Collaborative optimization method for entrance and exit of high-speed interchange under mixed traffic

Abstract

The invention relates to a high-speed interchange entrance and exit collaborative optimization method under mixed traffic, and the method comprises the steps: achieving the digital mapping of vehicles in an interchange region at an edge cloud; identifying a lane changing intention of a human-driven vehicle by using the steering lamp information and deducing a navigation path; recognizing a deceleration lane changing intention, dynamically distributing lane changing gaps, and constructing order constraints; carrying out passing order optimization and long time domain trajectory planning under a full network connection vehicle hypothesis; constructing a local game group for a human driving vehicle without a deceleration lane changing intention, solving a control strategy of an intelligent network connection vehicle based on a potential game model, and inhibiting a lane changing behavior which is not beneficial to the overall efficiency; when the track deviation is large, re-planning is carried out; a safety constraint is constructed by using short-time speed prediction of a human-driven vehicle, and a short-time-domain control instruction of the intelligent network-connected vehicle is generated based on long-time-domain trajectory planning. Therefore, the problems that in the prior art, straight-going vehicles cannot be used for cooperatively changing lanes to release the traffic capacity of a main road, the modeling and response capacity for uncertainty of human driving behaviors is lacked, and the traffic efficiency of an interchange road section in a complex traffic environment is low are solved.

Description

Technical Field

[0001] The present application relates to the field of intelligent connected vehicle planning and control technology, and in particular to a method for collaborative optimization of high-speed interchange entrances and exits under mixed traffic conditions. Background Art

[0002] As crucial hubs connecting different expressways, highway interchanges typically consist of adjacent ramp diverging and merging areas. These areas, due to the frequent need for vehicles to change lanes, are typical traffic bottlenecks on highways. Vehicles entering ramps from the main road, or merging from ramps into the main road, must change lanes within a limited space, which can easily lead to right-of-way conflicts, resulting in vehicle deceleration and even increased collision risk. In severe cases, this can cause traffic congestion. Furthermore, the diverging and merging areas are often coupled, and congestion in one area can quickly spread to the entire interchange, significantly reducing overall traffic efficiency.

[0003] In related technologies, a multi-vehicle collaborative optimization scheduling method is adopted for a highway interchange section including a ramp diverging area and a ramp merging area to solve the congestion problem caused by vehicle right-of-way conflicts in highway interchange scenarios.

[0004] However, in related technologies, the multi-vehicle collaborative scheduling method for optimizing the diverging and merging areas of highway interchanges fails to utilize the potential for collaborative lane changing by straight-moving vehicles on the main road, thereby limiting the further release of the main road's traffic capacity. At the same time, it does not consider how to adjust when vehicles do not drive according to the cloud-based optimization results. It lacks the ability to model and respond to the uncertainty of human driving behavior, making it difficult to apply to mixed traffic scenarios in reality where intelligent connected vehicles and human-driven vehicles coexist, and urgently needs to be improved. Summary of the Invention

[0005] This application provides a method for collaborative optimization of highway interchange entrances and exits under mixed traffic conditions to address the problems of related technologies failing to utilize collaborative lane changes of through-going vehicles to release the main road's capacity and lacking the ability to model and respond to the uncertainty of human driving behavior, resulting in low traffic efficiency of interchange sections in complex traffic environments.

[0006] The present application provides a method for collaborative optimization of highway interchange entrances and exits under mixed traffic, including the following steps: Step S1: Implementing digital mapping of vehicles in the interchange area on an edge cloud, recording the two main roads in the ramp divergence area as main roads M2 and M1 from the inside out, and recording the two main roads in the ramp merging area as main roads M4 and M3 from the inside out, respectively. The ramp and the auxiliary lanes connecting the main roads M1 and M3 at both ends are combined and recorded as lane R;

[0007] Step S2: Classifying the lane change intention of the human-driven vehicle based on the turn signal information of the human-driven vehicle. Human-driven vehicles with their turn signals on on the main roads M3 and M4 are both classified as having a straight lane change intention. Human-driven vehicles with their right turn signals on on the main road M2 are classified as having a diverging lane change intention, and their navigation path is assumed to be that of a diverging vehicle about to enter lane R. Furthermore, human-driven vehicles with their left turn signals on on the main road M1 are classified as having a straight lane change intention, and their navigation path is assumed to be that of a straight-moving vehicle traveling along the main road M1.

[0008] Step S3: Using the vehicle information aggregated from the cloud, the human-driven vehicle is treated as an intelligent connected vehicle that follows cloud-based planning recommendations, and multi-vehicle collaborative order optimization decision-making and trajectory planning are performed under the assumption of fully connected vehicles. During the collaborative decision-making and planning under the assumption of fully connected vehicles, the navigation path information of the intelligent connected vehicle is known, while the navigation path information of the human-driven vehicle is unknown. Human-driven vehicles on the main road M2 with no intention of diverging or changing lanes are treated as straight-moving vehicles traveling along the main road M2, and human-driven vehicles on the main road M1 with no intention of going straight or changing lanes are treated as diverging vehicles that will enter lane R.

[0009] Step S4: For the human-driven vehicle with the diverging lane-changing intention and the straight lane-changing intention, analyze whether it decelerates to search for a lane-changing gap in the adjacent lane. If it decelerates to search for a lane-changing gap in the adjacent lane, assign the nearest lane-changing gap behind the adjacent lane to the human-driven vehicle. The suggested lane-changing gap is recorded as (GAP_P, GAP_F), where GAP_P represents the vehicle in front of the lane-changing gap and GAP_F represents the vehicle behind the lane-changing gap. If the result is inconsistent with the order decision of step S3, it is recorded as an order constraint for replanning the passage order of each vehicle.

[0010] Step S5: Calculate the suggested lane change gaps (GAP_P, GAP_F) for other human-driven vehicles and construct a game group. For vehicles without a clear lane change intention, the lane change gap that causes the least delay in the target lane is calculated as the suggested lane change gap. For vehicles with a straight lane change intention but not slowing down to find a lane change gap in the adjacent lane, the suggested lane change gap is extracted from the sequential decision in step S3.

[0011] When the preceding vehicle CAV_P in the adjacent lane and the following vehicle CAV_F in the adjacent lane are both intelligent connected vehicles, the human-driven vehicle HDV and the preceding vehicle PV in the current lane, the preceding vehicle CAV_P in the adjacent lane, and the following vehicle CAV_F in the adjacent lane are constructed into a game group, which is recorded as {PV, HDV, CAV_P, CAV_F};

[0012] If (CAV_P, CAV_F) is different from (GAP_P, GAP_F), a potential game problem is constructed to solve the control strategy of CAV_P and CAV_F to reduce the front-to-back gap and constrain the human driver to not cut in between (CAV_P, CAV_F);

[0013] Step S6: When the following conditions are met: (1) the lane change gap is allocated to the human-driven vehicle in step S4; (2) the game problem is constructed and solved in step S5; (3) the deviation between the actual driving trajectory of the human-driven vehicle and the planning result of step S2 is greater than a preset condition, the collaborative decision-making and planning under the assumption of the fully connected vehicle in step S3 are re-performed; wherein the order decision result meets the order constraint of step S4, and the long-term speed planning result of the intelligent connected vehicle can simultaneously realize the control strategy calculated in the game problem in step S5;

[0014] Step S7: In each preset control cycle, the historical speed curve of the human-driven vehicle is used to predict the speed curve of the human-driven vehicle in a future control cycle, and the long-term planning result calculated in step S2 or step S6 is used as a reference target. Safety constraints are formed according to the prediction results of the human-driven vehicle, and the short-term control strategy of the intelligent connected vehicle is calculated.

[0015] In the embodiment of the present application, digital mapping of vehicles in the interchange area can be achieved through the edge cloud to determine lane information, and the lane change intention of human-driven vehicles can be identified based on the lane information. The cloud then regards the human-driven vehicles as intelligent connected vehicles that follow planning suggestions, and performs multi-vehicle collaborative order optimization decisions and trajectory planning under the assumption of full connectivity. By accurately grasping the lane status through digital mapping, combining lane change intention identification with collaborative decision-making under the assumption of full connectivity, a more efficient lane change coordination mechanism can be introduced between intelligent connected vehicles, and the traffic strategy of intelligent vehicles can be dynamically adjusted in the presence of human-driven vehicles, thereby improving the operating efficiency and traffic safety of the entrances and exits of high-speed interchanges.

[0016] Optionally, in one embodiment of the present application, the step S3 performs multi-vehicle collaborative sequence optimization decision-making and trajectory planning under the assumption of fully connected vehicles, including:

[0017] A certain length of the main road M1, the main road M2, the main road M3, and the main road M4 upstream of the ramp merging and ramp diverging areas is used as the overall planning area, and a rolling traversal planning is performed using triples to consider the lane change coordination of straight-moving vehicles;

[0018] The vehicles on the four lanes of the main road M1, the main road M2, the main road M3, and the main road M4 are recorded in order as {M1_1, M1_2, …, M1_p}, {M2_1, M2_2, …, M2_q}, {M3_1, M3_2, …, M3_n}, {M4_1, M4_2, …, M4_m}, {M1_1, M2_1, M3_1} is taken as the first triplet, and {M1_1, M2_1, M3_2} is taken as the second triplet;

[0019] For the triplet formed by M1_1, M2_1, and M3_1, calculate the vehicle that has priority when M3_1 does not change lanes to M4. This vehicle is denoted as G1, and the resulting travel time delay cost3_1 for subsequent vehicles of M1, M2, and M3 is given.

[0020] For the triple formed by M1_1, M2_1, and M3_2, calculate the vehicles that have priority in the merging area when M3_1 changes lanes to M4, and give the travel time delay cost3_2 for subsequent vehicles on the main roads M1, M2, and M3 caused by this. At the same time, calculate the feasible lane-changing gap in which M3_1 minimizes the travel time delay for vehicles on the main road M4. Assume that M3_1 should change lanes before M4_i, and the corresponding travel time delay caused to subsequent vehicles on the main road M4 is cost4_i.

[0021] Compare cost3_1 with cost3_2 + cost4_i. When cost3_1 > cost3_2 + cost4_i, M3_1 switches to the lane-changing gap before M4_i. Calculate M3_1's future lane-changing trajectory and speed plan. M3_2 and M3_3 are added to form new triplet number one and triplet number two, respectively.

[0022] When cost3_1<=cost3_2+cost4_i, M3_1 does not change lanes. If the navigation path information of M1_1 is to enter the main road M3 via R, G1 is given priority, and its future speed plan and lane change trajectory are calculated. The vehicle behind G1 is added to form a new triplet number one and triplet number two. If the navigation path information of M1_1 is to go straight along the main road M1, assuming that M1_1 changes lanes before M2_1, M1_2, M1_1, and M3_1 are used as triplet number three, and the vehicle that should have priority among M1_1 and M3_1 is calculated. The travel time delay cost1 caused to the subsequent vehicles on the main roads M1 and M3 at this time is given. At the same time, the travel time delay cost2 caused to the vehicles on the main road M2 before M1_1 changes lanes to M2_1 is calculated.

[0023] Compare cost3_1 and cost1+cost2. When cost3_1 <= cost1+cost2, M1_1 does not change lanes, giving G1 priority. Its future speed plan and lane-changing trajectory are calculated, and the vehicle behind G1 is added to form new triplet 1 and triplet 2. When cost3_1 > cost1+cost2, M1_1 changes lanes before M2_1. M1_1's future lane-changing trajectory and speed plan are calculated, and M1_2 is added to form new triplet 1 and triplet 2.

[0024] Repeat the above steps until the lane change trajectory and speed planning are calculated for all vehicles in the first planning area.

[0025] The embodiment of the present application can realize digital modeling of the physical traffic environment through the edge cloud, accurately extract lane information, and provide real-time and accurate environmental data support for subsequent lane change intention recognition and collaborative decision-making, thereby improving the adaptability of the entire optimization method to actual traffic conditions.

[0026] Optionally, in one embodiment of the present application, the step S4 of analyzing whether the human driver has slowed down to find a lane-changing gap in an adjacent lane includes:

[0027] The longitudinal driving behavior of the human-driven vehicle is considered as a control system. The deceleration behavior of the human-driven vehicle to find a lane-changing gap in the adjacent lane is determined by comparing the actual output of the control system with the estimated output. The input of the system is the distance between the ego vehicle and the initial position of the preceding vehicle, and the speed of the preceding vehicle. The output of the system is the acceleration, speed, and position of the human-driven vehicle.

[0028] The edge cloud collects the historical status data of the human-driven vehicle and the vehicle in front of it with a preset length of T in advance to form the feature matrix U p ,U f ,Y p ,Y f :

[0029]

[0030] Where u(i)=col(v l (i),p l (i)-p h (1)), y(i) = col(a h (i),v h (i),p h (i)); a h (i),v h (i),p h (i) and a l (i),v l (i),pl (i) The acceleration, velocity, and position of the human-driven vehicle and the vehicle in front of it, respectively. If there is no vehicle in front of the human-driven vehicle, a virtual vehicle in front is constructed to satisfy:

[0031] a l (i)=0,v l (i) = v max ,p l (i) = p h (i) +500,

[0032] From the current moment, we trace back the historical state data of the human-driven vehicle and the vehicle in front of it for a length of T+N. This data is used to obtain the system input information and actual system output in the estimated time domain N, as well as the system input and output in the historical time domain T before the estimated time domain.

[0033] In the historical time domain, the system input is r ini =col(r(1),r(2),…,r(N)), the system output is s ini =col(s(1),s(2),…,s(N)),

[0034] Where r(j)=col(v l (j),p l (j)-p h (1)), s(j) = col(a h (j),v h (j),p h (j));

[0035] In the estimation time domain, the system input is r = col (r (T + 1), r (T + 2), ..., r (T + N)), the actual system output is s = col (s (T + 1), s (T + 2), ..., s (T + N)), and the estimated output of the system to be solved is

[0036] A quadratic programming problem is constructed to solve and identify the human-driven vehicle's behavior of slowing down to find a lane-changing gap in the adjacent lane. The quadratic programming problem is as follows:

[0037]

[0038] Among them, I is The unit matrix of the same dimension, substitute the solved vector g into Calculated

[0039] Compare p(T+N) and The third component If the size ( is a set threshold greater than 0), it is considered that the human-driven vehicle is slowing down to find a lane-changing gap in the adjacent lane.

[0040] The embodiments of the present application can allocate the nearest lane-changing gap for human-driven vehicles that intend to diverge or go straight and change lanes and slow down to find a gap. In the event of a conflict, the order constraint is recorded for re-planning. By targeted processing of the lane-changing behavior of vehicles with specific lane-changing intentions, lane-changing resources are accurately allocated and the order constraints are dynamically adjusted, thereby reducing lane-changing conflicts and improving the coordination and smoothness of lane changes for vehicles at entrances and exits.

[0041] Optionally, in one embodiment of the present application, when constructing the potential game problem in step S5 to solve the control strategy of CAV_P and CAV_F, the method includes:

[0042] Step S501: Define the strategy space of the potential game problem: S = S P ×S F ×S H , where S P is the strategy space of CAV_P, S F is the strategy space of CAV_F, S H The strategy space for human-driven vehicles,

[0043] S P =A P ×W P ,

[0044] Among them, A P ={a|a min ≤a≤a max} is the acceleration action space of CAV_P, W P ={w|w min ≤w≤w max} is the lateral displacement action space of CAV_P. Similarly, there is S F =A F ×W F , S H =A H ×W H ;

[0045] Step S502: Construct the cost function of each game participant. Taking CAV_P as an example, the cost function is:

[0046]

[0047] Among them, θ 1,P ,θ 2,P ,θ 3,P is the weight factor,

[0048]

[0049] Among them, v P,d (τ) is the expected speed of CAV_P, x P,d (τ) is the expected longitudinal position of CAV_P, which is calculated based on the collaborative decision-making and planning under the assumption of fully connected vehicles, v P (τ) and x P (τ) is the speed and position of CAV_P to be solved in the game problem, LaneWidth is the lane width, lat P,H =|w P -w H | is the lateral distance between CAV_P and the human-driven vehicle, long P,H and long P,F are the longitudinal distances between CAV_P and human-driven vehicles, and between CAV_P and CAV_F, respectively. safe dis is the minimum safe following distance, T g It is the control time domain of the game problem;

[0050] The cost functions for CAV_F and human-driven vehicles can be defined separately and The goal of each game participant is to make his or her own cost function meet the preset conditions;

[0051] Step S503: Setting the constraints of the game problem, including safety constraints and efficiency constraints. The safety constraints are:

[0052] long L,H ≥safe dis,

[0053] Among them, long L,H is the longitudinal distance between the human-driven vehicle and the vehicle ahead of it in the same lane, and the efficiency constraint is:

[0054]

[0055] Where Δx P ,Δx F ,Δx H are the solved CAV_P, CAV_F, and the change in the final longitudinal position of the human-driven vehicle compared to the expected final longitudinal position, v max,P ,v max,F ,v max,H is the maximum speed limit of CAV_P, CAV_F and human-driven vehicles in their respective lanes; cost_threshold is the maximum travel time delay allowed during the game, which is an estimate of the travel time delay caused by human-driven vehicles cutting in;

[0056] Step S504: Solve The smallest w P ,w F ,w H ,a P ,a F ,a H , and calculate the corresponding longitudinal velocity sequence v P ,v F ,v H and the longitudinal position sequence x P ,x F ,x H , the result is the Nash equilibrium solution of the potential game problem.

[0057] The embodiment of the present application can calculate the recommended lane change gaps for other human-driven vehicles and build a game group. For vehicles without clear lane change intentions, the gap that minimizes the delay in the target lane is calculated. For vehicles with the intention to change lanes straight ahead and without slowing down, the gaps are extracted from the collaborative decision-making, thereby adapting to different lane change requirements, reducing interference with the target lane, and coordinating multiple parties through game to improve overall traffic efficiency.

[0058] The embodiment of the present application can realize digital mapping of vehicles in the interchange area through the edge cloud to determine lane information, identify the lane change intention of human-driven vehicles based on the lane information, and then treat the human-driven vehicles as intelligent connected vehicles following planning suggestions on the cloud side, and make multi-vehicle collaborative order optimization decisions and trajectory planning under the assumption of full connectivity. By accurately grasping the lane status through digital mapping, combining lane change intention identification with collaborative decision-making under the assumption of full connectivity, it can not only introduce a more efficient lane change coordination mechanism between intelligent connected vehicles, but also dynamically adjust the traffic strategy of intelligent vehicles in the presence of human-driven vehicles, thereby improving the operating efficiency and traffic safety of the entrances and exits of high-speed interchanges. Therefore, it solves the problem that the related technology fails to use the coordinated lane change of straight-moving vehicles to release the main road's traffic capacity and lacks the ability to model and respond to the uncertainty of human driving behavior, resulting in low traffic efficiency of interchange sections in complex traffic environments.

[0059] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0061] Figure 1 This is a flow chart of a method for collaborative optimization of highway interchange entrances and exits under mixed traffic conditions provided in accordance with an embodiment of the present application;

[0062] Figure 2A schematic diagram of vehicle-road-cloud collaboration for a highway interchange provided according to one embodiment of the present application;

[0063] Figure 3 A schematic diagram of a highway interchange scenario provided according to one embodiment of the present application;

[0064] Figure 4 A logical block diagram of decision-making planning for lane-changing coordination of straight-moving vehicles according to one embodiment of the present application;

[0065] Figure 5 A schematic diagram of a decision-making process for the passage order of each vehicle provided according to one embodiment of the present application;

[0066] Figure 6 A longitudinal position-time diagram of all vehicles provided according to one embodiment of the present application;

[0067] Figure 7 A schematic diagram of the driving trajectory of a vehicle that needs to change lanes according to one embodiment of the present application. DETAILED DESCRIPTION

[0068] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0069] The following describes a method for collaborative optimization of highway interchange entrances and exits under mixed traffic conditions with reference to the accompanying drawings. In response to the problems mentioned in the above background technology that the related technologies fail to utilize the coordinated lane change of straight-moving vehicles to release the main road's traffic capacity and lack the ability to model and respond to the uncertainty of human driving behavior, resulting in low traffic efficiency of interchange sections in complex traffic environments, the present application provides a method for collaborative optimization of highway interchange entrances and exits under mixed traffic conditions. In this method, digital mapping of vehicles in the interchange area can be achieved through the edge cloud to determine lane information. Based on the lane information, the lane change intention of human-driven vehicles is identified. The cloud then treats the human-driven vehicles as intelligent connected vehicles following planning recommendations, and performs multi-vehicle collaborative sequence optimization decision-making and trajectory planning under the assumption of full connectivity. By accurately grasping the lane status through digital mapping, combined with lane change intention identification and collaborative decision-making under the assumption of full connectivity, a more efficient lane change coordination mechanism can be introduced between intelligent connected vehicles. In addition, the traffic strategy of intelligent vehicles can be dynamically adjusted in the presence of human-driven vehicles, thereby improving the operating efficiency and traffic safety of highway interchange entrances and exits. This solves the problem that related technologies fail to utilize the coordinated lane changing of straight-moving vehicles to release the main road's traffic capacity, and lack the ability to model and respond to the uncertainty of human driving behavior, resulting in low traffic efficiency of interchange sections in complex traffic environments.

[0070] Specifically, Figure 1 A flow chart of a method for collaborative optimization of highway interchange entrances and exits under mixed traffic provided in an embodiment of the present application.

[0071] like Figure 1 As shown, the collaborative optimization method for high-speed interchange entrances and exits under mixed traffic includes the following steps:

[0072] In step S1, digital mapping of vehicles in the interchange area is implemented on the edge cloud. The two main roads in the ramp divergence area are recorded as main road M2 and main road M1 from the inside to the outside, respectively. The two main roads in the ramp merging area are recorded as main road M4 and main road M3 from the inside to the outside, respectively. The ramp and the auxiliary lanes connecting the main roads M1 and M3 at both ends are combined and recorded as lane R.

[0073] It can be understood that in the embodiment of the present application, digital mapping can be used to convert physical elements such as physical vehicles, lane lines, traffic signs, etc. in the interchange area into digital models through edge cloud technology; lane information can include the number of lanes, lane types (such as through lanes, diverging lanes), lane boundaries and real-time traffic status, which is the basis for subsequent vehicle behavior analysis.

[0074] During actual implementation, the embodiment of the present application can realize fusion perception and real-time digital mapping of vehicles in the interchange area in the edge cloud to determine lane information.

[0075] Specifically,Figure 2 The schematic diagram of the highway interchange vehicle-road-cloud collaboration provided in the embodiment of this application is as follows Figure 2 As shown, roadside equipment collects vehicle information from the interchange area. Connected vehicles in the interchange area then upload their own information. The edge cloud implements integrated perception and real-time digital mapping of the vehicles, summarizing their location, speed, heading angle, and navigation path information. The edge cloud makes decisions and plans for each vehicle's passage order and driving trajectory. The cloud also issues longitudinal speed adjustment strategies and reference driving trajectories to guide efficient passage.

[0076] Furthermore, Figure 3 The schematic diagram of a highway interchange scene including a ramp diverging area and a ramp merging area provided in the embodiment of the present application. Figure 3 As shown in the figure, the two main lanes in the ramp divergence area are marked as main lane M2 and main lane M1 from the inside to the outside, and the two main lanes in the ramp merging area are marked as main lane M4 and main lane M3 from the inside to the outside. The ramp and the auxiliary lanes connecting the main lanes M1 and M3 at both ends are combined and marked as R. In particular, the speed limit of lane R is set to v R = 20 m / s, and the lane speed limits for vehicles in lanes M1 and M2 are set to v1 = 25 m / s and v2 = 30 m / s. Specifically, vehicles in each lane can be named using the "lane name_vehicle ID" formula, such as "M1_1" for vehicle ID 1 in main lane M1. Specifically, within a platoon, vehicles are named using the "lane name_leader vehicle ID_platoon internal ID" formula, such as "M2_3_2" for the second following vehicle in the platoon with ID 3 in main lane M2.

[0077] Further, in Figure 3 Set up a planning area and filter the vehicles to be planned. Select a certain length of road upstream of the ramp divergence area and a certain length of road upstream of the ramp confluence area as the overall planning area. When a new vehicle enters the planning area, all vehicles in the planning area are replanned.

[0078] The embodiment of the present application can realize digital modeling of the physical traffic environment through the edge cloud, accurately extract lane information, and provide real-time and accurate environmental data support for subsequent lane change intention recognition and collaborative decision-making, thereby improving the adaptability of the entire optimization method to actual traffic conditions.

[0079] In step S2, the lane-changing intention of the human-driven vehicle is classified according to the turn signal information of the human-driven vehicle, and the human-driven vehicles on the main road M3 and the main road M4 with the turn signal on are classified as having straight lane-changing intention, the human-driven vehicle on the main road M2 with the right turn signal on is classified as having split lane-changing intention, and the navigation path thereof is assumed to be a split vehicle that will enter the lane R, and the human-driven vehicle on the main road M1 with the left turn signal on is classified as having straight lane-changing intention, and the navigation path thereof is assumed to be a straight vehicle that travels along the main road M1.

[0080] It can be understood that the lane-changing intention category in the embodiments of the present application can be split lane-changing (driving off the current road), straight lane-changing (changing to an adjacent straight lane), and no explicit lane-changing intention.

[0081] For example, the embodiments of the present application can classify the lane-changing intention of the human-driven vehicle according to the turn signal information of the human-driven vehicle for vehicles located in the planning area. The human-driven vehicles on the main road M3 and the main road M4 with the turn signal on are classified as having straight lane-changing intention. The human-driven vehicle on the main road M2 with the right turn signal on is classified as having split lane-changing intention, and the navigation path thereof is assumed to be a split vehicle that will enter the lane R. The human-driven vehicle on the main road M1 with the left turn signal on is classified as having straight lane-changing intention, and the navigation path thereof is assumed to be a straight vehicle that travels along the main road M1.

[0082] The embodiments of the present application can accurately identify the lane-changing intention category of the human-driven vehicle through the combination of the turn signal information and the lane environment, and provide a basis for subsequent differentiated collaborative strategies, thereby reducing traffic conflicts caused by misjudgment of intention.

[0083] In step S3, the human-driven vehicle is regarded as an intelligent connected vehicle that follows the cloud planning suggestion for driving by using the vehicle information collected by the cloud, and multi-vehicle collaborative sequence optimization decision and trajectory planning under the assumption of full connected vehicles are performed, wherein in the collaborative decision and planning under the assumption of full connected vehicles, the navigation path information of the intelligent connected vehicle is known information, and the navigation path information of the human-driven vehicle is unknown information, the human-driven vehicle on the main road M2 without split lane-changing intention is regarded as a straight vehicle that travels along the main road M2, and the human-driven vehicle on the main road M1 without straight lane-changing intention is regarded as a split vehicle that will enter the lane R.

[0084] It can be understood that in the embodiments of the present application, the assumption of full connected vehicles can be that the cloud virtually regards the human-driven vehicle as an intelligent connected vehicle that follows the cloud planning suggestion; the multi-vehicle collaborative sequence optimization decision can be to solve the global vehicle priority based on mixed integer programming, and the goal is to minimize the total travel time; and the trajectory planning can refer to generating a conflict-free driving path for each vehicle.

[0085] During actual implementation, embodiments of the present application can utilize vehicle information aggregated from the cloud, treating human-driven vehicles as intelligent connected vehicles following cloud-based planning recommendations, and perform multi-vehicle collaborative order optimization decision-making and trajectory planning under the assumption of fully connected vehicles. During collaborative decision-making and planning under the assumption of fully connected vehicles, the navigation path information of the intelligent connected vehicles is known, while the navigation path information of the human-driven vehicles is unknown. Human-driven vehicles on main road M2 with no intention of diverging or changing lanes are treated as vehicles traveling straight along main road M2, and human-driven vehicles on main road M1 with no intention of changing lanes straight are treated as vehicles diverting into R.

[0086] Specifically, when making multi-vehicle collaborative order optimization decisions and trajectory planning under the assumption of fully connected vehicles, Figure 3 The planning area in is the overall planning area, and the rolling traversal planning considering the lane change coordination of straight vehicles is carried out using triples. The overall idea is as follows Figure 4 shown.

[0087] Optionally, in one embodiment of the present application, step S3 performs multi-vehicle collaborative sequence optimization decision-making and trajectory planning under the assumption of fully connected vehicles, including:

[0088] A certain length of main road M1, M2, M3, and M4 upstream of the ramp merge and ramp divergence areas is taken as the overall planning area, and triples are used to perform rolling traversal planning considering the lane-changing coordination of straight-moving vehicles.

[0089] The vehicles on the four lanes of main road M1, main road M2, main road M3 and main road M4 are recorded in order as {M1_1, M1_2,…, M1_p}, {M2_1, M2_2,…, M2_q}, {M3_1, M3_2,…, M3_n}, {M4_1, M4_2,…, M4_m}, {M1_1, M2_1, M3_1} is regarded as triplet number one, and {M1_1, M2_1, M3_2} is regarded as triplet number two.

[0090] For the triple formed by M1_1, M2_1, and M3_1, calculate the vehicle that has priority when M3_1 does not change lanes to M4. This vehicle is denoted as G1, and the resulting travel time delay cost3_1 for subsequent vehicles on the main roads M1, M2, and M3 is given.

[0091] For the triplet formed by M1_1, M2_1, M3_2, the vehicle that has priority to pass the merging area when M3_1 changes lane to main lane M4 is calculated, and the travel time delay cost3_2 of the subsequent vehicles of main lane M1, main lane M2 and main lane M3 caused at this time is given, at the same time, the feasible changing lane gap when the travel time delay of the vehicle of main lane M4 caused by M3_1 is the minimum is calculated, which is set as the changing lane gap before M4_i, and the travel time delay of the subsequent vehicle of main lane M4 caused by M3_1 is cost4_i.

[0092] The size of cost3_1 and cost3_2+cost4_i is compared, when cost3_1>cost3_2+cost4_i, M3_1 changes lane to the changing lane gap before M4_i, the future changing lane trajectory and speed planning of M3_1 is calculated, M3_2 and M3_3 are respectively supplemented to form a new first triplet and second triplet.

[0093] When cost3_1<=cost3_2+cost4_i, M3_1 does not change lane, if the navigation path information of M1_1 is to enter main lane M3 through R, G1 is allowed to have priority to pass, and the future speed planning and changing lane trajectory of G1 is calculated, the vehicles behind G1 are supplemented to form a new first triplet and second triplet, if the navigation path information of M1_1 is to straight along main lane M1, assuming that M1_1 changes lane to before M2_1, M1_2, M1_1, M3_1 are taken as a third triplet, the vehicle that should have priority to pass among M1_1 and M3_1 is calculated, and the travel time delay cost1 of the subsequent vehicles of main lane M1 and main lane M3 caused at this time is given, at the same time, the travel time delay cost2 of the vehicle of main lane M2 caused by M1_1 changing lane to before M2_1 is calculated.

[0094] The size of cost3_1 and cost1+cost2 is compared, when cost3_1<=cost1+cost2, M1_1 does not change lane, G1 is allowed to have priority to pass, and the future speed planning and changing lane trajectory of G1 is calculated, the vehicles behind G1 are supplemented to form a new first triplet and second triplet, when cost3_1>cost1+cost2, M1_1 changes lane to before M2_1, the future changing lane trajectory and speed planning of M1_1 is calculated, M1_2 is supplemented to form a new first triplet and second triplet.

[0095] The above steps are repeated until the changing lane trajectory and speed planning of all vehicles in the first planning area are calculated.

[0096] Specifically, each triplet sequential programming problem is formulated as a mixed integer linear programming problem, using the same structure as in Patent 202410871180.4. The optimization objective is the weighted sum of travel time delays for each lane and road section. For a triplet, the overall optimization objective is as follows:

[0097] J=α M11 C M11 +β M12 C M12 +α M21 C M21 +β M22 C M22 +β M3 C M3 ,

[0098] Among them, C M11 and C M21 are the travel time delays of vehicles on M1 and M2 from their current locations to the ramp diversion area, C M12 and C M22 The time delay for vehicles on M1 and M2 that need to go through ramp R to the main road M3 from the start of the ramp to the ramp merging area is C M3 The time delay for vehicles on the main road M3 from their current location to the time they pass through the ramp merging area. M11 ,α M21 ,β M12 ,β M22 ,β M3 is the weight corresponding to each road segment. The higher the vehicle speed, the greater the weight. When there is no vehicle on a road segment, the corresponding weight is set to 0.

[0099] Optionally, the travel time delay of each lane is calculated based on the impact of the vehicle on the following vehicles in the current lane. The latest lane change point p from the main road M2 to the main road M1 is set before the ramp divergence area. lc,d , set the latest lane change point p from ramp R to main road M3 in the ramp merging area lc,m Assume that the time it takes for the first vehicle in a lane to reach the latest lane-changing point at its current speed is t. When it needs to slow down to make room for vehicles in other lanes to change lanes, the vehicle can only reach a position Δs away from the latest lane-changing point after time t. Δs is called the coordinated displacement. The impact of Δs on all vehicles in the planned area of ​​the current lane is taken as the travel time delay C of this lane and road section.

[0100] Use d i is the distance between vehicle i and vehicle i+1 from the front and back, d v,i is the minimum following distance that vehicle i needs to follow the vehicle in front at the current speed.

[0101] The first car produces a coordinated displacement Δs, and the effect on itself is Then consider the impact C2 caused by Δs on the second car. v,2 When, record thre,2 =d1-d v,2 ,but

[0102]

[0103] Among them, v max is the maximum speed limit of the current lane. v,2 hour:

[0104] Similarly, consider the impact C caused by Δs on the i-th (i>2) vehicle i When d i >d v,i+1 hour:

[0105] Notes thre,i =(d1-d v,2 )ζ1+…+(d j -d v,j+1 )ζ j +…+d i -d v,i+1 ,but

[0106]

[0107] in, When d i ≤d v,i+1 hour:

[0108] Assume that there are N vehicles in the lane in the planning area, then C=C1+C2+…+C N .

[0109] Specifically, the constraints for each triplet sequential planning problem must ensure that each vehicle group does not collide at the two latest lane change points. The construction of safety constraints is based on patent 202410871180.4. For example, the safety constraints for vehicles on M1 and M2 in the ramp divergence zone are constructed as follows:

[0110] (s 1,1 -Δs 1,1 )-L1-(s 2,1 -Δs 2,1 )+(1-a)M≥d safe ,

[0111] (s 2,1 -Δs 2,1 )-(s 1,1 -Δs1,1 )-L2+aM≥d safe ,

[0112] Among them, s 1,1 and s 2,1 The latest lane-changing point p for vehicles on M1 and M2 to arrive at the ramp diversion area lc,d Position at time Δs 1,1 and Δs 2,1 is the coordinated displacement of the two vehicles, L1 and L2 are the lengths of the vehicles or queues, a is a 0-1 decision variable indicating the order of M1 and M2, M is a sufficiently large number, and d safe It is the minimum distance between the vehicle and the vehicle in the target lane when changing lanes.

[0113] Optionally, when planning the longitudinal velocity of the vehicle, the FVDM following model shown below is used:

[0114]

[0115] Among them, v max The maximum speed limit for the current lane.

[0116] Specifically, when a vehicle needs to change lanes, it changes lanes when the expected lane-changing gap meets the safe lane-changing conditions, and the safe lane-changing conditions use a fixed time interval.

[0117] Optionally, when calculating the lateral lane-changing trajectory of the vehicle, a quintic polynomial model as shown below is used:

[0118]

[0119] Among them, λ k is a parameter related to the road boundary, and y(0) is the initial lateral position at the start of the lane change. Vehicles that need to change lanes should complete the lane change as early as possible before the latest lane change point.

[0120] Optionally, when there is a vehicle queue on the road, the queue is considered as a whole to participate in the traffic order decision, and the PLF communication topology is adopted within the queue.

[0121] The embodiments of the present application can unify collaborative decision-making standards through the full network hypothesis, and combine cloud-based global optimization to achieve collaborative planning of multi-vehicle sequences and trajectories, thereby breaking through the control barriers between human-driven vehicles and intelligent connected vehicles, and improving the overall traffic efficiency of highway interchange entrances and exits.

[0122] In step S4, for human-driven vehicles with diverging lane-changing intentions and straight-ahead lane-changing intentions, the system analyzes whether they decelerate to search for a lane-changing gap in the adjacent lane. If so, the nearest lane-changing gap behind the adjacent lane is assigned to the human-driven vehicle. The suggested lane-changing gap is recorded as (GAP_P, GAP_F), where GAP_P represents the vehicle in front of the lane-changing gap and GAP_F represents the vehicle behind the lane-changing gap. If the result is inconsistent with the order decision made in step S3, it is recorded as an order constraint for replanning the passage order of each vehicle.

[0123] It can be understood that in the embodiment of the present application, the diverging lane-changing intention can be the vehicle's plan to leave the current main road and enter the diverging lane, and the straight lane-changing intention can be the vehicle's plan to change to the adjacent straight lane; the lane-changing gap can be the spatial interval between two vehicles in adjacent lanes that can be used for safe lane changing; the order constraint can be the priority passage relationship between vehicles recorded when the local lane-changing demand conflicts with the global coordination order.

[0124] For example, the embodiment of the present application can analyze whether there is any behavior of slowing down to find a lane-changing gap in the adjacent lane for human-driven vehicles with the intention of diverging lane changes and the intention of going straight. If there is a behavior of slowing down to find a lane-changing gap in the adjacent lane, the nearest lane-changing gap behind the adjacent lane is allocated to the human-driven vehicle, and the recommended lane-changing gap is recorded as (GAP_P, GAP_F). Among them, GAP_P represents the vehicle in front of the lane-changing gap, and GAP_F represents the vehicle behind the lane-changing gap. If the result is inconsistent with the order decision of step S3, it is recorded as an order constraint for re-planning the passage order of each vehicle. By allocating lane-changing gaps in this way, the conservative driving behavior of human-driven vehicles to slow down and change lanes can be avoided, which reduces the speed of the current lane and affects the traffic efficiency of the vehicles behind the current lane.

[0125] Optionally, in one embodiment of the present application, when analyzing in step S4 whether the human driver has slowed down to find a lane-changing gap in an adjacent lane, the following steps are included:

[0126] The longitudinal driving behavior of the human-driven vehicle is regarded as a control system. The actual output of the control system is compared with the estimated output to determine the human-driven vehicle's deceleration behavior to find a gap in the adjacent lane. The input of the system is the distance between the vehicle and the initial position of the preceding vehicle, and the speed of the preceding vehicle. The output of the system is the acceleration, speed, and position of the human-driven vehicle.

[0127] The edge cloud collects the historical status data of the human-driven vehicle and the vehicle in front of it with a preset length of T in advance to form the feature matrix U p ,U f ,Y p ,Y f :

[0128]

[0129]

[0130] Where u(i)=col(v l (i),p l (i)-p h (1)), y(i) = col(a h (i),v h (i),p h (i)); a h (i),v h (i),p h (i) and a l (i),v l (i),p l (i) are the acceleration, speed, and position of the human-driven vehicle and the vehicle in front of it, respectively. It can be understood that the preset length T can be a time length set according to a specific application scenario (such as urban roads, highways), the complexity of the driving behavior to be analyzed, and the data volume required. The preset length can be set by those skilled in the art based on actual conditions and is not specifically limited here.

[0131] If there is no vehicle in front of the human-driven vehicle, a virtual vehicle is constructed to satisfy:

[0132] a l (i)=0,v l (i) = v max ,p l (i) = p h (i) +500,

[0133] From the current moment, we can trace back the historical state data of the human-driven vehicle and the vehicle in front of it for a length of T+N, and sort out the system input information and actual system output in the estimated time domain N, as well as the system input and system output in the historical time domain T before the estimated time domain. In the historical time domain, the system input is r ini =col(r(1),r(2),…,r(N)), the system output is s ini =col(s(1),s(2),…,s(N)),

[0134] Where r(j)=col(v l (j),p l (j)-p h (1)), s(j) = col(a h (j),v h (j),p h (j));

[0135] In the estimation time domain, the system input is r = col(r(T+1), r(T+2), ..., r(T+N)), the actual system output is s = col(s(T+1), s(T+2), ..., s(T+N)), and the estimated output of the system to be solved is

[0136] A quadratic programming problem is constructed to solve and identify the human driver's behavior of slowing down to find a gap in the adjacent lane to change lanes. The quadratic programming problem is as follows:

[0137]

[0138] Among them, I is The unit matrix of the same dimension, substitute the solved vector g into Calculated

[0139] Compare p(T+N) and The third component If the size ( is a set threshold greater than 0), it is considered that the human-driven vehicle is slowing down to find a lane-changing gap in the adjacent lane.

[0140] The embodiments of the present application can prioritize the passage needs of vehicles with clear lane-changing intentions and actively decelerating by allocating lane-changing gaps in a targeted manner and dynamically recording order constraints. At the same time, a conflict adjustment mechanism is used to reduce the conflict between lane changing and the global order, thereby improving the smoothness and safety of the lane-changing process.

[0141] In step S5, the recommended lane change gaps (GAP_P, GAP_F) of other human-driven vehicles are calculated and a game group is constructed. For vehicles without a clear lane change intention, the lane change gap that causes the smallest delay in the target lane is calculated as the recommended lane change gap. For vehicles with a straight lane change intention but without slowing down to find a lane change gap in the adjacent lane, the recommended lane change gap is extracted from the sequential decision in step S3.

[0142] When the front vehicle CAV_P in the adjacent lane and the rear vehicle CAV_F in the adjacent lane are both intelligent connected vehicles, the human-driven vehicle HDV and the front vehicle PV in the current lane, the front vehicle CAV_P in the adjacent lane, and the rear vehicle CAV_F in the adjacent lane are constructed into a game group, which is recorded as {PV, HDV, CAV_P, CAV_F}.

[0143] If (CAV_P, CAV_F) is different from (GAP_P, GAP_F), a potential game problem is constructed to solve the control strategy of CAV_P and CAV_F to reduce the front-to-back gap and constrain the human-driven vehicle so that it does not squeeze in between (CAV_P, CAV_F).

[0144] It is understandable that in the embodiment of the present application, the game team can include vehicles with potential lane change conflicts into the same coordination unit, and balance the interests of all parties through multi-objective optimization; the recommended lane change gap can be the optimal lane change timing and spatial interval recommended by the cloud for the vehicle.

[0145] Optionally, in one embodiment of the present application, when constructing the potential game problem in step S5 to solve the control strategy of CAV_P and CAV_F, the following steps are included:

[0146] Step S501: Define the strategy space of the potential game problem: S = S P ×S F ×S H , where S P is the strategy space of CAV_P, S F is the strategy space of CAV_F, S H The strategy space for human-driven vehicles,

[0147] S P =A P ×W P ,

[0148] Among them, A P ={a|a min ≤a≤a max} is the acceleration action space of CAV_P, W P ={w|w min ≤w≤w max} is the lateral displacement action space of CAV_P. Similarly, there is S F =A F ×W F , S H =A H ×W H .

[0149] Step S502: Construct the cost function of each game participant. Taking CAV_P as an example, its cost function is:

[0150]

[0151] Among them, θ 1,P ,θ 2,P ,θ 3,P is the weight factor,

[0152]

[0153]

[0154] Among them, v P,d (τ) is the expected speed of CAV_P, x P,d (τ) is the expected longitudinal position of CAV_P, which is calculated based on the collaborative decision-making and planning under the assumption of fully connected vehicles. P (τ) and x P (τ) is the speed and position of CAV_P to be solved in the game problem, LaneWidth is the lane width, lat P,H =|w P -w H | is the lateral distance between CAV_P and the human-driven vehicle, long P,H and long P,F are the longitudinal distances between CAV_P and human-driven vehicles, and the longitudinal distances between CAV_P and CAV_F, respectively. safedis is the minimum safe following distance, T g It is the control time domain of the game problem;

[0155] The cost functions for CAV_F and human-driven vehicles can be defined separately and The goal of each game participant is to make his or her own cost function meet the preset conditions.

[0156] It can be understood that the preset conditions can be the target constraints for each participant to minimize the cost function value by optimizing its own control variables during the game process. The preset conditions can be set by those skilled in the art according to actual conditions and are not specifically limited here.

[0157] Step S503: Set the constraints of the game problem, including safety constraints and efficiency constraints. The safety constraints are:

[0158] long L,H ≥safe dis,

[0159] Among them, long L,H is the longitudinal distance between the human-driven vehicle and the vehicle ahead of it in the same lane, and the efficiency constraint is:

[0160]

[0161] Where Δx P ,Δx F ,Δx H are the solved CAV_P, CAV_F, and the change in the final longitudinal position of the human-driven vehicle compared to the expected final longitudinal position, v max,P ,vmax,F ,v max,H is the maximum speed limit of CAV_P, CAV_F and human-driven vehicles in their respective lanes, cost_threshold is the maximum travel time delay allowed during the game, and is the estimated value of the travel time delay caused by human-driven vehicles cutting in.

[0162] Step S504: Solve The smallest w P ,w F ,w H ,a P ,a F ,a H , and calculate the corresponding longitudinal velocity sequence v P ,v F ,v H and the longitudinal position sequence x P ,x F ,x H , the result is the Nash equilibrium solution of the potential game problem.

[0163] The embodiment of the present application can calculate the lane change gap differentially and construct a game group, taking into account the rationality of lane changes of vehicles without clear intentions and the execution of global collaborative decision-making, thereby reducing the interference of lane changes on the target lane and improving the stability of the overall traffic flow.

[0164] In step S6, when the following conditions are met: (1) step S4 allocates a lane change gap to the human-driven vehicle; (2) step S5 constructs a game problem and has a solution; (3) the actual driving trajectory of the human-driven vehicle deviates from the planning result of step S2 by more than the preset condition, the collaborative decision-making and planning under the assumption of a fully connected vehicle in step S3 are re-performed; wherein, the sequential decision result complies with the sequential constraint of step S4, and the long-term speed planning result of the intelligent connected vehicle can simultaneously realize the control strategy calculated in the game problem in step S5.

[0165] It can be understood that the preset condition for the deviation in the embodiment of the present application can be that the position deviation between the actual driving trajectory of the vehicle and the planned trajectory exceeds 1 meter. The preset condition can be set by technical personnel in this field according to actual conditions and is not specifically limited here.

[0166] During actual implementation, the embodiment of the present application can re-perform collaborative decision-making and planning under the assumption of fully connected vehicles when the trajectory deviation reaches a preset condition. At this time, the order decision result complies with the order constraint, and the long-term speed planning result of the intelligent connected vehicle can simultaneously implement the control strategy for the game problem calculation. The purpose of re-planning is to combine the global optimization results of the full network assumption with the strategy adjustment made according to the driving behavior of human-driven vehicles to obtain the optimized control strategy for each intelligent connected vehicle under mixed traffic.

[0167] Specifically, at the time of re-planning, the planning results for the first 5s of the calculated control strategy, on the basis of which the subsequent long-term planning results are calculated.

[0168] Specifically, Figure 5 The vehicle passing order decision is provided according to the embodiments of the present application. As shown in the figure, M2_3 and M3_2 are human-driven vehicles, and the rest are intelligent connected vehicle queues. In the queue, the solid circle represents the head vehicle of the queue, and the hollow circle represents the following vehicle of the queue. Figure 5 The arrow points to the vehicle with a higher passing order, and the dashed arrow indicates that the vehicle at the end of the arrow will leave the current lane. Figure 5

[0169] The embodiments of the present application can dynamically respond to trajectory deviation and construct safety constraints combined with short-time speed prediction, so that the control instructions of intelligent connected vehicles are more in line with the actual behavior of human-driven vehicles, thereby improving the flexibility and safety of trajectory tracking and reducing sudden conflicts.

[0170] Step S7, in each preset control period, the historical speed curve of the human-driven vehicle is used to predict the speed curve of the human-driven vehicle in the future control period, the long-term planning result calculated in step S2 or step S6 is taken as the reference target, the safety constraint is formed according to the prediction result of the human-driven vehicle, and the short-term control strategy of the intelligent connected vehicle is calculated.

[0171] It can be understood that the preset control period in the embodiments of the present application can be the time interval (such as 5 seconds) for updating the control strategy in the cloud, and the preset control period can be set by those skilled in the art according to the actual situation, which is not specifically limited here.

[0172] For example, the embodiments of the present application can use the historical speed curve of the human-driven vehicle to predict the speed curve of the human-driven vehicle in the future control period in each control period T, take the calculated long-term planning result as the reference target, form the safety constraint according to the prediction result of the human-driven vehicle, and calculate the short-term control strategy of the intelligent connected vehicle.

[0173] Specifically, the short-term control strategy can construct a planning problem solution, the optimization target is that the speed deviation relative to the long-term planning result is as small as possible, the safety constraint is that the vehicles cannot collide with each other, and the decision variable is the short-term speed sequence of the vehicle, and the control strategy for the future 3s is calculated.

[0174] Specifically, Figure 6 The longitudinal position-time diagram of all vehicles is provided according to the embodiments of the present application, Figure 7 The driving trajectory of the vehicle that needs to change lanes is provided according to the embodiments of the present application. For intelligent connected vehicles,​Figure 6 and Figure 7 The data in is the planning result of this embodiment, and it is considered that the planning result is the same as the actual execution result. Figure 6 and Figure 7 The data is the actual driving data of the vehicle driven by humans in this embodiment.

[0175] Figure 6 In the figure, the horizontal axis is time t, and the vertical axis is the distance s that the vehicle travels along the road direction from the initial moment. Figure 6 The gray areas in sub-figures (a) and (b) represent ramp divergence areas, and the gray areas in sub-figures (c) and (d) represent ramp merging areas. The color of the ST curves in the figures changes with vehicle speed, with red indicating slow speeds and green indicating fast speeds. Each ST curve represents a vehicle, and the red dots on the curves are lane change markers. Some curves are disconnected at the red dots, indicating that the corresponding vehicle changed lanes to the adjacent lane at the red dots; some curves are continued at the red dots, indicating that the corresponding vehicle changed lanes to the current lane at the red dots. Figure 6 The curve with red dots in (a) and Figure 6 The corresponding curve with red dots in (b) represents vehicles changing lanes from the main road M2 to the main road M1. Figure 6 The curve in (c) is disconnected at the red point and Figure 6 The corresponding curve with red dots in (d) represents vehicles changing lanes from the main road M3 to the main road M4. Figure 6 The curve with red dots in (c) that remains continuous is for vehicles that start from the main road M1 and the main road M2 and enter the main road M3 via R.

[0176] Figure 7 In the middle, the horizontal axis is Figure 3 The horizontal axis X coordinate in the coordinate system of the scene diagram is Figure 3 The Y coordinate of the vertical axis in the coordinate system of the scene diagram. Figure 7 In the figure, each subgraph is the driving trajectory of each vehicle that needs to change lanes. For the vehicle queue, only the driving trajectory of the leading vehicle is drawn.

[0177] This embodiment of the application periodically utilizes historical speed curves of human-driven vehicles to predict their future speeds, establishes safety constraints based on long-term planning, and calculates short-term control strategies for intelligent connected vehicles. The technical effect: Adapting to human driving characteristics while taking into account global collaborative goals, enhancing the adaptability of control strategies, and ensuring the stability and safety of multi-vehicle collaboration.

[0178] The embodiment of the present application can realize digital mapping of vehicles in the interchange area through the edge cloud to determine lane information, identify the lane change intention of human-driven vehicles based on the lane information, and then treat the human-driven vehicles as intelligent connected vehicles following planning suggestions on the cloud side, and make multi-vehicle collaborative order optimization decisions and trajectory planning under the assumption of full connectivity. By accurately grasping the lane status through digital mapping, combining lane change intention identification with collaborative decision-making under the assumption of full connectivity, it can not only introduce a more efficient lane change coordination mechanism between intelligent connected vehicles, but also dynamically adjust the traffic strategy of intelligent vehicles in the presence of human-driven vehicles, thereby improving the operating efficiency and traffic safety of the entrances and exits of high-speed interchanges. Therefore, it solves the problem that the related technology fails to use the coordinated lane change of straight-moving vehicles to release the main road's traffic capacity and lacks the ability to model and respond to the uncertainty of human driving behavior, resulting in low traffic efficiency of interchange sections in complex traffic environments.

[0179] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0180] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of this application, "N" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0181] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing a custom logical function or process step, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed in a different order than shown or discussed, including performing functions in a substantially simultaneous manner or in a reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application pertain.

Claims

1. A collaborative optimization method for high-speed interchange entrances and exits under mixed traffic conditions, characterized in that: The following steps are involved: Step S1: Implement digital mapping of vehicles in the interchange area on the edge cloud. The two main roads in the ramp divergence area are marked as main roads M2 and M1 from the inside out. The two main roads in the ramp merging area are marked as main roads M4 and M3 from the inside out. The ramp and the auxiliary lanes connecting the main roads M1 and M3 at both ends are combined and marked as lane R. Step S2: Classifying the lane change intention of the human-driven vehicle based on the turn signal information of the human-driven vehicle. Human-driven vehicles with their turn signals on on the main roads M3 and M4 are both classified as having a straight lane change intention. Human-driven vehicles with their right turn signals on on the main road M2 are classified as having a diverging lane change intention, and their navigation path is assumed to be that of a diverging vehicle about to enter lane R. Furthermore, human-driven vehicles with their left turn signals on on the main road M1 are classified as having a straight lane change intention, and their navigation path is assumed to be that of a straight-moving vehicle traveling along the main road M1. Step S3: Using the vehicle information aggregated from the cloud, the human-driven vehicle is treated as an intelligent connected vehicle that follows cloud-based planning recommendations, and multi-vehicle collaborative order optimization decision-making and trajectory planning are performed under the assumption of fully connected vehicles. During the collaborative decision-making and planning under the assumption of fully connected vehicles, the navigation path information of the intelligent connected vehicle is known, while the navigation path information of the human-driven vehicle is unknown. Human-driven vehicles on the main road M2 with no intention of diverging or changing lanes are treated as straight-moving vehicles traveling along the main road M2, and human-driven vehicles on the main road M1 with no intention of going straight or changing lanes are treated as diverging vehicles that will enter lane R. Step S4: For the human-driven vehicle with the diverging lane-changing intention and the straight lane-changing intention, analyze whether it decelerates to search for a lane-changing gap in the adjacent lane. If it decelerates to search for a lane-changing gap in the adjacent lane, assign the nearest lane-changing gap behind the adjacent lane to the human-driven vehicle. The suggested lane-changing gap is recorded as (GAP_P, GAP_F), where GAP_P represents the vehicle in front of the lane-changing gap and GAP_F represents the vehicle behind the lane-changing gap. If the result is inconsistent with the order decision of step S3, it is recorded as an order constraint for replanning the passage order of each vehicle. Step S5: Calculate the suggested lane change gaps (GAP_P, GAP_F) for other human-driven vehicles and construct a game group. For vehicles without a clear lane change intention, the lane change gap that causes the least delay in the target lane is calculated as the suggested lane change gap. For vehicles with a straight lane change intention but not slowing down to find a lane change gap in the adjacent lane, the suggested lane change gap is extracted from the sequential decision in step S3. When the preceding vehicle CAV_P in the adjacent lane and the following vehicle CAV_F in the adjacent lane are both intelligent connected vehicles, the human-driven vehicle HDV and the preceding vehicle PV in the current lane, the preceding vehicle CAV_P in the adjacent lane, and the following vehicle CAV_F in the adjacent lane are constructed into a game group, which is recorded as {PV, HDV, CAV_P, CAV_F}; If (CAV_P, CAV_F) is different from (GAP_P, GAP_F), a potential game problem is constructed to solve the control strategy of CAV_P and CAV_F to reduce the front-to-back gap and constrain the human driver to not cut in between (CAV_P, CAV_F); Step S6: When the following conditions are met: (1) the lane change gap is allocated to the human-driven vehicle in step S4; (2) the game problem is constructed and solved in step S5; (3) the deviation between the actual driving trajectory of the human-driven vehicle and the planning result of step S2 is greater than a preset condition, the collaborative decision-making and planning under the assumption of the fully connected vehicle in step S3 are re-performed; wherein the order decision result meets the order constraint of step S4, and the long-term speed planning result of the intelligent connected vehicle can simultaneously realize the control strategy calculated in the game problem in step S5; Step S7: In each preset control cycle, the historical speed curve of the human-driven vehicle is used to predict the speed curve of the human-driven vehicle in a future control cycle, and the long-term planning result calculated in step S2 or step S6 is used as a reference target. Safety constraints are formed according to the prediction results of the human-driven vehicle, and the short-term control strategy of the intelligent connected vehicle is calculated.

2. The method according to claim 1, characterized in that Step S3 performs multi-vehicle collaborative sequence optimization decision-making and trajectory planning under the assumption of fully connected vehicles, including: A certain length of the main road M1, the main road M2, the main road M3, and the main road M4 upstream of the ramp merging and ramp diverging areas is used as the overall planning area, and a rolling traversal planning is performed using triples to consider the lane change coordination of straight-moving vehicles; The vehicles on the four lanes of the main road M1, the main road M2, the main road M3, and the main road M4 are recorded in order as {M1_1, M1_2, …, M1_p}, {M2_1, M2_2, …, M2_q}, {M3_1, M3_2, …, M3_n}, {M4_1, M4_2, …, M4_m}, {M1_1, M2_1, M3_1} is taken as the first triplet, and {M1_1, M2_1, M3_2} is taken as the second triplet; For the triplet formed by M1_1, M2_1, and M3_1, calculate the vehicle that would have priority if M3_1 did not change lanes to the main road M4. This vehicle is denoted as G1, and the resulting travel time delay cost3_1 for subsequent vehicles on the main roads M1, M2, and M3 is given. For the triple formed by M1_1, M2_1, and M3_2, calculate the vehicles that have priority in the merging area when M3_1 changes lanes to M4, and give the travel time delay cost3_2 for subsequent vehicles on the main roads M1, M2, and M3 caused by this. At the same time, calculate the feasible lane-changing gap in which M3_1 minimizes the travel time delay for vehicles on the main road M4. Assume that M3_1 should change lanes before M4_i, and the corresponding travel time delay caused to subsequent vehicles on the main road M4 is cost4_i. Compare cost3_1 with cost3_2 + cost4_i. When cost3_1 > cost3_2 + cost4_i, M3_1 switches to the lane-changing gap before M4_i. Calculate M3_1's future lane-changing trajectory and speed plan. M3_2 and M3_3 are added to form new triplet number one and triplet number two, respectively. When cost3_1<=cost3_2+cost4_i, M3_1 does not change lanes. If the navigation path information of M1_1 is to enter the main road M3 via R, G1 is given priority, and its future speed plan and lane change trajectory are calculated. The vehicle behind G1 is added to form a new triplet number one and triplet number two. If the navigation path information of M1_1 is to go straight along the main road M1, assuming that M1_1 changes lanes before M2_1, M1_2, M1_1, and M3_1 are used as triplet number three, and the vehicle that should have priority among M1_1 and M3_1 is calculated. The travel time delay cost1 caused to the subsequent vehicles on the main roads M1 and M3 at this time is given. At the same time, the travel time delay cost2 caused to the vehicles on the main road M2 before M1_1 changes lanes to M2_1 is calculated. Compare cost3_1 and cost1+cost2. When cost3_1 <= cost1+cost2, M1_1 does not change lanes, giving G1 priority. It also calculates its future speed plan and lane-changing trajectory, and adds the vehicle behind G1 to form new triplet 1 and triplet 2. When cost3_1 > cost1+cost2, M1_1 changes lanes before M2_1. The future lane-changing trajectory and speed plan of M1_1 are calculated, and M1_2 is added to form new triplet 1 and triplet 2. Repeat the above steps until the lane change trajectory and speed planning are calculated for all vehicles in the first planning area.

3. The method according to claim 1, characterized in that The analysis in step S4 of whether the human driver has slowed down to find a lane-changing gap in the adjacent lane includes: The longitudinal driving behavior of the human-driven vehicle is considered as a control system. The deceleration behavior of the human-driven vehicle to find a lane-changing gap in the adjacent lane is determined by comparing the actual output of the control system with the estimated output. The input of the system is the distance between the ego vehicle and the initial position of the preceding vehicle, and the speed of the preceding vehicle. The output of the system is the acceleration, speed, and position of the human-driven vehicle. The edge cloud collects the historical status data of the human-driven vehicle and the vehicle in front of it with a preset length of T in advance to form the feature matrix U p ,U f ,Y p ,Y f : Where u(i)=col(v l (i),p l (i)-p h (1)), y(i) = col(a h (i),v h (i),p h (i)); a h (i),v h (i),p h (i) and a l (i),v l (i),p l (i) The acceleration, velocity, and position of the human-driven vehicle and the vehicle in front of it, respectively. If there is no vehicle in front of the human-driven vehicle, a virtual vehicle in front is constructed to satisfy: a l (i)=0,v l (i)=v max ,p l (i)=p h (i)+500, From the current moment, we trace back the historical state data of the human-driven vehicle and the vehicle in front of it for a length of T+N. This data is used to obtain the system input information and actual system output in the estimated time domain N, as well as the system input and output in the historical time domain T before the estimated time domain. In the historical time domain, the system input is r ini =col(r(1),r(2),…,r(N)), the system output is s ini =col(s(1),s(2),…,s(N)), Where r(j)=col(v l (j),p l (j)-p h (1)), s(j) = col(a h (j),v h (j),p h (j)); In the estimation time domain, the system input is r = col (r (T + 1), r (T + 2), ..., r (T + N)), the actual system output is s = col (s (T + 1), s (T + 2), ..., s (T + N)), and the estimated output of the system to be solved is A quadratic programming problem is constructed to solve and identify the human-driven vehicle's behavior of slowing down to find a lane-changing gap in the adjacent lane. The quadratic programming problem is as follows: Among them, I is The unit matrix of the same dimension, substitute the solved vector g into Calculated Compare p(T+N) and The third component If the size ( is a set threshold greater than 0), it is considered that the human-driven vehicle is slowing down to find a lane-changing gap in the adjacent lane.

4. The method according to claim 1, wherein When constructing the potential game problem in step S5 to solve the control strategy of CAV_P and CAV_F, it includes: Step S501: Define the strategy space of the potential game problem: S = S P ×S F ×S H , where S P is the strategy space of CAV_P, S F is the strategy space of CAV_F, S H The strategy space for human-driven vehicles, S P =A P ×W P , Among them, A P ={a|a min ≤a≤a max } is the acceleration action space of CAV_P, W P ={w|w 0in ≤w≤w max } is the lateral displacement action space of CAV_P. Similarly, there is S F =A F ×W F , S H =A H ×W H ; Step S502: Construct the cost function of each game participant. Taking CAV_P as an example, the cost function is: Among them, θ 1,P ,θ 2,P ,θ 3,P is the weight factor, Among them, v P,d (τ) is the expected speed of CAV_P, x P,d (τ) is the expected longitudinal position of CAV_P, which is calculated based on the collaborative decision-making and planning under the assumption of fully connected vehicles, v P (τ) and x P (τ) is the speed and position of CAV_P to be solved in the game problem, LaneWidth is the lane width, lat P,H =|w P -w H | is the lateral distance between CAV_P and the human-driven vehicle, long P,H and long P,F are the longitudinal distances between CAV_P and human-driven vehicles, and the longitudinal distances between CAV_P and CAV_F, respectively. safedis is the minimum safe following distance, T g It is the control time domain of the game problem; The cost functions for CAV_F and human-driven vehicles can be defined separately and The goal of each game participant is to make his or her own cost function meet the preset conditions; Step S503: Setting the constraints of the game problem, including safety constraints and efficiency constraints. The safety constraints are: long L,H ≥safe dis, Among them, long L,H is the longitudinal distance between the human-driven vehicle and the vehicle ahead of it in the same lane, and the efficiency constraint is: Where Δx P ,Δx F ,Δx H are the solved CAV_P, CAV_F, and the change in the final longitudinal position of the human-driven vehicle compared to the expected final longitudinal position, v max,P ,v max,F ,v max,H is the maximum speed limit of CAV_P, CAV_F and human-driven vehicles in their respective lanes; cost_threshold is the maximum travel time delay allowed during the game, which is an estimate of the travel time delay caused by human-driven vehicles cutting in; Step S504: Solve The smallest w P ,w F ,w H ,a P ,a F ,a H , and calculate the corresponding longitudinal velocity sequence v P ,v F ,v H and the longitudinal position sequence x P ,x F ,x H , the result is the Nash equilibrium solution of the potential game problem.

Images