

[0144]

[0145] Among them, v orig is the initial head vehicle speed of the fleet M, v new is the speed of the first vehicle in the platoon M after the inserted vehicle, if v new ≥v orig , the team accelerates, using the hyperbolic tangent function and normalizing the efficiency index; if v new <vorig , then a Gaussian function is used to penalize the speed drop. τ is the speed fluctuation tolerance, which is generally set to 20% to control the speed drop trend.

[0146] Step 46: Calculate the lane change impact index of the dual-queue smoothness

[0147] The introduction of transition smoothness evaluation into the evaluation index aims to address abnormal and sudden changes in spacing caused by rapid speed changes during lane changes. This evaluation unit uses a cosine constraint function to suppress the sudden changes in speed adjustment. The formula is as follows:

[0148]

[0149] Where Δv X is the speed difference of vehicle X before and after the lane change, max_speed_adjust is the maximum allowable speed adjustment value, which is initially defined as 5.0km / h. Indicates that the absolute value of the speed adjustment is normalized to the interval [0,1]. X When |>max_speed_adjust, the smoothness index is set to 0, indicating a completely uneven lane change; when |Δv X When |≤max_speed_adjust, the cosine function is used to calculate the smoothness index; when |Δv X When |=0, S transition =0 means completely smooth lane change.

[0150] Step 5: Based on the environmental information data of the road section where the vehicle queue passes in step 1, such as weather visibility, wind speed, road friction coefficient, etc., the weight of each dimensional evaluation unit in step 4 is dynamically adjusted. The specific implementation steps of step 5 are as follows:

[0151] Step 51, define the original weight W base =[ω safety ,ω stability ,ω efficiency ,ω transition ]=[1,1,1,1], define the environmental factor F env =[f safety ,f stability ,f efficiency ,f transition ].

[0152] Step 52: Calculate the adjustment factors of each weight and set the environmental parameters φ,ψ, where It represents the road friction coefficient, for example, 0.8 for dry road, 0.4 for wet road, and 0.1 for icy road; φ represents weather visibility, for example, 1.0 for sunny day, 0.6 for rainy day, 0.3 for foggy day, and 0.2 for snowy day; ψ represents wind speed, for example, 0 for no wind and 0.8 for strong wind (different values ​​are set according to different wind strengths).

[0153] Because the safety weight increases significantly when the friction coefficient is low, visibility is poor, and wind speed is high; the stability weight increases when the road friction coefficient is low or lateral disturbance is large; the efficiency weight is higher when the road friction coefficient is high and visibility is good; when the road friction coefficient is high and there is no lateral disturbance, transition smoothness is more important, so the following formula is defined:

[0154]

[0155] in, φ and ψ are the setting environment parameters, represents the road friction coefficient, φ represents the weather visibility, and ψ represents the wind speed.

[0156] Step 53: Set the adjusted weights. The expression is as follows:

[0157] W adjusted =softmax(α·W base +β·F env )

[0158]

[0159] s i =α·W base +β·F env

[0160] Among them, α and β are hyperparameters that control the balance between the original weights and the environmental influence (default α = 1.0, β = 0.5), ensuring that the sum of the adjusted weights in the softmax function is 1.

[0161] Step 54, substitute Calculate the adjustment factor for the environmental parameter values ​​of φ and ψ, and then (α·W base +β·F env ) is substituted into the softmax function to obtain the adjusted weight W adjusted This enables dynamic adjustment of the weights of the four-dimensional evaluation indicators of safety, stability, efficiency and smoothness.

[0162] Step 6: traverse each lane change insertion position in the candidate set, perform a weighted average of the lane change impact index obtained in step 4 and the corresponding weight in step 5, and obtain a comprehensive index for each lane change insertion position. Compare the comprehensive scores at each expected lane change position to generate the optimal lane change strategy. The specific implementation steps of step 6 are as follows:

[0163] Step 61: Determine the comprehensive score calculation formula, which is as follows:

[0164] S total =W adjusted ·(S safety ,S stability ,S efficiency ,S transition )

[0165] Among them, S total Give an overall score for this lane change position.

[0166] In step 62, each lane change insertion position in the candidate set is traversed, and the weight obtained in step 5 is added to the various indicators calculated in step 4 to obtain a comprehensive indicator for each lane change insertion position. The comprehensive scores at each expected lane change position are compared, and the insertion position with the highest score and its lane change decision are selected as the optimal lane change strategy.

[0167] The above steps describe the implementation process of the present invention in detail, but the present invention is not limited to the specific details of the above embodiments. Anything within the scope of the present invention should not be excluded from the scope of protection of the present invention.

Claims

1. An intelligent lane-changing decision method for an autonomous vehicle based on multi-dimensional evaluation, characterized in that: The following steps are involved: Step 1: Use sensing equipment and technology to obtain road environment information for the original queue (represented as queue N) and the target queue (represented as queue M). This includes data such as the number of vehicles in the two queues, the distance between vehicles, vehicle speeds, and the motion status of the vehicle requesting the lane change. Step 2: Based on the data of the road environment information of the two queues in step 1, and considering the influence of the speed difference and the desired spacing between vehicles in the target queue on the lane change insertion position, an optimized decision space vector is constructed to determine the candidate set of positions for the lane-changing vehicle to insert into the convoy; Step 3: Select an appropriate lane-changing trajectory curve and trajectory tracking control to realize the vehicle lane changing. The present invention uses a sinusoidal function lane-changing trajectory curve and Model Prediction Control (MPC); Step 4: Adjust the vehicle parameters of the two platoons after the lane change, including speed, absolute position, and distance between the front and rear vehicles. Based on the candidate set of vehicle insertion positions determined in Step 2, construct a lane change impact assessment module, including a safety assessment unit, a platoon stability assessment unit, a traffic efficiency assessment unit, and a transition smoothness assessment unit. The calculation results of each unit are normalized. Step 5: Based on the data of the environmental information of the road section where the vehicle queue passes in step 1, such as weather visibility, wind speed, road friction coefficient, etc., dynamically adjust the weight of each dimensional evaluation unit in step 4 in the comprehensive index; In step 6, each lane change insertion position in the candidate set is traversed. The lane change impact index obtained in step 4 is weighted averaged with the corresponding weight in step 5 to obtain the comprehensive index of each lane change insertion position. The comprehensive scores at each expected lane change position are compared, and the insertion position with the highest score and its lane change decision are selected as the optimal lane change strategy.

2. The method according to claim 1, wherein: The environmental information of the road section where the two convoys pass, including the number of vehicles in the two queues, the distance between vehicles, the speed of vehicles, and the motion status of vehicles requesting lane change, is obtained, thereby determining the number of vehicles m in the current convoy M and the coordinate x of the kth vehicle's forward direction along the lane line. M,k and speed V M,k , the distance d between the kth car and the k+1th car M,k , the number of vehicles in the team N is n, and the coordinates of the k-th vehicle moving along the lane line are x N,k With speed V N,k , the distance d between the kth car and the k+1th car N,k , the lane-changing vehicle X is at position x in the convoy N. Assume that the length of all vehicles is L and the minimum safe distance between adjacent vehicles is d global_safe_min =2L. Set the absolute value of the speed fluctuation of a single vehicle to no more than v global_max =5km / h.

3. The method according to claim 1, wherein: In step 2, based on the data of the environmental information of the two queues' traffic sections in step 1, and considering the influence of the speed difference and the desired spacing between vehicles in the target queue on the lane change insertion position, an optimized decision space vector is constructed to determine the candidate set of positions for the lane change vehicle to be inserted into the convoy. The implementation steps are as follows: (1) First, construct the optimized decision space vector Θ. The specific expression is as follows: Θ={(k,Δv,d safe )} Where k is the position number of the vehicle X that is about to be inserted into the convoy M (not considering the position before or after the front vehicle). The number of the vehicle before position k is set to k, the number of the vehicle after position k is set to k+1, and so on. Δv is the speed adjustment value, which is used to simulate the maximum speed change of vehicle X during the lane change process. v Mmin 、v Mmax They represent the minimum and maximum speeds of the M fleet, v X represents the speed of vehicle X at the time of lane change, and Δv is limited to ensure the speed v of the lane-changing vehicle after the lane change. X +Δv in [v Mmin ,v Mmax ] within the range; d safe is the safe distance that the lane-changing vehicle X and the vehicles in the convoy M should meet before and after the insertion position during the lane change, d safe_max Set by road conditions, d safe_min is the minimum safe distance between the lane-changing vehicle X and the vehicles in front and behind the platoon M during the lane change and μ is the road friction coefficient; (2) Monte Carlo sampling is used to generate h random points for each insertion gap k and put them into the initial candidate strategy set, as follows: Among them, p is the location point of lane change insertion, k n is a randomly generated point for each insertion gap k, h is the total number of random points generated for the insertion gap k, Δv (kn) To insert k in lane-changing vehicle X n The maximum speed adjustment of lane-changing vehicle X when inserting k n The maximum value of the continuous speed change of the lane-changing vehicle X during the process, is the insertion point k of the team M during the lane change process n The safety distance that should be maintained between the front and rear vehicles and the lane-changing vehicle X; (3) Finally, for the lane change insertion point k n , if Δv and d are safe If the following conditions are met, the vehicle is included in the candidate set for the queue position: Among them, x M,k+1 、x M,k They represent the lane change insertion point k n The absolute positions of the car behind k+1 and the car in front k, Indicates the insertion point k n The absolute position, v M,k+1 、v M,k 、v X They represent the speeds of the following vehicle, the leading vehicle, and the lane-changing vehicle at the time of lane change, and L represents the length of the vehicle.

4. The method according to claim 1, wherein: In step 4, the parameters of the two convoy vehicles after the lane change are adjusted as follows: (1) Assume that the lane-changing vehicle changes lanes between the i-th vehicle and the i+1-th vehicle in the convoy M. After the lane change, adjust the speed and absolute position of the vehicles in the convoy M. Adjust the speed of all vehicles before the lane-changing vehicle to Within the range, and adjust the absolute position to: x M,k +d global_safe_min / 2(k≤i); adjust the speed of all vehicles after the lane-changing vehicle to [v M,k -v global_max ,v X +Δv (in) ], the absolute position is adjusted to: x M,k -d global_safe_min / 2(k>i); the speed of the lane-changing vehicle is adjusted to v X +Δv (in) , the absolute vertical position is adjusted to: (x M,i +x M,i+1 ) / 2. Among them, v X 、v M,k is the speed of the lane-changing vehicle and the corresponding vehicle in the team M at the lane-changing time, x M,k 、x M,i 、x M,i+1 is the absolute position of the corresponding vehicle in the fleet M at the time of lane change; (2) After changing lanes, adjust the speed and absolute position of all vehicles in convoy N. The speed and absolute position of all vehicles before the lane-changing vehicle relative to the position of convoy N remain unchanged; adjust the speed of all vehicles after the lane-changing vehicle relative to the position of convoy N to [v X-1 ,v N,k +v global_max ] range, the absolute position is adjusted to: x N,k +L+d global_safe_min (k>x). Among them, v X-1 represents the speed of the preceding vehicle X-1 of the platoon N of the lane-changing vehicle X at the time of lane change, and L represents the length of the vehicle; By limiting the speed and absolute distance between the vehicles in convoys M and N, and ensuring that the distance between two adjacent vehicles during the lane change is not less than d global_safe_min , which can realize the safe and stable operation of dual fleets.

5. The method according to claim 1, wherein: In step 4, the four-dimensional lane-changing impact indicators of the two queues, namely safety, stability, efficiency, and smoothness, are calculated as follows: (1) Calculate the safety lane change impact index of the double queue: a. First, we traverse all vehicles in fleets M and N and calculate the TTC (time to collision) between each pair of leading and trailing vehicles after the lane change: Where L is the vehicle length, v k is the speed of the kth vehicle, v k-1 is the speed of the car in front of car k, x k is the coordinate of the k car's forward direction along the lane line, x k-1 is the coordinate of the front car's forward direction along the lane line. First, compare the speeds of the two cars. When the speed of car k is v k Less than or equal to the speed of the vehicle in front v k-1 When , the safe collision time is set to infinity (no collision), and traverse to the next step to compare the speeds of car k and car k+1; b. The normalized scoring mechanism for collision time is as follows: Among them, TTC threshold It is the value recommended by the U.S. Highway Safety Administration, generally between 2.5 and 3 seconds. MAX It is the upper limit of the actual observation safety time interval, generally 10s, TTC min is the minimum TTC of all adjacent vehicles in the two platoons k ; (2) Calculate the stability lane change impact index of the dual queue: a. Introducing the coefficient of variation of the spacing between the M and N queues of the fleet: Where u represents the fleet M or N, σ gaps (u) is the standard deviation of the distance between vehicles in this platoon, μ gaps (u) is the average distance between vehicles in this platoon. This formula reflects the degree of dispersion of the distance between vehicles in this platoon; b. Introducing velocity gradient factor Calculate the speed gradient difference of the front and rear vehicles at the lane change insertion position: Among them, V front is the speed of the vehicle in front of the lane-changing position, V rear is the speed of the vehicle behind the lane-changing position, V insert is the lane-changing vehicle speed, Δx front is the distance between the lane-changing vehicle and the preceding vehicle, Δx rear is the distance between the lane-changing vehicle and the following vehicle. This formula can be used to measure the magnitude of the speed difference between the vehicles before and after the insertion point. c. Stability evaluation formula: Among them, ζ and η are the weights of the coefficient of variation of the spacing between the M and N queues of the fleet, and are generally set to ζ = 0.5 and η = 0.

5. ω1 and ω2 are the weight coefficients of the total spacing coefficient of variation and the speed gradient difference, and are generally set to ω1 = 0.7 and ω2 = 0.

3. (3) Calculate the efficiency lane change impact index of the double queue: ratio=v new / v orig Among them, v orig is the initial head vehicle speed of the fleet M, v new is the speed of the first vehicle in the platoon M after the inserted vehicle, if v new ≥v orig , the team accelerates, using the hyperbolic tangent function and normalizing the efficiency index; if v new <v orig , then the Gaussian function is used to penalize the speed drop. τ is the speed fluctuation tolerance, which is generally set to 20% to control the speed drop trend; (4) Calculate the smooth lane change impact index of the double queue: Where Δv X is the speed difference of vehicle X before and after the lane change, Indicates that the absolute value of the speed adjustment is normalized to the [0,1] interval.

6. The method according to claim 1, wherein: The step 5 is implemented as follows: (1) Define the original weight W base =[ω safety ,ω stability ,ω efficiency ,ω transition ]=[1,1,1,1], define the environmental factor F env =[f safety ,f stability ,f efficiency ,f transition ]; (2) Calculate the adjustment factor of each weight: in, φ and ψ are the setting environment parameters, represents the road friction coefficient, φ represents the weather visibility, and ψ represents the wind speed; (3) Set the adjusted weights as: W adjusted =softmax(α·W base +β·F env ) s i =α·W base +β·F env Among them, α and β are hyperparameters that control the balance between the original weights and the environmental influence (default α = 1.0, β = 0.5), and the softmax function ensures that the sum of the adjusted weights is 1.

7. The method according to claim 1, wherein: The step 6 is implemented as follows: (1) The formula for calculating the comprehensive score is as follows: S total =W adjusted ·(S safety ,S stability ,S efficiency ,S transition ) Among them, S total For this lane change position, a comprehensive score is given; (2) Traverse each lane-changing insertion position in the candidate set, add the weight obtained in step 5 to the various indicators calculated in step 4 to obtain the comprehensive score of the lane-changing insertion position, compare the comprehensive scores of the fleet after lane changing at each lane-changing position in the candidate set, and select the insertion position with the highest score and its lane-changing decision as the optimal lane-changing strategy.

Images