Left-turn operation optimization method for intelligent networked vehicle

Abstract

The invention relates to an intelligent network connection vehicle left-turn operation optimization method, which comprises the following steps: extracting the speed, curvature and turning angle of a left-turn vehicle from an intersection vehicle driving video, classifying left-turn driving tracks by adopting a turning track classification algorithm based on rules, and fitting different types of left-turn driving tracks; constructing a conflict point prediction model based on time-space characteristics of multiple tracks, and predicting conflict points; based on the conflict point prediction result, a variable speed adjustment strategy is adopted to generate a vehicle turning speed; and based on different types of left-turn driving fitting trajectories and vehicle turning data, travel time and parking delay are comprehensively considered, and a vehicle multi-trajectory optimal strategy is constructed for trajectory planning. Compared with the prior art, the method has the advantages of high flexibility and reliability, and can meet the requirements of trajectory classification, speed adjustment and multi-trajectory collaborative optimization at the same time.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to a method for optimizing left-turn operation of an intelligent networked vehicle. Background Art

[0002] With the rapid development of connected and autonomous vehicles (CAVs) technology, optimizing the trajectory and speed of left-turning vehicles at intersections has become the key to improving traffic operation efficiency. However, existing studies have simplified the trajectory of such vehicles into an ideal curve and set the vehicle to travel at a constant speed along this trajectory, failing to fully consider the fluctuations in the trajectory and speed of the vehicle during the turning interaction process and the impact of the vehicle size on its operating dynamics. This limitation has led to insufficient optimization of CAVs during intersection turning.

[0003] Existing methods lack consideration of the impact of background traffic flow and vehicle body size on the CAV's operating trajectory and speed, and fail to fully solve the problems of trajectory optimization and dynamic speed adjustment.

[0004] Therefore, it is urgent to design a left-turn operation scheme for autonomous driving vehicles at intersections that combines trajectory classification, speed adjustment and multi-trajectory collaborative optimization. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a flexible and reliable left-turn operation optimization method for intelligent connected vehicles, which can simultaneously meet the requirements of trajectory classification, speed adjustment and multi-trajectory collaborative optimization.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] The present invention provides a method for optimizing left-turn operation of an intelligent networked vehicle, comprising:

[0008] The speed, curvature and turning angle of left-turning vehicles are extracted from the intersection vehicle driving video, and the left-turn driving trajectory is classified using a rule-based turning trajectory classification algorithm, and different types of left-turn driving trajectories are fitted;

[0009] Based on the spatiotemporal characteristics of multiple trajectories, a conflict point prediction model is constructed to predict conflict points;

[0010] Based on the conflict point prediction results, a variable speed adjustment strategy is used to generate the vehicle turning speed;

[0011] Based on different types of left-turn driving fitting trajectories and vehicle turning data, the optimal vehicle multi-trajectory strategy is constructed for trajectory planning by comprehensively considering travel time and parking delay.

[0012] Preferably, the fitting of different types of vehicle running trajectories is specifically: fitting different types of vehicle running trajectories using exponential functions.

[0013] Preferably, the categories of the left-turn driving trajectory include a nearby crossing trajectory, a normal crossing trajectory and a detour crossing trajectory.

[0014] Preferably, according to the actual trajectory before the conflict point, a rule-based turning trajectory classification algorithm is used to classify the left-turn driving trajectory, specifically:

[0015] For the curvature k i , speed v i and the turning angle β i , set the maximum curvature k 0 , parking speed v stop and the maximum turning angle β 0 :

[0016] 1) If there exists k i ∈(-∞,0)∪(k 0 ,+∞) and Q C ={(x i ,y i )∈C i |x i <x Q ,y i <y Q}≠{}, it is determined to be a nearby crossing trajectory, Q C Indicates that the point of conflict (x Q ,y Q ) The vehicle running trajectory point set at the lower left; x Q ,y Q is the horizontal and vertical coordinates of the conflict point Q: if there is v i ≤v stop , it is determined to be a parking and nearby crossing trajectory; if all v i >v stop , it is determined as a non-stop nearby crossing trajectory;

[0017] 2) If all k i ∈[0,k 0 ] or Q C ={(x i ,y i )∈C i |x i <x Q ,y i <y Q}={}, and there is β i <β 0 , it is determined to be a normal crossing trajectory: if there is vi ≤v stop , it is determined to be a parking ordinary crossing trajectory; if all v i >v stop , it is determined as a non-parking ordinary crossing trajectory;

[0018] 3) If all k i ∈[0,k 0 ] or Q C ={(x i ,y i )∈C i |x i <x Q ,y i <y Q}={}, and all β i ≥β 0 , it is determined to be a detour crossing trajectory: if there is v i ≤v stop , it is determined to be a parking detour crossing trajectory; if all v i >v stop , it is determined to be a non-stop detour crossing trajectory.

[0019] Preferably, based on the spatiotemporal characteristics of multiple trajectories, a conflict point prediction model is established to predict the conflict points. The conflict point prediction model is specifically:

[0020]

[0021] Where: v s is the speed of the straight-moving vehicle, x s ,y s v is the horizontal and vertical coordinates of the straight-moving vehicle; l is the speed of the left-turning vehicle, x l ,y l is the horizontal and vertical coordinates of the left-turning vehicle; t is the predicted conflict time; T s 0 Adjust the time zone for the speed of the straight-moving vehicle, T l 0 is the speed adjustment time area for left-turning vehicles; B is the width of the vehicle; L is the length of the vehicle; S 0 is the safety distance before the detection area, S l is the safety distance behind the detection area; x in is the horizontal coordinate of the point where the vehicle enters the conflict, x out is the horizontal coordinate of the vehicle leaving the conflict point; h' is the path slope of the left-turning vehicle trajectory, which indicates the degree of inclination of the trajectory at a certain point and the rate of change of the direction of the path at the current point; g(t) is the dynamic path function of the left-turning vehicle at time t; L 0is the distance from the conflict point to the stop line of the exit road; x c0 and x c1 They are the horizontal coordinates of the left and right boundaries of the lane where the through vehicle is located in the intersection.

[0022] Preferably, the conflict time window is used to record the speed v of the straight-moving vehicle. s , Speed ​​adjustment time zone T for straight-moving vehicles s 0 , left-turn vehicle speed adjustment time zone T l 0 , predict the conflict time t and its corresponding conflict point coordinates (x l ,y l );

[0023] Among them, the conflict time window T i C and the accessible time window T i a The specific search process is:

[0024] 1) Set the time increment Δt and initialize the conflict time set T c ={};

[0025] 2) Time for each left-turn vehicle Perform the following steps:

[0026] Time for each straight vehicle in And R s ={(T s ,k 0 ,v s ,k)},k=1,2,...:for time t∈[max(T l 0 ,T s ,j), is the maximum travel time for turning left through an intersection, is the maximum travel time for going straight through the intersection, if f S (x s ,y s ,v s ,t)=f L (x l ,y l ,v l ,t), then add the corresponding solution to the conflict time set T c =T c ∪{v s ,j,T s ,j 0 ,T l0 ,t,x l ,y l}, time t is updated to t+Δt;

[0027] 3) Return the conflict time set T c .

[0028] Preferably, it also includes performing a time security check, specifically:

[0029] Combined with the passable time window T i a Length And the safety time length Δt safe , when the following conditions are met, all the passable time windows T are given i a′ ∈{T a′}:

[0030]

[0031] Where: c is the constant coefficient set; x c0 With x c1 is the horizontal coordinate boundary of the straight trajectory distribution;

[0032] If in the current cycle If a passable time window cannot be found within the specified time period, the left-turning vehicle will stop at the potential stop line and wait for the next passable time window search result.

[0033] Preferably, the step of adopting a variable speed adjustment strategy to generate a vehicle turning speed specifically includes:

[0034] Construct a vehicle turning speed adjustment interval, wherein the parameters of the vehicle turning speed adjustment interval include the maximum turning speed v of the vehicle 0 , minimum speed v min and acceleration a;

[0035] When the motion state needs to be adjusted, time node b 1 ,b 2 ,b 3 They represent the time points of completing deceleration, starting acceleration, and completing acceleration respectively; speed adjustment must be completed before the conflict point, and the final speed v f Should be in the interval [v max ,v 0 ], and the time to reach the conflict point is T l Should be in the interval t 1 For a left-turning vehicle, the current speed v 0 Slow down to the target speed v min Time required;

[0036] According to the vehicle motion state, the speed adjustment strategy is divided into incomplete deceleration strategy, complete deceleration strategy and parking strategy. The index parameters involved include the initial time T l 0 , time to complete deceleration b 1 ', time to start acceleration b 2 ', time to complete acceleration b 3 ′, time to reach the stop line b 4 ′, time to reach the conflict point b 5 ′, initial velocity v 0 and the final velocity v f .

[0037] Preferably, the speed adjustment strategy specifically includes:

[0038] Incomplete deceleration: If T l ∈{T a ′} and T l 0 +t 1 <T l ≤T l 0 +t 3 , the vehicle needs to be at time T l 0 Start to decelerate until reaching the final speed v f , v f ≥v min , the vehicle state is between the optimal state and the critical state, and the corresponding threshold and strategy calculation expressions are:

[0039]

[0040]

[0041] t 3 =v 0 / v min ·[t 1 -v 0 / 2a·(1-v min / v 0 ) 2 ]

[0042] b 1 =b 2 =b 3 =T l 0 +(v 0 -v min ) / a

[0043] t=T l =T l 0

[0044]

[0045] b 4 ′=T l ,b 5 ′=T l -L 1 / v

[0046] v=v 0 -a(b′-T l 0 )

[0047] Where: t 3 For left turning vehicles from v min Accelerate to target speed v max Time required;

[0048] Complete deceleration strategy: If T l ∈{T a ′} and T l 0 +t 3 <T l <T l 0 +t 4 , the vehicle needs to be at time T l 0 Start to decelerate until time b 1 ', but does not stop completely, and then immediately accelerates until it reaches the final speed v min , the vehicle state is between the critical state and the optimal parking state, and the corresponding threshold and strategy calculation expressions are:

[0049]

[0050] b 1 =b 2 =T l 0 +v 0 / a,b 3 =b 2 +v min / a

[0051]

[0052] b 4 ′=T l ,b 5 ′=T l -L 1 / v

[0053] Where: t 4 The time required for a left-turning vehicle to travel from its initial position to the conflict point;

[0054] Parking strategy: If T l ∈{T a '}and The vehicle needs to slow down to a complete stop and wait for a period of time after stopping. The vehicle state is between the critical parking state and the worst parking state. The corresponding threshold and strategy calculation expressions are:

[0055]

[0056] b 1 =T l 0 +v 0 / a,b 2 =b 3 -v min / a

[0057]

[0058] b 1 ′=T l 0 +v 0 / a,b′ 2 =b 3 ′-v min / a

[0059] b 3 ′=T l 0 +t-(t 4 -v 0 / av min / a)

[0060] b′ 4 =T l ,b 5 ′=T l -L 1 / v

[0061] If the left-turning vehicle cannot find a passing strategy in the current cycle, it will stop before entering the road stop line and wait for the next cycle until it passes. The parking position calculation expression is:

[0062]

[0063] Where: L 0 It is the straight-line path length of the left-turning vehicle from the intersection entrance to the conflict point.

[0064] Preferably, the comprehensive consideration of travel time and parking delay to construct a vehicle multi-trajectory optimal strategy for trajectory planning specifically includes:

[0065] Taking into account the vehicle travel time and parking delay, the optimization goal is to minimize the total travel time of left-turning vehicles and optimize the target TT. 0 The function expression is:

[0066] TT 0 =min i,m TT i,m =min i,m {t i,m +(L 2,i +L) / ν_t i,m +Δt stop,i,m}

[0067]

[0068] Where: TT i,m is the actual travel time of the mth vehicle under the i-th type of trajectory; Δt stop,i,m Delay time for parking; L 1,i L is the adjustment distance from the detection area to the conflict point; 2,i is the curve distance from the conflict point to the stop line of the exit road; The length of the time window for left-turning vehicles is used to determine whether left-turning vehicles have enough time to pass through the intersection; v_t i,m For a left-turning vehicle at time t i,m The initial speed at time t represents the speed of the vehicle when it enters the current time window; b′ 1,i,m The time point when the left-turning vehicle completes deceleration from the initial speed ν_t i,m The time required to decelerate to the target speed; b' 2,i,m is the time point when the left-turning vehicle starts to accelerate; t s Time lost in starting the vehicle.

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] (1) Left-turn trajectory modeling is more realistic: Vehicle size and dynamic characteristics are introduced into trajectory planning modeling for optimization, making the trajectory closer to the actual operating status.

[0071] (2) More accurate and stable trajectory description: Based on the characteristics of the left-turn trajectory that the angle and speed gradually change and tend to be stable during the turning process, an exponential function is used to fit the trajectories of different types of vehicles, thereby achieving a more accurate and stable trajectory description.

[0072] (3) Accurate trajectory classification: Multi-index classification is performed based on speed, curvature, and steering angle to achieve accurate distinction between nearby crossing trajectories, normal crossing trajectories, and detour crossing trajectories.

[0073] (4) Speed ​​optimization is more flexible: By dynamically adjusting the speed range, vehicles can pass through intersections at the optimal speed, effectively reducing parking delays and traffic congestion.

[0074] (5) Strong multi-trajectory collaborative optimization capability: Utilizing the multi-trajectory interaction model, it can flexibly predict and respond to potential time-space conflict points, significantly improving the operational efficiency and safety of intersections. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 Optimize the strategy framework for left turns;

[0076] Figure 2 Layout for intersections;

[0077] Figure 3 Results of left-turn vehicle trajectory clustering and trajectory fitting;

[0078] Figure 4 Classify the interaction behavior of left-turn vehicles;

[0079] Figure 5 T i c ,T i a Schematic diagram of the construction process;

[0080] Figure 6 A schematic diagram of speed adjustment conditions for left-turning vehicles;

[0081] Figure 7 is the probability of trajectory distribution;

[0082] Figure 8 is the cumulative frequency of the velocity distribution. DETAILED DESCRIPTION

[0083] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0084] Example

[0085] like Figure 1 As shown, this embodiment provides a method for optimizing left-turn operation of an intelligent connected vehicle to jointly optimize the turning trajectory and turning speed of the vehicle, and the method includes:

[0086] The speed, curvature and turning angle of left-turning vehicles are extracted from the intersection vehicle driving video, and the left-turn driving trajectory is classified using a rule-based turning trajectory classification algorithm, and different types of left-turn driving trajectories are fitted;

[0087] Based on the spatiotemporal characteristics of multiple trajectories, a conflict point prediction model is constructed to predict conflict points;

[0088] Based on the conflict point prediction results, a variable speed adjustment strategy is used to generate the vehicle turning speed;

[0089] Based on different types of left-turn driving fitting trajectories and vehicle turning data, the optimal vehicle multi-trajectory strategy is constructed for trajectory planning by comprehensively considering travel time and parking delay.

[0090] Next, the method of this embodiment is introduced in detail.

[0091] The method of this embodiment is described in detail below.

[0092] 1. Classification of left-turn driving trajectories

[0093] The speed, curvature and turning angle indicators of left-turning vehicles are extracted from the vehicle driving video at the intersection. The left-turn driving trajectories are classified using a rule-based turning trajectory classification algorithm. The left-turn driving trajectories (i.e., left-turn behaviors) are divided into three categories: nearest crossing trajectories, ordinary crossing trajectories and detour crossing trajectories.

[0094] This embodiment collects the left-turn vehicle driving trajectory data at a certain intersection in a certain city. Figure 2 First, a plane rectangular coordinate system is established in the unsignalized intersection. Figure 4 The characteristic information of various left-turning vehicle behaviors at the intersection is summarized, including the characteristics and thresholds of trajectory attributes, where the speed indicates the severity of the conflict between the turning vehicle and the oncoming straight vehicle. The curvature k and abnormal coordinates indicate that the vehicle has detoured during the turning process, and the table shows that the maximum k is the maximum value when the turning trajectory is a circular curve. 0 is 0.05. Turning angle Describes the change in angle during the entire turning process. This embodiment uses this indicator to distinguish the types of turning trajectories. Based on the analysis of the turning trajectory of the vehicle in the video, the maximum turning angle β of the vehicle is max =130°, and this angle occurs at the conflict point. i ,b i ,c are line segments |BC i |,|AC i |, the length of AB|.

[0095] According to the actual trajectory before the conflict point, a rule-based turning trajectory classification algorithm is used to classify the complete left-turn driving trajectory, including:

[0096] For the curvature k i , speed v i and the turning angle β i , set the maximum curvature k 0 , parking speed v stop and the maximum turning angle β 0 :

[0097] 1) If there exists k i ∈(-∞,0)∪(k 0 ,+∞) and Q C ={(x i ,y i )∈C i |x i <x Q ,y i <y Q}≠{}, it is determined to be a nearby crossing trajectory, Q C Indicates that the point of conflict (x Q ,y Q ) The vehicle trajectory point set at the lower left; x Q ,y Q is the horizontal and vertical coordinates of the conflict point Q: if there is v i ≤v stop , it is determined to be a parking and nearby crossing trajectory; if all v i >v stop , it is determined as a non-stop nearby crossing trajectory;

[0098] 2) If all k i ∈[0,k 0 ] or Q C ={(x i ,y i )∈C i |x i <x Q ,y i <y Q}={}, and there is β i <β 0 , it is determined to be a normal crossing trajectory: if there is v i ≤v stop , it is determined to be a parking ordinary crossing trajectory; if all v i >v stop , it is determined as a non-parking ordinary crossing trajectory;

[0099] 3) If all k i ∈[0,k0 ] or Q C ={(x i ,y i )∈C i |x i <x Q ,y i <y Q}={}, and all β i ≥β 0 , it is determined to be a detour crossing trajectory: if there is v i ≤v stop , it is determined to be a parking detour crossing trajectory; if all v i >v stop , it is determined to be a non-stop detour crossing trajectory.

[0100] The left-turn driving trajectory classification results and fitting curves are shown in Figure 3 shown.

[0101] In this embodiment, 585 straight driving trajectories and 64 left-turn driving trajectories are effectively extracted from the 100-minute intersection vehicle driving video, where the number of three types of left-turn driving trajectories is 6, 40, and 18 respectively. The functions of the three types of left-turn curves fitted by exponential distribution are as follows:

[0102] Nearby crossing track h 1 (x):y=286.6e -0.19x +21.03

[0103] Normal crossing trajectory h 2 (x):y=906.8e -0.26x +19.33

[0104] Detour through trajectory h 3 (x):y=2.35×10 9 e -1.28x +19.38

[0105] The analysis of these three types of trajectories shows that the common crossing trajectories are concentrated in the middle area of ​​the intersection and change relatively smoothly, such as Figure 3 The middle trajectory in . When the oncoming straight traffic flow is small, since drivers tend to turn earlier to reduce the driving distance, the turning time is earlier and closer to the stop line. This trajectory is the nearest crossing trajectory. When the oncoming straight traffic flow is large, in order to avoid collision or reduce parking time, the turning vehicle will turn later than the normal crossing trajectory, that is, when it approaches the intersection, it will go straight for a distance first, and when there is a traversable gap in the object straight traffic flow, the driver will quickly turn the steering wheel to complete the turn. This turning trajectory is the bypass crossing trajectory.

[0106] 2. Conflict point prediction

[0107] Based on the spatiotemporal characteristics of multiple trajectories, a conflict point prediction model is established to improve the ability to flexibly respond to potential traffic conflicts.

[0108] In the non-signal control scenario, this embodiment assumes that the CAV vehicles are of the same size and ignores the impact of pedestrians and non-motor vehicles. The control center designs a passing strategy based on the status of left-turning vehicles and adjusts their speed before the potential stop line, while straight-moving vehicles have absolute priority.

[0109] The multi-trajectory set contains different spatial conflict points, which significantly improves the probability of finding an acceptable gap compared to the traditional fixed trajectory. The key time nodes describing the left-turning vehicle in the process of passing through the intersection include: t 1 - Left turning vehicle from current speed v 0 Slow down to the target speed v min The time required to describe the first phase of a left-turning vehicle's speed adjustment to ensure that the vehicle can enter the intersection safely; t 2 -The time when the left-turning vehicle starts to accelerate again after deceleration; t 3 - Left turning vehicles from v min Accelerate to target speed v max The time required to describe the phase in which a vehicle accelerates back to normal speed when passing through an intersection; t 4 - The time required for a left-turning vehicle to travel from its initial position to the conflict point is used to determine the conflict window of the vehicle, which, combined with the time window of the straight-moving vehicle, optimizes the safety and efficiency of the intersection; 5 -The total time required for left-turning vehicles to enter the intersection and completely leave the intersection is used to calculate the total time that left-turning vehicles occupy the intersection, providing a basis for multi-vehicle coordination and path optimization.

[0110] This embodiment proposes a conflict point prediction model suitable for straight and left-turn trajectories based on the time-space dimension, as shown in formula (1). The width and length of the vehicle are used to ensure spatial safety, and time is used as the time dimension of safety quantification:

[0111]

[0112] Where: v s is the speed of the straight-moving vehicle, x s ,y s v is the horizontal and vertical coordinates of the straight-moving vehicle; l is the speed of the left-turning vehicle, x l ,y l is the horizontal and vertical coordinates of the left-turning vehicle; t is the predicted conflict time; T s 0 Adjust the time zone for the speed of the straight-moving vehicle, Tl 0 is the speed adjustment time zone for left-turning vehicles; B is the width of the vehicle, in this embodiment B = 1.6, unit: meter; L is the length of the vehicle, in this embodiment L = 4, unit: meter; S 0 is the safety distance before the detection area. In this embodiment, S 0 =50, unit: meter; S l is the safety distance behind the detection area. In this embodiment, S l =40, unit: meter; x in is the horizontal coordinate of the point where the vehicle enters the conflict, x out is the horizontal coordinate of the vehicle leaving the conflict point. In this embodiment, x in =14.18,x out =32, unit: meter; h' is the path slope of the left-turning vehicle trajectory, which indicates the degree of inclination of the trajectory at a certain point, and indicates the rate of change of the direction of the path at the current point; g(t) is the dynamic path function of the left-turning vehicle at time t; L 0 is the distance from the conflict point to the stop line of the exit road. In this embodiment, L 0 =15, unit: meter; x c0 and x c1 are the horizontal coordinates of the left and right boundaries of the lane where the straight-moving vehicle is located in the intersection. In addition, the embodiment also sets a safety vehicle body width B' (B'>B) and a safety vehicle body length L' (L'>L).

[0113] A. Conflict time window:

[0114] According to the distribution characteristics of the straight track and its lateral fluctuation, the straight track is distributed at x c0 With x c1 In this embodiment, the horizontal axis range is set to [18, 21]. The conflict time window records the speed v of the straight vehicle. s , the speed adjustment time zone T for vehicles going straight and turning left s 0 , T l 0 , predict the conflict time t, and its corresponding conflict point coordinates (x l ,y l ). When no conflict point is predicted, t and (x l ,y l ) is empty. Set the time increment Δt to 0.1s, and the maximum travel time for going straight through the intersection It can be calculated by the formula:

[0115]

[0116] The following is the time window T used to search for conflicts: iC and the accessible time window T i a Design, Figure 5 The specific construction process of the time window is shown, including:

[0117] 1) Initialization: Set the time increment Δt = 0.1s and initialize the conflict time set T c ={}.

[0118] 2) Search loop:

[0119] 21) Set the time increment Δt and initialize the conflict time set T c ={};

[0120] 22) Time for each left-turn vehicle Perform the following steps:

[0121] Time for each straight vehicle in And R s ={(T s ,k 0 ,v s ,k)},k=1,2,...:for time t∈[max(T l 0 ,T s ,j), is the maximum travel time for turning left through an intersection, is the maximum travel time for going straight through the intersection, if f S (x s ,y s ,v s ,t)=f L (x l ,y l ,v l ,t), then add the corresponding solution to the conflict time set T c =T c ∪{v s ,j,T s ,j 0 ,T l 0 ,t,x l ,y l}, time t is updated to t+Δt;

[0122] 3) Return: Return the conflict time set T c .

[0123] B. Time security verification:

[0124] Combined with the passable time window Ti a The length Δt m a And the safety time length Δt safe , when the following conditions are met, all accessible time windows T are given i a′ ∈{T a′}:

[0125]

[0126] Where, c = 1.5s;

[0127] If in the current cycle If a passable time window cannot be found within the specified time period, the left-turning vehicle will stop at the potential stop line and wait for the next passable time window search result.

[0128] Speed ​​adjustment strategy: Design a dynamic speed adjustment strategy to optimize the vehicle acceleration and deceleration process according to different trajectory types and traffic flow characteristics, and generate the optimal traffic plan.

[0129] 3. Optimize left-turn vehicle speed

[0130] Taking travel time and parking delay into comprehensive consideration, the objective function is designed to jointly optimize the trajectory and speed of the vehicle to achieve a coordinated improvement in efficiency and safety.

[0131] In order to ensure that left-turning vehicles can pass through the intersection at an appropriate variable speed and minimize delays, a vehicle turning speed adjustment range is designed. This range includes the maximum turning speed v 0 (35km / h), minimum speed v min (10km / h) and acceleration / deceleration a(3.5m / s 2 )

[0132] The ideal vehicle turning strategy is that the vehicle has an initial speed v 0 Next, at time T l =T l 0 +t 1 When the motion state needs to be adjusted, time node b 1 ,b 2 ,b 3 They represent the time points of completing deceleration, starting acceleration, and completing acceleration. Speed ​​adjustment must be completed before the conflict point, and the final speed v f Should be in the interval [v max ,v 0 ], and the time to reach the conflict point is T l Should be in the interval

[0133] According to the different vehicle motion states, the speed adjustment strategies are divided into the following three types: incomplete deceleration, complete deceleration and parking strategy. The strategy set φ contains the following eight key indicators:

[0134] T l 0 --- Initial time; b 1 ′---time to complete deceleration; b 2 ′---time to start acceleration; b 3 ′---time to complete acceleration; b 4 ′---time to reach the stop line; b 5 ′---the time to reach the conflict point; v 0 ---initial velocity; v f ---Final velocity.

[0135] like Figure 6 As shown in the figure, the speed adjustment types specifically include:

[0136] Type 1: Incomplete deceleration:

[0137] If T l ∈{T′ a} and T l 0 +t 1 <T l ≤T l 0 +t 3 , the vehicle needs to be at time T l 0 Start to decelerate until reaching the final speed v f (≥v min ). This state is between the optimal state (marked as ①) and the critical state (marked as ③), such as Figure 5 As shown in ②.

[0138] Under this condition, the threshold and strategy φ are calculated as follows:

[0139]

[0140] Type 2: Complete deceleration:

[0141] If T l ∈{T a ′} and T l 0 +t 3 <T l <T l 0 +t 4 , the vehicle needs to be at time T l0 Start to decelerate until time b 1 ', but does not stop completely; it then accelerates immediately until it reaches the final speed v min This state is between the critical state ③ and the optimal parking state ④.

[0142] Under this condition, the threshold and strategy φ calculation expressions are:

[0143]

[0144] Type 3: Parking strategy:

[0145] If T l ∈{T a '}and The vehicle needs to slow down to a complete stop and wait for a while after stopping. This state is between the critical parking state ④ and the worst parking state ⑤. In this case, the calculation expression of the threshold and strategy φ is:

[0146]

[0147] If the left-turning vehicle cannot find a passing strategy in this cycle, it needs to stop before the stop line of the road and wait for the next cycle until it passes. The parking position calculation expression is:

[0148]

[0149] 4. Joint Optimization:

[0150] Taking travel time and parking delay into comprehensive consideration, the objective function is designed to jointly optimize the trajectory and speed of the vehicle to achieve a coordinated improvement in efficiency and safety.

[0151] Based on comprehensive consideration of vehicle travel time and parking delay, the optimization objective function TT is constructed in this embodiment. 0 , in order to minimize the total travel time of left-turn vehicles, the objective function is expressed as:

[0152]

[0153] Among them, TT i,m is the actual travel time of the mth vehicle under the i-th type of trajectory; Δt stop,i,m Delayed parking time.

[0154] The vehicle driving distance consists of two parts: the adjustment distance L from the detection area to the conflict point 1,i The curve distance L from the conflict point to the exit road stop line 2,i .

[0155] In order to compare the performance of the multi-trajectory model proposed in this paper and the traditional fixed trajectory (normal trajectory) model in terms of left-turn traffic efficiency, this embodiment constructs three headway distribution scenarios of straight-moving vehicles to describe the arrival time of oncoming straight-moving vehicles, specifically:

[0156] 1) Shifted negative exponential distribution (aS)

[0157] This distribution is often used to describe the scenario of small traffic arriving randomly without overtaking. Its average headway is 3.6s and the minimum headway τ is 2.0s, which ensures the safety of the vehicle following state.

[0158] 2) Uniform distribution (bU)

[0159] The headway time range of this distribution is [2.0s, 5.6s], which is used to describe the scenario of random arrival of large traffic, in which the impact of left-turning vehicles on straight traffic is small.

[0160] 3) Pulse distribution (cP)

[0161] This distribution is composed of two uniform distributions, where the headway intervals are [2.0s, 5.6s] and [6.0s, 10.0s]. In this embodiment, the straight traffic generated in the previous interval accounts for 40%60%. This distribution usually describes the periodic alternation of dense and sparse arrivals of straight vehicles in time when upstream and downstream intersections are controlled by signals.

[0162] Through these three different headway distributions, this embodiment simulates the impact of opposite through traffic on the traffic efficiency of left-turn vehicles under different arrival modes. The construction of the simulation scenario aims to: A. verify the adaptability of the multi-trajectory model in different traffic scenarios; B. compare the performance differences between the traditional fixed trajectory model and the multi-trajectory model in optimizing left-turn travel time and reducing the number of stops.

[0163] (1) Vehicle delay time in different scenarios

[0164] In Table 2, the relative optimization effects of the three scenarios are shown in the average driving time T real and parking rate p stop Overall, the optimization ranges of the three scenarios are approximately 10% and 26.1%, respectively. Among them, the first scenario aS shows the most significant optimization effect. This embodiment uses the following formula to define delay to quantify the improvement of these two indicators:

[0165]

[0166] Among them, p i is the selection frequency of the i-th trajectory; T i,real ,T i,ideal is the average travel time of the ith trajectory under actual and ideal conditions; p stop is the parking frequency; t s =2s is the starting loss time.

[0167] In the aS scenario, the delay was reduced from 8.39s to 6.73s. The delay optimization of the multi-trajectory model reached 19.8% compared with the fixed-trajectory model. The following table 1 is a detailed summary of the test set optimization results:

[0168] Table 1

[0169]

[0170] (2) Trajectory and speed adjustment results

[0171] Further analysis shows that under different arrival distribution conditions, the frequency distribution of the selected trajectories is basically consistent with the actual situation, as shown in the figure below. Most vehicles choose the normal trajectory and the yield trajectory, while relatively few vehicles choose the rush trajectory. This phenomenon shows that when entering the time window, left-turning vehicles usually cannot accelerate enough to pass the conflict point first, so they need to wait for the next passable time window, at which time vehicles often arrive at the detour conflict point and the normal conflict point in turn.

[0172] Stopping will cause large delays, so drivers usually avoid this trajectory. In contrast, the conflict points of other trajectories are relatively far away, giving the vehicle enough time to gradually approach the upcoming time window at a lower speed without stopping.

[0173] Figure 8 The cumulative frequency distribution of left turn speed in the aS scenario is shown. The results show that:

[0174] The final speeds of 90% of the rush-through trajectories and normal crossing trajectories are concentrated in the intervals of [10.0, 19.0] m / s and [12.0, 22.0] m / s, respectively; the final speed of the detour crossing trajectory is concentrated in the interval of [17.0, 30.0] m / s.

[0175] This result shows that the multi-trajectory model covers the entire feasible range of speed adjustment, making speed adjustment more flexible and reasonable.

[0176] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for optimizing left-turn operation of an intelligent connected vehicle, characterized in that: include: The speed, curvature and turning angle of left-turning vehicles are extracted from the intersection vehicle driving video, and the left-turn driving trajectory is classified using a rule-based turning trajectory classification algorithm, and different types of left-turn driving trajectories are fitted; Based on the spatiotemporal characteristics of multiple trajectories, a conflict point prediction model is constructed to predict conflict points; Based on the conflict point prediction results, a variable speed adjustment strategy is used to generate the vehicle turning speed; Based on different types of left-turn driving fitting trajectories and vehicle turning data, the optimal vehicle multi-trajectory strategy is constructed for trajectory planning by comprehensively considering travel time and parking delay.

2. The method for optimizing left-turn operation of an intelligent connected vehicle according to claim 1, characterized in that: The fitting of different types of vehicle running trajectories is specifically: fitting different types of vehicle running trajectories using exponential functions.

3. The method for optimizing left-turn operation of an intelligent connected vehicle according to claim 1, characterized in that: The categories of the left-turn driving trajectory include a nearby crossing trajectory, a normal crossing trajectory and a detour crossing trajectory.

4. The method for optimizing left-turn operation of an intelligent connected vehicle according to claim 3, characterized in that: According to the actual trajectory before the conflict point, the left-turn driving trajectory is classified using a rule-based turning trajectory classification algorithm, specifically: For the curvature k i , speed v i and the turning angle β i , set the maximum curvature k0 and parking speed v stop And the maximum turning angle β0: 1) If there exists k i ∈(-∞,0)∪(k0,+∞) and Q C ={(x i ,y i )∈C i |x i <x Q ,y i <y Q }≠{}, it is determined to be a nearby crossing trajectory, Q C Indicates that the point of conflict (x Q ,y Q ) The vehicle running trajectory point set at the lower left; x Q ,y Q is the horizontal and vertical coordinates of the conflict point Q: if there is v i ≤v stop , it is determined to be a parking and nearby crossing trajectory; if all v i >v stop , it is determined as a non-stop nearby crossing trajectory; 2) If all k i ∈[0,k0] or Q C ={(x i ,y i )∈C i |x i <x Q ,y i <y Q }={}, and there is β i <β0, it is judged as a normal crossing trajectory: if there is v i ≤v stop , it is determined to be a parking ordinary crossing trajectory; if all v i >v stop , it is determined as a non-parking ordinary crossing trajectory; 3) If all k i ∈[0,k0] or Q C ={(x i ,y i )∈C i |x i <x Q ,y i <y Q }={}, and all β i ≥β0, it is determined to be a detour trajectory: if there is v i ≤v stop , it is determined to be a parking detour crossing trajectory; if all v i >v stop , it is determined to be a non-stop detour crossing trajectory.

5. The method for optimizing left-turn operation of an intelligent networked vehicle according to claim 4, characterized in that: Based on the spatiotemporal characteristics of multiple trajectories, a conflict point prediction model is established to predict the conflict points. The conflict point prediction model is specifically as follows: Where: v s is the speed of the straight-moving vehicle, x s ,y s v is the horizontal and vertical coordinates of the straight-moving vehicle; l is the speed of the left-turning vehicle, x l ,y l is the horizontal and vertical coordinates of the left-turning vehicle; t is the predicted conflict time; Adjust the time zone for the speed of the straight-moving vehicle, is the speed adjustment time area for left-turning vehicles; B is the vehicle width; L is the vehicle length; S0 is the safety distance before the detection area, S l is the safety distance behind the detection area; x in is the horizontal coordinate of the point where the vehicle enters the conflict, x out is the horizontal coordinate of the vehicle leaving the conflict point; h' is the path slope of the left-turning vehicle trajectory, which indicates the degree of inclination of the trajectory at a certain point and the rate of change of the direction of the path at the current point; g(t) is the dynamic path function of the left-turning vehicle at time t; L0 is the distance from the conflict point to the stop line of the exit road; x c0 and x c1 They are the horizontal coordinates of the left and right boundaries of the lane where the through vehicle is located in the intersection.

6. A method for optimizing left-turn operation of an intelligent networked vehicle according to claim 5, characterized in that: Use the conflict time window to record the speed v of the straight-moving vehicle s , Speed ​​adjustment time zone for straight-moving vehicles Speed ​​adjustment time zone for left-turning vehicles Predict the conflict time t and its corresponding conflict point coordinates (x l ,y l ); The conflict time window and the accessible time window The specific search process is: 1) Set the time increment Δt and initialize the conflict time set T c ={}; 2) Time for each left-turn vehicle Perform the following steps: For each straight-moving vehicle, time T s , Where T s , And R s ={(T s ,k 0 ,v s ,k)},k=1,2,...:for time is the maximum travel time for turning left through an intersection, is the maximum travel time for going straight through the intersection, if f S (x s ,y s ,v s ,t)=f L (x l ,y l ,v l ,t), then the corresponding solution is added to the conflict time set Time t is updated to t+Δt; 3) Return the conflict time set T c .

7. The method for optimizing left-turn operation of an intelligent networked vehicle according to claim 6, characterized in that: It also includes time security verification, specifically: Combined with the passable time window Length And the safety time length Δt safe , when the following conditions are met, all the passable time windows are given Where: c is the constant coefficient set; x c0 With x c1 is the horizontal coordinate boundary of the straight trajectory distribution; If in the current cycle If a passable time window cannot be found within the specified time period, the left-turning vehicle will stop at the potential stop line and wait for the next passable time window search result.

8. The method for optimizing left-turn operation of an intelligent connected vehicle according to claim 6, characterized in that: The use of a variable speed adjustment strategy to generate a vehicle turning speed specifically includes: Construct a vehicle turning speed adjustment interval, wherein the parameters of the vehicle turning speed adjustment interval include the maximum turning speed v0, the minimum turning speed v min and acceleration a; When the motion state needs to be adjusted, time nodes b1, b2, and b3 represent the time points of completing deceleration, starting acceleration, and completing acceleration respectively; speed adjustment must be completed before the conflict point, and the final speed v f Should be in the interval [v max ,v0], and the time to reach the conflict point is T l Should be in the interval t1 is the time when the left-turning vehicle decelerates from the current speed v0 to the target speed v min Time required; According to the vehicle's motion state, the speed adjustment strategy is divided into incomplete deceleration strategy, complete deceleration strategy and parking strategy. The indicator parameters involved include initial time The time to complete deceleration b1′, the time to start acceleration b2′, the time to complete acceleration b3′, the time to reach the stop line b4′, the time to reach the conflict point b5′, the initial speed v0 and the final speed v f .

9. The method for optimizing left-turn operation of an intelligent networked vehicle according to claim 8, characterized in that: The speed adjustment strategy specifically includes: Incomplete deceleration: If T l ∈{T a '}and The vehicle must be in time Start to decelerate until reaching the final speed v f , v f ≥v min , the vehicle state is between the optimal state and the critical state, and the corresponding threshold and strategy calculation expressions are: t3=v0 / v min ·[t1-v0 / 2a·(1-v min / v0) 2 ] b4′=T l ,b5′=T l -L1 / v Where: t3 is the time from v to v min Accelerate to target speed v max Time required; Complete deceleration strategy: If T l ∈{T a '}and The vehicle must be in time It starts to decelerate until time b1′, but does not stop completely, and then immediately accelerates until it reaches the final speed v min , the vehicle state is between the critical state and the optimal parking state, and the corresponding threshold and strategy calculation expressions are: b4′=T l ,b5′=T l -L1 / v Where: t4 is the time required for the left-turning vehicle to travel from the initial position to the conflict point; Parking strategy: If T l ∈{T a '}and The vehicle needs to slow down to a complete stop and wait for a period of time after stopping. The vehicle state is between the critical parking state and the worst parking state. The corresponding threshold and strategy calculation expressions are: b′4=T l ,b5′=T l -L1 / v If the left-turning vehicle cannot find a passing strategy in the current cycle, it will stop before entering the road stop line and wait for the next cycle until it passes. The parking position calculation expression is: Where: L0 is the straight path length of the left-turning vehicle from the intersection entrance to the conflict point.

10. The method for optimizing left-turn operation of an intelligent connected vehicle according to claim 1, characterized in that: The method comprehensively considers travel time and parking delay, constructs a vehicle multi-trajectory optimal strategy for trajectory planning, and specifically includes: Taking into account the vehicle travel time and parking delay, the optimization goal is to minimize the total travel time of left-turning vehicles. The function expression of the optimization goal TT0 is: TT0=min i,m TT i,m =min i,m {t i,m +(L 2,i +L) / ν-t i,m +Δt stop,i,m } Where: TT i,m is the actual travel time of the mth vehicle under the i-th type of trajectory; Δt stop,i,m Delay time for parking; L 1,i is the adjustment distance from the detection area to the conflict point; L 2,i is the curve distance from the conflict point to the stop line of the exit road; The length of the time window for left-turning vehicles is used to determine whether left-turning vehicles have enough time to pass through the intersection; v_t i,m For a left-turning vehicle at time t i,m The initial speed at time t represents the speed of the vehicle when it enters the current time window; b' 1,i,m The time point when the left-turning vehicle completes deceleration from the initial speed ν_t i,m The time required to decelerate to the target speed; b' 2,i,m is the time point when the left-turning vehicle starts to accelerate; t s Time lost in starting the vehicle.

Images