A dynamic lane changing decision optimization method and system for fully connected autonomous driving scenarios

Abstract

This invention discloses a dynamic lane-changing decision optimization method and system for fully connected autonomous driving scenarios, relating to the fields of traffic engineering and intelligent transportation systems. The method includes: deriving the flow-density relationship in equilibrium based on a car-following model for connected autonomous vehicles, obtaining a basic triangular macroscopic diagram, and determining the maximum flow, maximum inflow rate, and congestion density of a single lane. Taking the road network as the object, the planning time period is discretized into decision intervals, with lane configuration as the decision variable and capacity parameters as constraints, to establish a dynamic lane-changing optimization model that minimizes the total system travel time. This model includes constraints on road segment transmission / reception capacity, flow conservation, total number of lanes, lane change amplitude, and lane difference between adjacent road segments. At the beginning of each decision interval, state correction, short-term demand forecasting, model solving, and scheme implementation are executed sequentially. Only the current control time-domain scheme is applied before proceeding to the next time step, and this process continues until the planning is completed.

Description

Technical Field

[0001] This invention belongs to the technical field of traffic engineering and intelligent transportation systems, and in particular relates to a dynamic lane changing decision optimization method and system for fully connected autonomous driving scenarios. Background Technology

[0002] In urban road traffic networks, roads typically consist of two parallel, opposing road segments, and the number of lanes is a crucial factor affecting the capacity of these segments. As travel demand constantly changes in time and space, especially during peak hours and under conditions of localized congestion, two-way traffic flow often exhibits significant directional differences and time-varying characteristics. Therefore, dynamically adjusting the permitted direction of lanes based on traffic demand to achieve supply-demand matching is a vital means of improving road network efficiency. However, in traditional manually driven vehicle scenarios, lane changing requires hardware such as zipper trucks and movable water-filled barriers, as well as professional personnel, which significantly limits the frequency and flexibility of lane changing management measures.

[0003] In recent years, with the rapid development of vehicle-to-everything (V2X) communication technology, vehicle-road cooperative technology, and autonomous driving technology, the large-scale application of connected and autonomous vehicles (CAVs) has provided new conditions for the realization of dynamic lane management. In the scenario of fully connected and autonomous vehicles, relying on vehicle-to-everything (V2X) technology and autonomous driving technology, connected and autonomous vehicles can receive and respond to dynamic lane driving direction management schemes in real time. This allows the number of lanes in each driving direction on the road to be dynamically adjusted according to time-varying travel demand, without relying on hardware facilities such as zipper trucks or movable water-filled barriers. This provides technical support for high-frequency and refined dynamic lane management.

[0004] Meanwhile, urban road traffic exhibits significant network connectivity. Changes in the number of lanes on a single road segment not only affect the traffic conditions of that segment but also further impact traffic flow connections between upstream and downstream segments, influencing traffic distribution across the entire road network by altering travelers' route choices. Therefore, for fully connected autonomous vehicle scenarios, it is necessary to study the dynamic lane-changing decision optimization problem at the road network level.

[0005] In existing technologies, Duell et al. used the Cell Transmission Model (CTM) as the underlying traffic flow model, taking the number of lanes in different directions of travel in each cell as the decision variable, and constructed a lane-changing model applicable only to a single road with the goal of system optimization. The aforementioned existing technologies have the following drawbacks: First, they optimize only a single road, failing to consider the traffic flow propagation relationship and capacity matching relationship between upstream and downstream adjacent road segments at the road network level, which can easily create bottlenecks again in the road network; second, they are based on static origin-destination travel demand, failing to consider the characteristics of origin-destination travel demand changing over time. When traffic demand fluctuates, lane configuration is still optimized based on previously predicted origin-destination travel demand, making it difficult to respond promptly to changes in actual traffic conditions, thus weakening the effectiveness of the optimization scheme. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a dynamic lane-changing decision optimization method for fully connected autonomous driving scenarios, comprising: The car-following behavior of connected autonomous vehicles is characterized by a car-following model. The relationship between traffic flow and traffic flow density of a road segment is derived under the equilibrium state of the car-following model, resulting in a triangular macroscopic basic diagram of traffic flow in the scenario of fully connected autonomous vehicles. Based on the triangular macroscopic basic diagram of traffic flow, the maximum flow rate, maximum inflow rate, and congestion density of each road segment are determined. Taking a given road network as the optimization object, the planning period is discretized into several decision time intervals. Multiple parameters of the given road network are used as decision variables. The maximum flow rate of a single lane, the maximum inflow rate of a single lane, and the congestion density are used as traffic capacity parameters. Under the conditions of satisfying the constraints of road segment transmission capacity and road segment receiving capacity of the transmission model, as well as the constraints of flow conservation, road segment on total lane volume, lane change amplitude of adjacent decision time intervals, and lane difference between adjacent road segments, a dynamic lane reversal optimization model is established with the goal of minimizing the total system travel time. At the beginning of each decision time interval, the system state correction, short-term origin-destination travel demand prediction, dynamic lane reversal optimization model solution and lane configuration scheme implementation are executed sequentially. Only the lane configuration scheme in the current control time domain is implemented in the road traffic network before being advanced to the next decision time. This process is repeated until the end of the planning period.

[0007] Furthermore, the decision variables include: the number of lanes allocated to road segments in the fixed road network, the cumulative inflow of road segments, the cumulative outflow of road segments, and the cumulative transfer flow between road segments.

[0008] Furthermore, the velocity update equation of the car-following model is: The following error equation of the following car model is: in, For the control system step size of connected autonomous vehicles, For connected autonomous vehicles The speed at any given moment For connected autonomous vehicles The speed at any given moment for The following error between the actual vehicle spacing and the expected vehicle spacing at any given time. for The first derivative with respect to time, for The actual headway of a connected autonomous vehicle. This refers to the body length of a connected autonomous vehicle. To ensure the minimum safe parking distance, To maintain a constant inter-vehicle distance for connected and autonomous vehicles, This is the proportional control coefficient for the car-following error. This is the differential control coefficient for the car-following error.

[0009] Furthermore, the macroscopic basic diagram of triangular traffic flow in a fully connected autonomous vehicle scenario includes: In the free-flow section, the traffic flow of the road segment Traffic flow density The relationship between them: in, For free flow velocity; During congested sections, the car-following model is brought to equilibrium, and the traffic flow density in the equilibrium state is calculated. : Calculate the relationship between traffic flow and traffic flow density in congested sections: Critical density At the intersection of the free-flow section and the congested section, through simultaneous equations... and form Calculate the critical density And calculate the maximum flow rate. ,make The value is zero, thus obtaining the blockage density. and according to Calculate the propagation speed of congested sections ; With free flow velocity and the speed of transmission in congested areas Two slopes, with maximum flow rate Using the vertex as the point, a basic macroscopic diagram of the triangular traffic flow is formed in the traffic flow-traffic flow density coordinate system; The maximum flow rate per lane for each road section is taken as follows: The maximum inflow rate for a single lane is set to [value]. .

[0010] Furthermore, discretizing the planning period into several decision time intervals also includes: discretizing the planning period into decision time intervals and basic time intervals; The constraints on the transmission capacity and reception capacity of road segments are established through a road segment transmission model, including: Section In the Segment transmission capacity within a basic time interval for: Section In the Segment reception capability within a basic time interval for: For the road section and subsequent road sections In the Within each basic time interval, the road segment Rear connecting road sections The actual traffic flow transmitted was: in, As of the date The cumulative number of entries into the road segment at the end of each basic time interval The number of vehicles, As of the date The cumulative number of entries into the road segment at the end of each basic time interval The number of vehicles, As of the date The cumulative number of road sections exited at the end of each basic time interval The number of vehicles, As of the date The cumulative number of road sections exited at the end of each basic time interval The number of vehicles, As of the date The cumulative number of road sections exited at the end of each basic time interval The number of vehicles is used as the cumulative outflow for a road segment. For road section Free-flow propagation time, For road section The transmission time during congested periods For road section The length of the road segment For road section In the The outflow capacity of road segments within a basic time interval. For road section In the Road segment inflow capacity within a basic time interval This is a set of non-terminating road segments. For road section The following road segments are collected. It is a set of indexes based on the basic time interval.

[0011] Furthermore, establishing road segment transmission capacity constraints and road segment reception capacity constraints through the road segment transmission model also includes: performing equivalent linearization on the road segment transmission capacity constraints and the road segment reception capacity constraints; and, under the conditions of satisfying the constraints after equivalent linearization, as well as the constraints of traffic conservation, road segment on total lane volume, lane change amplitude constraints of adjacent decision time intervals, and lane difference constraints of adjacent road segments, establishing a dynamic lane reversing optimization model with the objective of minimizing the total system travel time.

[0012] Furthermore, the characteristic feature is that the flow conservation constraint includes: For the initial road segment, the cumulative inflow to the segment equals the cumulative travel demand generated by the initial road segment at both the origin and destination points: For non-starting road segments, the cumulative inflow to the segment equals the inflow to all preceding road segments. The sum of the cumulative transfer traffic between road sections: For non-terminating road segments, the cumulative outflow of the segment equals the segment's total outflow. The sum of the cumulative transfer flows between road segments leading to all subsequent road segments: The non-negativity constraints and initial value constraints for the cumulative transfer flow between road segments are as follows: in, As of the date The cumulative number of entries into the road segment at the end of each basic time interval The number of vehicles, as the road segment Cumulative inflow As of the date The final segment of each basic time interval The cumulative amount of travel demand generated between origin and destination. For the initial set of road segments, This is the complete collection of road sections. For road section The preceding road segment collection, As of the date The basic time interval is not determined by the road segment. Flowing section The cumulative transfer traffic between road sections As of the date The basic time interval is not determined by the road segment. Flow to subsequent road sections The cumulative traffic flow transferred between road sections.

[0013] Furthermore, the total number of lanes on the road segment is constrained as follows: Among them, the road section From the road section and road sections Road sections that are parallel to each other in opposite directions and have the same length and free-flow velocity. constitute, For the road section The total number of lanes on the corresponding road. For road segment pairs, For the set of decision time interval indices, It is a set of positive integers. For road section In the Number of lanes allocated to road segments within each decision time interval For road section In the Number of lanes allocated to a road segment within a decision time interval; The lane change range constraint for adjacent decision time intervals is: in, For road section In the Number of lanes allocated to road segments within each decision time interval For road section The maximum number of lane changes allowed between two adjacent decision time intervals; The lane difference constraint between adjacent road segments is: in, Road segments with connecting relationships and subsequent road sections The maximum difference in the number of lane assignments allowed within the same decision time interval.

[0014] Furthermore, the dynamic lane-changing optimization model includes: in, Let be the objective function. The decision variable is the base time interval duration. .

[0015] This invention also proposes a dynamic lane-changing decision optimization system for fully connected autonomous driving scenarios, comprising: The module for generating a macroscopic basic map is used to characterize the car-following behavior of connected autonomous vehicles through the car-following model of connected autonomous vehicles. Under the equilibrium state of the car-following model, the relationship between traffic flow and traffic flow density of the road segment is derived to obtain a triangular macroscopic basic map of traffic flow in the scenario of fully connected autonomous vehicles. Based on the triangular macroscopic basic map of traffic flow, the maximum flow rate of a single lane, the maximum inflow rate of a single lane, and the congestion density of each road segment are determined. A dynamic lane-changing optimization model module is established to take a given road network as the optimization object, discretize the planning time period into several decision time intervals, use multiple parameters of the given road network as decision variables, and use the maximum flow rate of a single lane, the maximum inflow rate of a single lane, and the congestion density as capacity parameters. Under the conditions of satisfying the constraints of road segment transmission capacity and road segment reception capacity established by the road segment transmission model, as well as the constraints of flow conservation, road segment on total lane volume, lane change amplitude of adjacent decision time intervals, and lane difference between adjacent road segments, the dynamic lane-changing optimization model is established with the goal of minimizing the total travel time of the system. The decision module is used to sequentially perform system state correction, short-term origin-destination travel demand prediction, dynamic lane reversal optimization model solution, and lane configuration scheme implementation at the beginning of each decision time interval. Only the lane configuration scheme in the current control time domain is implemented in the road traffic network before proceeding to the next decision time. This process is repeated until the end of the planning period.

[0016] Compared with the prior art, the present invention has the following advantages and technical effects: (1) Using the Link Transmission Model (LTM) as a traffic flow propagation description tool can effectively characterize the traffic flow propagation relationship and capacity matching relationship between upstream and downstream adjacent road segments in the road network, avoid creating new bottlenecks in the road network due to local optimization, and achieve coordinated operation of the entire road network.

[0017] (2) The planning time period is innovatively discretized in two layers: the basic time interval is used to finely characterize the traffic flow propagation process, and the decision time interval is used to determine the update frequency of the lane configuration scheme. The two work together through a mapping relationship, so that the same decision scheme is applied to the corresponding set of basic time intervals, which controls the scale of the decision problem while ensuring the accuracy of traffic flow description.

[0018] (3) The rolling time domain optimization algorithm is adopted to decompose the long-term dynamic lane management decision problem into a series of short-term dynamic lane management sub-problems, which can speed up the solution, avoid the accumulation of long-term prediction errors, and respond to time-varying traffic conditions in a timely manner.

[0019] (4) Numerical experiments show that, compared with the fixed lane benchmark scheme (3 lanes in each direction), the average travel time of the dynamic lane changing scheme proposed in this invention is reduced by about 39.8%, and the total travel time of the system is reduced by about 40.6%, which significantly improves the traffic efficiency of the road network. Attached Figure Description

[0020] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention; Figure 2 This is a system structure diagram of Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the basic macroscopic traffic flow of a triangle in the scenario of a fully connected and autonomous vehicle according to the present invention; Figure 4 This is a schematic diagram of the rolling time-domain optimization algorithm of the present invention; Figure 5 This is a schematic diagram of the improved ND road network topology of the present invention; Figure 6 This is the decision-making moment of the invention. Schematic diagram of lane configuration results for the road network; Figure 7 This is the decision-making moment of the invention. Schematic diagram of lane configuration results for the road network (peak demand from west to east); Figure 8 This is the decision-making moment of the invention. Schematic diagram of lane configuration results for the road network (peak demand from east to west). Detailed Implementation

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0023] The main terms used in this invention are defined as follows: Connected autonomous vehicles (CAVs): Autonomous vehicles equipped with a cooperative adaptive cruise control system that can receive and execute dynamic lane management schemes in real time through vehicle-to-infrastructure communication technology; Road segment: A directed connection unit between two nodes in a road network; A road segment pair is a combination of two parallel road segments of the same length and free-flow velocity, corresponding to the same two-way road. Cumulative inflow to a road segment: The total number of vehicles that have entered the road segment up to the end of a certain basic time interval; Cumulative outflow of a road segment: The total number of vehicles that have exited the road segment up to the end of a certain basic time interval; Cumulative transfer flow between road segments: The total number of vehicles that have flowed from one road segment to the next road segment up to the end of a certain basic time interval; Lane allocation for a road segment: The number of lanes allocated to the direction of travel on a road segment within a certain decision time interval; Segment sending capacity: Within a certain basic time interval, the maximum number of vehicles that a segment can send to its downstream successor segment under the current traffic conditions; Road segment receiving capacity: Within a certain basic time interval, the maximum number of vehicles that a road segment can receive from the upstream preceding road segment under the current traffic conditions; Basic time interval: The smallest unit of time used to characterize the propagation process of traffic flow, with a duration of [duration missing]. ; Decision interval: A time unit used to determine the frequency of lane configuration plan updates, with a duration of [duration missing]. ,in The number of base time intervals included in each decision time interval; Decision time domain: The time range considered in the future during each rolling time domain optimization, with a duration of [duration missing]. ; Control Time Domain: The actual execution time range in each rolling time domain optimization, with a duration of [duration missing]. This is equal to the duration of a decision-making time interval; Total System Travel Time (TSTT): The sum of the travel times of all vehicles entering the road network and ultimately reaching their destination; Average Trip Time (ATT): Total trip time of the system divided by the total number of vehicles.

[0024] Example 1 like Figure 1 As shown in the figure, this embodiment proposes a dynamic lane changing decision optimization method for fully connected autonomous driving scenarios, which specifically includes the following steps: Step S1: Characterize the car-following behavior of connected autonomous vehicles through the car-following model, derive the relationship between traffic flow and traffic flow density of road segments under the equilibrium state of the car-following model, obtain the macroscopic basic diagram of triangular traffic flow in the scenario of fully connected autonomous vehicles, and determine the maximum flow rate, maximum inflow rate and congestion density of each road segment based on the macroscopic basic diagram of triangular traffic flow. Specifically, the velocity update equation of the car-following model is: (1) The following error equation of the following car model is: (2) in, For the control system step size of connected autonomous vehicles, For connected autonomous vehicles The speed at any given moment For connected autonomous vehicles The speed at any given moment for The following error between the actual vehicle spacing and the expected vehicle spacing at any given time. for The first derivative with respect to time, for The actual headway of a connected autonomous vehicle. This refers to the body length of a connected autonomous vehicle. To ensure the minimum safe parking distance, To maintain a constant inter-vehicle distance for connected and autonomous vehicles, This is the proportional control coefficient for the car-following error. This is the differential control coefficient for the car-following error.

[0025] Preferably, formula (1) describes the speed update process of a connected autonomous vehicle under car-following control: the vehicle updates its speed based on the current car-following error. and its rate of change Adjust the driving speed to maintain the desired distance from the vehicle in front. Formula (2) defines the following error, which is the current actual headway. Distance from the target safe front end The deviation between them, where the target safe front-end distance is determined by the vehicle length. Minimum safe parking distance and the following distance that should be maintained when driving at the current speed It consists of three parts.

[0026] Specifically, the macroscopic basic diagram of triangular traffic flow in a fully connected autonomous vehicle scenario includes: Traffic flow in free-flow sections (low-density free-flow conditions) Traffic flow density The relationship between them is ,in, For free flow velocity; In congested sections (under congestion conditions), the basic macroscopic diagram of homogeneous traffic flow (composed of vehicles of the same type) can be derived from the car-following model under equilibrium conditions. The specific process is as follows: When the car-following model reaches equilibrium, the speed difference between adjacent time points is set to zero, thus allowing the calculation of the corresponding headway. Then based on traffic flow density Distance from the front of the car The reciprocal relationship yields the traffic flow density in equilibrium. : (3) Then based on traffic flow Traffic flow density With driving speed The relationship between the three ( ), calculate the relationship between traffic flow and traffic flow density in congested sections: (4) Critical density At the intersection of the free-flow section and the congested section, through simultaneous equations... and form Calculate the critical density And calculate the maximum flow rate. ,make The value is zero, thus obtaining the blockage density. and according to Calculate the propagation speed of congested sections ; With free flow velocity and the speed of transmission in congested areas Two slopes, with maximum flow rate Using the vertex as the point, a basic macroscopic traffic flow diagram (e.g., a triangle) is formed in the traffic flow-traffic flow density coordinate system. Figure 3 (as shown) The maximum flow rate per lane for each road section is taken as follows: The maximum inflow rate for a single lane is set to [value]. .

[0027] Step S2: Taking the given road network as the optimization object, the planning time period is discretized into several decision time intervals. Multiple parameters of the given road network are used as decision variables. The maximum flow rate of a single lane, the maximum inflow rate of a single lane, and the congestion density are used as traffic capacity parameters. Under the conditions of satisfying the road segment transmission capacity constraints and road segment reception capacity constraints established by the road segment transmission model, as well as the flow conservation constraints, road segment constraints on the total number of lanes, lane change amplitude constraints of adjacent decision time intervals, and lane difference constraints of adjacent road segments, a dynamic lane reversal optimization model is established with the goal of minimizing the total system travel time. Preferably, assuming a given road network ,in For a set of nodes, This is a set of road segments. In a given road network, the node that generates origin-destination travel demand is called the origin node, and the node that the vehicle ultimately arrives at is called the destination node. and These represent the starting set of nodes and the ending set of nodes, respectively. The remaining nodes are called general nodes, and are represented by... express.

[0028] In a given road network, road segments are divided into three categories: starting segments connecting starting nodes, general road segments, and ending segments connecting ending nodes. A starting (ending) node connects to only one starting (ending) segment, and a starting (ending) segment connects to only one starting (ending) node. Both starting and ending segments are virtual road segments, assumed to have unlimited traffic demand receiving capacity, and it is assumed that vehicles are ultimately stored in the ending segment. Represents the initial set of road segments. Indicates the set of terminating road segments. This is the set of non-terminating road segments. For any general road segment... If there exists a road segment that is parallel to it in the opposite direction and has the same length and free-flow velocity... Then it is called For one road segment. For road segment pairs.

[0029] Specifically, the decision variables include: the number of lanes allocated to road segments in the defined road network, the cumulative inflow to road segments, the cumulative outflow to road segments, and the cumulative transfer flow between road segments.

[0030] Specifically, discretizing the planning period into several decision time intervals also includes: discretizing the planning period into decision time intervals and basic time intervals; Preferably, this embodiment plans the time period. First, it is discretized into a series of basic time intervals, and the set of basic time interval indices is... The basic time interval is And satisfy Secondly, from the perspective of traffic managers, based on the basic time interval duration... The frequency of dynamic adjustment of lane configuration schemes is determined by integer multiples of the planning period. It is further discretized into a series of decision time intervals, and the set of decision time interval indices is: The decision interval is ,in, The number of base time intervals included in each decision time interval.

[0031] Because lane configuration schemes are updated at the decision interval level, while traffic flow propagation is characterized at the base time interval level, therefore, within the same decision interval... Within this range, all corresponding basic time intervals share the same lane configuration scheme. Basic Time Interval Index Index of decision time intervals The correspondence between them is as follows: (5) in, This indicates the rounding up operation.

[0032] Specifically, traffic flow propagation in a road network is mainly described by the transmission and reception capabilities of road segments, as determined by Newell's simplified moving wave theory. This embodiment establishes constraints on transmission and reception capabilities of road segments using a road segment transmission model (Newell's simplified moving wave theory), including: Section In the Segment transmission capacity within a basic time interval for: (6) Section In the Segment reception capability within a basic time interval for: (7) For the road section and subsequent road sections In the Within each basic time interval, the road segment Rear connecting road sections The actual traffic flow transmitted was: (8) in, As of the date The cumulative number of entries into the road segment at the end of each basic time interval The number of vehicles, As of the date The cumulative number of entries into the road segment at the end of each basic time interval The number of vehicles, As of the date The cumulative number of road sections exited at the end of each basic time interval The number of vehicles, As of the date The cumulative number of road sections exited at the end of each basic time interval The number of vehicles, As of the date The cumulative number of road sections exited at the end of each basic time interval The number of vehicles is used as the cumulative outflow for a road segment. For road section The free-flow propagation time (in terms of the number of basic time intervals). For road section The propagation time of congested segments (in terms of the base time interval). For road section The length of the road segment For road section In the The outflow capacity of road segments within a basic time interval. ,in, For road section In the Maximum flow rate of a single lane within a basic time interval For road section In the Road segment inflow capacity within a basic time interval ,in, For road section In the Maximum inflow rate of a single lane within a basic time interval This is a set of non-terminating road segments. For road section The following road segments are collected. It is a set of indexes based on the basic time interval.

[0033] In formula (6), This indicates that under free-flow propagation conditions, in the first... Within a basic time interval, the road segment is ready for departure. The number of vehicles under the given conditions, the segment's sending capacity is taken as this value and the segment's outflow capacity. The smaller of the two; in formula (7), Indicates road segment The remaining unused storage space; the road segment's receiving capacity is calculated by taking this value and the road segment's inflow capacity. The smaller one.

[0034] Specifically, establishing road segment transmission capacity constraints and road segment reception capacity constraints through the road segment transmission model also includes: performing equivalent linearization on the road segment transmission capacity constraints and the road segment reception capacity constraints; and, under the conditions of satisfying the constraints after equivalent linearization, as well as the constraints of traffic conservation, road segment on total lane volume, lane change amplitude constraints of adjacent decision time intervals, and lane difference constraints of adjacent road segments, establishing a dynamic lane reversing optimization model with the goal of minimizing the total system travel time.

[0035] Preferably, formulas (6) to (8) are nonlinear expressions containing minimum value operations, which can lead to a nonconvex model and increase the difficulty of solving the problem. Therefore, formulas (6) to (8) are equivalently linearized to the following constraints: (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) in, , , , As an auxiliary binary variable, it is used to identify which of the four candidate minimum values ​​is the actual minimum value. For subsequent road sections The length of the road segment; constraints (9) to (12) ensure that the actual transmission flow does not exceed any one of the four candidate values ​​(upper bound constraint); constraints (13) to (16) combined with auxiliary binary variables ensure that the actual transmission flow takes the smallest of the four candidate values ​​(lower bound constraint); constraint (17) ensures that exactly one auxiliary binary variable takes the value 1.

[0036] The constraints (9) to (18) above equate to linearizing the minimum value operation. To help those skilled in the art understand the correctness of this equivalent linearization, the following supplementary explanation is provided: When ( When the actual transmission flow is 100%, the lower bound constraints in constraints (13) to (16) are valid (i.e., the actual transmission flow is 100%). No. (1 candidate values), while the other three constraints are automatically satisfied due to the existence of a large M (non-tight constraints); at the same time, constraints (9) to (12) guarantee the actual transmission flow. Each of the four candidate values. Therefore, when At that time, the actual transmission flow is equal to the first The condition is that there are three candidate values, and this value is less than or equal to the other three candidate values, thus achieving the effect of taking the minimum value. Constraints (17) and (18) guarantee that there is exactly one candidate value. Thus, the entire linearized system is equivalent to the original minimum value operation.

[0037] Specifically, flow conservation constraints include: For the initial road segment, the cumulative inflow to the segment equals the cumulative travel demand generated by the initial road segment at both the origin and destination points: (19) For non-starting road segments, the cumulative inflow to the segment equals the inflow to all preceding road segments. The sum of the cumulative transfer traffic between road sections: (20) For non-terminating road segments, the cumulative outflow of the segment equals the segment's total outflow. The sum of the cumulative transfer flows between road segments leading to all subsequent road segments: (twenty one) The non-negativity constraints and initial value constraints for the cumulative transfer flow between road segments are as follows: (twenty two) (twenty three) in, As of the date The cumulative number of entries into the road segment at the end of each basic time interval The number of vehicles, as the road segment Cumulative inflow As of the date The final segment of each basic time interval The cumulative amount of travel demand generated between origin and destination. For the initial set of road segments, This is the complete collection of road sections. For road section The preceding road segment collection, As of the date The basic time interval is not determined by the road segment. Flowing section The cumulative transfer traffic between road sections As of the date The basic time interval is not determined by the road segment. Flow to subsequent road sections The cumulative traffic flow transferred between road sections.

[0038] When the road section There are multiple subsequent road segments (i.e.) In this case, a feasible approach is to assume that each subsequent road segment is allocated resources from the road segment according to its receiving capacity. The outflow of vehicles, that is, for the road section Multiple subsequent road sections Diverted to subsequent road sections The traffic ratio is .

[0039] Specifically, the total number of lanes on the road segment is constrained as follows: (twenty four) Among them, the road section From the road section and road sections Road sections that are parallel to each other in opposite directions and have the same length and free-flow velocity. constitute, For the road section The total number of lanes on the corresponding road. For road segment pairs, For the set of decision time interval indices, It is a set of positive integers. For road section In the Number of lanes allocated to road segments within each decision time interval For road section In the Number of lanes allocated to a road segment within a decision time interval; The change in the number of lanes allocated on the same road segment between adjacent decision-making time intervals must not exceed a preset upper limit to ensure the smoothness of lane changing. (25) The absolute value constraint is linearized into a lane change magnitude constraint between adjacent decision time intervals, i.e.: (26) (27) in, For road section In the Number of lanes allocated to road segments within each decision time interval For road section The maximum number of lane changes allowed between two adjacent decision time intervals; The number of lanes allocated between adjacent road segments should maintain a certain continuity within the same decision-making time interval to prevent new traffic bottlenecks caused by mismatches in upstream and downstream traffic capacity. (28) The absolute value constraint is linearized into a lane difference constraint between adjacent road segments, i.e.: (29) (30) in, Road segments with connecting relationships and subsequent road sections The maximum difference in the number of lane assignments allowed within the same decision time interval.

[0040] Specifically, the objective of the dynamic lane-changing optimization model is to minimize the Total System Travel Time (TSTT). TSTT refers to the sum of the travel times of all vehicles entering the road network and ultimately reaching their destination. This embodiment obtains the average travel time of vehicles on the road segment by simulating the cumulative traffic flow curve in discrete time. The objective function is: (31) in, Let be the objective function. The decision variable is the base time interval duration. .

[0041] Based on the above objective function and constraints, the Dynamic Lane Reversing Optimization (DLR) model for fully connected autonomous vehicles is expressed as follows: (32) Constraints: Formulas (9) to (18) (linearized forms of road segment sending capacity constraints and road segment receiving capacity constraints), Formulas (19) to (21) (flow conservation constraints), Formulas (22) to (23) (non-negativity and initial value constraints of cumulative transfer flow between road segments), Formula (24) (road segment constraints on total lane volume), Formulas (26) to (27) (linearized forms of lane change range constraints for adjacent decision time intervals), Formulas (29) to (30) (linearized forms of lane difference constraints between adjacent road segments).

[0042] Step S3: At the beginning of each decision time interval, the system state correction, short-term origin-destination travel demand prediction, dynamic lane reversal optimization model solution and lane configuration scheme implementation are executed sequentially. Only the lane configuration scheme in the current control time domain is implemented in the road traffic network before proceeding to the next decision time. This process is repeated until the planning period ends.

[0043] Preferably, since the travel demand and traffic conditions at origin and destination points continuously change over time, optimizing the entire planning period based on long-term origin-destination travel demand predictions would result in long solution times and potential error accumulation. Therefore, this embodiment introduces a rolling time-domain optimization algorithm, which can accelerate the solution process, avoid long-term prediction errors, and respond promptly to time-varying traffic conditions.

[0044] This embodiment divides the rolling time-domain optimization process into a decision-making time domain and a control time domain. The decision-making time domain represents the time range considered in the future during each optimization, used to generate subsequent lane management plans by combining current traffic conditions and short-term demand forecasts. The control time domain represents the actual time range of execution of the above plans, i.e., only implementing the lane configuration plan for the first decision time interval of the current stage. Therefore, a decision must be made at the beginning of each decision time interval until the entire planning period is covered (e.g., ...). Figure 4 (As shown).

[0045] The rolling temporal optimization consists of the following four operations: Operation 1: Current State Estimation and Correction (System State Correction) In the At the start of each decision time interval, traffic state observations (cumulative inflow and outflow of road segments) are obtained from the real-time road traffic state data uploaded by connected autonomous vehicles at the end of the previous control time domain. These observations are then combined with the terminal state corresponding to the lane configuration scheme implemented in the previous control time domain (e.g., by directly replacing model predictions with observations, or by using a weighted average method (given the weights of the observations)). and model prediction weights (Alternatively, state estimation methods such as Kalman filtering can be used) to correct the system state. The corrected system state is then used as the initial state for the current rolling time-domain optimization, ensuring that the optimization process is always based on the latest traffic conditions.

[0046] Operation 2: Short-term origin-destination travel demand forecast Given the corrected initial state conditions, the short-term origin-destination travel demand within the current decision-making time domain is predicted using a demand forecasting model (e.g., a statistical forecasting method based on historical traffic data, a recursive forecasting method based on Kalman filtering, or a forecasting method based on machine learning). The predicted traffic demand of the starting segment is used as the exogenous input of the dynamic lane-changing optimization model (i.e., in formula (19)). ), used to characterize the expected traffic inflow within the future decision-making time domain.

[0047] Operation 3: Rolling time-domain optimization of lane configuration scheme (dynamic lane reversal optimization model) Based on the corrected initial system state and the predicted origin-destination travel demand, a dynamic lane-changing optimization model is constructed and solved within the current decision-making time domain to determine the optimal lane configuration scheme for each decision time interval within the current decision-making time domain. This invention is implemented using Python programming and uses the Gurobi solver to solve the mixed-integer linear programming (MILP) model.

[0048] Operation 4: Control Implementation and Time-Domain Update (Lane Configuration Scheme Implementation) After solving the dynamic lane-changing optimization model, the lane configuration schemes for the first control time domain (i.e., the current decision time interval) are implemented on the road traffic network. Subsequently, under the influence of the implemented lane configuration schemes, the system progresses to the start of the next decision time interval. At this moment, new real-time traffic data is collected again, the system state is updated, and operations one through four are repeated in a rolling manner until the end of the planned time period. Through the above operations, the original long-term dynamic lane management decision problem is decomposed into a series of short-term dynamic lane management sub-problems.

[0049] The following are the numerical experimental results of this embodiment. This embodiment conducts numerical experiments on an improved Nguyen and Dupuis (ND) road network, the road network topology of which is as follows: Figure 5 As shown. The experimental road network consists of 19 nodes, including 14 general nodes, 3 starting nodes, and 2 ending nodes. The lengths of each road segment are set as follows: 335m segments include r1-1, r2-4, 14-S1, r3-14, and 4-S2; 600m segments include 1-12, 7-8, 7-11, 9-13, 8-2, 10-11, 3-14, and 2-14; 750m segments include 12-8; and 800m segments include 1-5, 4-9, 5-9, 6-10, 11-2, 12-6, 4-5, 5-6, 6-7, 9-10, 11-3, and 13-3.

[0050] The experiment simultaneously set up three origin-endpoint pairs: r1→S1, r2→S1, and r3→S2. The first two types of origin-endpoint pairs can be categorized as traffic demand merging into node S1 from west to east, while the third type corresponds to traffic demand merging into node S2 from east to west. r1→1, r2→4, and r3→14 are virtual starting segments, while 14→S1 and 4→S2 are virtual ending segments. Except for the virtual segments, which only support one-way traffic, the other ordinary roads consist of segments in two directions, allowing for the reallocation of lane resources between the two directions.

[0051] Time-varying travel demand for each origin-destination pair Described in the following functional form: in, For origin and destination pairs The basic demand rate and These are the amplification factors for the first and second demand peaks, respectively. and These are the indexes of the basic time intervals between the occurrence of the first and second demand peaks, respectively. This refers to the peak demand diffusion parameter (unit: base time intervals). The demand parameter settings for the three start-end point pairs are as follows: Origin and end point pair r1→S1: , , , , , ; Origin and end point pair r2→S1: , , , , , ; Origin and end point pair r3→S2: , , , , , ; in, This represents the total number of basic time intervals within the planned time period.

[0052] Based on the calibration results of the actual vehicle test, the following error proportional control coefficient Carry-following error differential control coefficient Vehicle body length Control system step size Minimum safe parking distance Expected workshop time distance The free-flow velocity is obtained based on the Intelligent Car-Following Model (IDM). The critical density was calculated. Maximum flow rate per lane Blocking density The total duration of the planned time period is set at 3600 seconds, and the basic time interval is [duration missing]. Basic time interval index Decision interval duration ,therefore Decision time interval index Decision-making timeframe duration Controlling the duration of the time domain Initially, all roads are configured with three lanes in each direction, such as... Figure 6 As shown.

[0053] Experimental results show that in the first half of the planning period, due to the high traffic demand at the origin and destination points for r1→S1 and r2→S1, the road network as a whole exhibits a dominant traffic flow from west to east, with lane resources mainly tilted towards the west and east. In the second half of the planning period, the traffic demand at the origin and destination points for r3→S2 gradually increases and becomes dominant, the traffic flow direction of the road network changes to east to west, and lane resources gradually tilt towards the east and west. When the difference in traffic demand weakens, the lane configuration can return to a relatively balanced state. This is the period when demand peaks from west to east. For the peak hours of eastbound to westbound demand, the corresponding road network lane configuration results are as follows: Figure 7 and Figure 8 As shown.

[0054] To demonstrate the effectiveness of this embodiment in improving road network traffic efficiency, a fixed-lane baseline scenario (3 lanes in each direction) was further constructed under the same road network structure, demand input, and traffic flow parameters. The comparison results of the total system travel time and average travel time of the two schemes are as follows: Fixed-lane scheme , Dynamic lane changing scheme , The dynamic lane-changing scheme reduces the average travel time by approximately 39.8% and the total system travel time by approximately 40.6% compared to the fixed lane scheme, demonstrating a significant advantage.

[0055] Example 2 like Figure 2 As shown, this embodiment proposes a dynamic lane changing decision optimization system for fully connected autonomous driving scenarios, which specifically includes the following modules: The module for generating a macroscopic basic map is used to characterize the car-following behavior of connected autonomous vehicles through the car-following model of connected autonomous vehicles. Under the equilibrium state of the car-following model, the relationship between traffic flow and traffic flow density of the road segment is derived to obtain a triangular macroscopic basic map of traffic flow in the scenario of fully connected autonomous vehicles. Based on the triangular macroscopic basic map of traffic flow, the maximum flow rate of a single lane, the maximum inflow rate of a single lane, and the congestion density of each road segment are determined. A dynamic lane-changing optimization model module is established to take a given road network as the optimization object, discretize the planning time period into several decision time intervals, use multiple parameters of the given road network as decision variables, and use the maximum flow rate of a single lane, the maximum inflow rate of a single lane, and the congestion density as capacity parameters. Under the conditions of satisfying the constraints of road segment transmission capacity and road segment reception capacity established by the road segment transmission model, as well as the constraints of flow conservation, road segment on total lane volume, lane change amplitude of adjacent decision time intervals, and lane difference between adjacent road segments, the dynamic lane-changing optimization model is established with the goal of minimizing the total travel time of the system. The decision module is used to sequentially perform system state correction, short-term origin-destination travel demand prediction, dynamic lane reversal optimization model solution, and lane configuration scheme implementation at the beginning of each decision time interval. Only the lane configuration scheme in the current control time domain is implemented in the road traffic network before proceeding to the next decision time. This process is repeated until the end of the planning period.

[0056] Since the system technical solution of this embodiment 2 is based on the technical solution of embodiment 1, it will not be described again.

[0057] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A dynamic lane-changing decision optimization method for fully connected autonomous driving scenarios, characterized in that, include: The car-following behavior of connected autonomous vehicles is characterized by a car-following model. The relationship between traffic flow and traffic flow density of a road segment is derived under the equilibrium state of the car-following model, resulting in a triangular macroscopic basic diagram of traffic flow in the scenario of fully connected autonomous vehicles. Based on the triangular macroscopic basic diagram of traffic flow, the maximum flow rate, maximum inflow rate, and congestion density of each road segment are determined. Taking a given road network as the optimization object, the planning period is discretized into several decision time intervals. Multiple parameters of the given road network are used as decision variables. The maximum flow rate of a single lane, the maximum inflow rate of a single lane, and the congestion density are used as traffic capacity parameters. Under the conditions of satisfying the constraints of road segment transmission capacity and road segment reception capacity established by the road segment transmission model, as well as the constraints of flow conservation, road segment to total lane volume, lane change amplitude of adjacent decision time intervals, and lane difference between adjacent road segments, a dynamic lane reversal optimization model is established with the goal of minimizing the total system travel time. At the beginning of each decision time interval, the system state correction, short-term origin-destination travel demand prediction, dynamic lane reversal optimization model solution and lane configuration scheme implementation are executed sequentially. Only the lane configuration scheme in the current control time domain is implemented in the road traffic network before being advanced to the next decision time. This process is repeated until the end of the planning period.

2. The dynamic lane changing decision optimization method for fully connected autonomous driving scenarios as described in claim 1, characterized in that, The decision variables include: the number of lanes allocated to road segments in the defined road network, the cumulative inflow to road segments, the cumulative outflow to road segments, and the cumulative transfer flow between road segments.

3. The dynamic lane changing decision optimization method for fully connected autonomous driving scenarios as described in claim 1, characterized in that, The velocity update equation for the car-following model is: The following error equation of the following car model is: in, For the control system step size of connected autonomous vehicles, For connected autonomous vehicles The speed at any given moment For connected autonomous vehicles The speed at any given moment for The following error between the actual vehicle spacing and the expected vehicle spacing at any given time. for The first derivative with respect to time, for The actual headway of a connected autonomous vehicle. This refers to the body length of a connected autonomous vehicle. To ensure the minimum safe parking distance, To maintain a constant inter-vehicle distance for connected and autonomous vehicles, This is the proportional control coefficient for the car-following error. This is the differential control coefficient for the car-following error.

4. The dynamic lane changing decision optimization method for fully connected autonomous driving scenarios as described in claim 3, characterized in that, The macroscopic basic diagram of triangular traffic flow in a fully connected autonomous vehicle scenario includes: In the free-flow section, the traffic flow of the road segment Traffic flow density The relationship between them: in, For free flow velocity; During congested sections, the car-following model is brought to equilibrium, and the traffic flow density under equilibrium conditions is calculated. : Calculate the relationship between traffic flow and traffic flow density in congested sections: Critical density At the intersection of the free-flow section and the congested section, through simultaneous equations... and form Calculate the critical density And calculate the maximum flow rate. ,make The value is zero, thus obtaining the blockage density. and according to Calculate the propagation speed of congested sections ; With free flow velocity and the speed of transmission in congested areas Two slopes, with maximum flow rate Using the vertex as the point, a basic macroscopic diagram of the triangular traffic flow is formed in the traffic flow-traffic flow density coordinate system; The maximum flow rate per lane for each road section is taken as follows: The maximum inflow rate for a single lane is set to [value]. .

5. The dynamic lane changing decision optimization method for fully connected autonomous driving scenarios as described in claim 4, characterized in that, Discretizing the planning period into several decision time intervals also includes: discretizing the planning period into decision time intervals and basic time intervals; The constraints on the transmission capacity and reception capacity of road segments are established through a road segment transmission model, including: Section In the Segment transmission capacity within a basic time interval for: Section In the Segment reception capability within a basic time interval for: For the road section and subsequent road sections In the Within each basic time interval, the road segment Rear connecting road sections The actual traffic flow transmitted was: in, As of the date The cumulative number of entries into the road segment at the end of each basic time interval The number of vehicles, As of the date The cumulative number of entries into the road segment at the end of each basic time interval The number of vehicles, As of the date The cumulative number of road sections exited at the end of each basic time interval The number of vehicles, As of the date The cumulative number of road sections exited at the end of each basic time interval The number of vehicles, As of the date The cumulative number of road sections exited at the end of each basic time interval The number of vehicles is used as the cumulative outflow for a road segment. For road section Free-flow propagation time, For road section The transmission time during congested periods For road section The length of the road segment For road section In the The outflow capacity of road segments within a basic time interval. For road section In the Road segment inflow capacity within a basic time interval This is a set of non-terminating road segments. For road section The following road segments are collected. It is a set of indexes based on the basic time interval.

6. The dynamic lane changing decision optimization method for fully connected autonomous driving scenarios as described in claim 1, characterized in that, Establishing road segment transmission capacity constraints and road segment reception capacity constraints through the road segment transmission model also includes: performing equivalent linearization on the road segment transmission capacity constraints and the road segment reception capacity constraints; and, under the conditions of satisfying the constraints after equivalent linearization, as well as the constraints of traffic conservation, road segment on total lane volume, lane change amplitude constraints of adjacent decision time intervals, and lane difference constraints of adjacent road segments, establishing a dynamic lane reversing optimization model with the objective of minimizing the total system travel time.

7. The dynamic lane changing decision optimization method for fully connected autonomous driving scenarios as described in claim 5, characterized in that, Flow conservation constraints include: For the initial road segment, the cumulative inflow to the segment equals the cumulative travel demand generated by the initial road segment at both the origin and destination points: For non-starting road segments, the cumulative inflow to the segment equals the inflow to all preceding road segments. The sum of the cumulative transfer traffic between road sections: For non-terminating road segments, the cumulative outflow of the segment equals the segment's total outflow. The sum of the cumulative transfer flows between road segments leading to all subsequent road segments: The non-negativity constraints and initial value constraints for the cumulative transfer flow between road segments are as follows: in, As of the date The cumulative number of entries into the road segment at the end of each basic time interval The number of vehicles, as the road segment Cumulative inflow As of the date The final segment of each basic time interval The cumulative amount of travel demand generated between origin and destination. For the initial set of road segments, This is the complete collection of road sections. For road section The preceding road segment collection, As of the date The basic time interval is not determined by the road segment. Flowing section The cumulative transfer traffic between road sections As of the date The basic time interval is not determined by the road segment. Flow to subsequent road sections The cumulative traffic flow transferred between road sections.

8. The dynamic lane changing decision optimization method for fully connected autonomous driving scenarios as described in claim 7, characterized in that, The total number of lanes on the road section is constrained as follows: Among them, the road section From the road section and road sections Road sections that are parallel to each other in opposite directions and have the same length and free-flow velocity. constitute, For the road section The total number of lanes on the corresponding road. For road segment pairs, For the set of decision time interval indices, It is a set of positive integers. For road section In the Number of lanes allocated to road segments within each decision time interval For road section In the Number of lanes allocated to a road segment within a decision time interval; The lane change range constraint for adjacent decision time intervals is: in, For road section In the Number of lanes allocated to road segments within each decision time interval For road section The maximum number of lane changes allowed between two adjacent decision time intervals; The lane difference constraint between adjacent road segments is: in, Road segments with connecting relationships and subsequent road sections The maximum difference in the number of lane assignments allowed within the same decision time interval.

9. The dynamic lane changing decision optimization method for fully connected autonomous driving scenarios as described in claim 7, characterized in that, The dynamic lane-changing optimization model includes: in, Let be the objective function. The decision variable is the base time interval duration. .

10. A dynamic lane-changing decision optimization system for fully connected autonomous driving scenarios, characterized in that, include: The module for generating a macroscopic basic map is used to characterize the car-following behavior of connected autonomous vehicles through the car-following model of connected autonomous vehicles. Under the equilibrium state of the car-following model, the relationship between traffic flow and traffic flow density of the road segment is derived to obtain a triangular macroscopic basic map of traffic flow in the scenario of fully connected autonomous vehicles. Based on the triangular macroscopic basic map of traffic flow, the maximum flow rate of a single lane, the maximum inflow rate of a single lane, and the congestion density of each road segment are determined. A dynamic lane-changing optimization model module is established to take a given road network as the optimization object, discretize the planning time period into several decision time intervals, use multiple parameters of the given road network as decision variables, and use the maximum flow rate of a single lane, the maximum inflow rate of a single lane, and the congestion density as capacity parameters. Under the conditions of satisfying the constraints of road segment transmission capacity and road segment reception capacity established by the road segment transmission model, as well as the constraints of flow conservation, road segment on total lane volume, lane change amplitude of adjacent decision time intervals, and lane difference between adjacent road segments, the dynamic lane-changing optimization model is established with the goal of minimizing the total travel time of the system. The decision module is used to sequentially perform system state correction, short-term origin-destination travel demand prediction, dynamic lane reversal optimization model solution, and lane configuration scheme implementation at the beginning of each decision time interval. Only the lane configuration scheme in the current control time domain is implemented in the road traffic network before proceeding to the next decision time. This process is repeated until the end of the planning period.

Images