Macro-micro integrated alliance decision-making method based on intelligent network connection vehicle cluster

Abstract

The invention relates to the technical field of intelligent network connection vehicle cluster decision making, in particular to an intelligent network connection vehicle cluster macro-micro integrated alliance decision making method, which constructs a cluster alliance decision making cost function by constructing an initial path of an intelligent network connection vehicle cluster and considering safety, comfort, efficiency and energy consumption targets. Under the condition that kinematics constraint and comfort conditions of the intelligent network connection automobile are met, an adaptive coevolution algorithm is used for solving, a diversity-based search distribution mechanism is adopted in the adaptive coevolution algorithm for population diversity distribution and traffic flow guide variation, a global exploration module and a local optimization module are optimized respectively, and a global optimization module and a local optimization module are optimized respectively. And operations such as periodic global optimization are carried out to generate an optimal intelligent network connection vehicle cluster decision, so that the vehicle can reduce energy consumption and improve comfort on the premise of ensuring the safety to the maximum extent and improving the efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent connected vehicle cluster decision-making, and particularly relates to a macro-micro integrated coalition decision-making method for intelligent connected vehicle clusters. Background Art

[0002] Intelligent driving decision-making is the core link during vehicle driving. It receives and analyzes traffic information between vehicles in real time, makes precise judgments and responses, so as to achieve safe and efficient interaction between vehicles. This process not only concerns the safety of driving, ensuring that every step during the driving process is stable and reliable, but is also crucial for improving driving efficiency. It can optimize traffic flow, reduce congestion, and thus increase the overall driving speed.

[0003] In recent years, a series of theories for intelligent connected vehicle cluster decision-making have been proposed at home and abroad. The research on intelligent connected vehicle cluster decision-making can be divided into two types from the mechanism level: centralized decision-making and distributed decision-making. The centralized intelligent connected vehicle cluster decision-making mechanism refers to a decision-making mode in which a global central controller formulates a traffic plan for all intelligent connected vehicle clusters. The distributed decision-making mechanism of intelligent connected vehicle clusters refers to a decision-making mode in which vehicles in the traffic system independently formulate a traffic plan after collecting global / local traffic information without relying on a central controller.

[0004] Cluster decision-making is a complex non-deterministic problem, and its computational difficulty increases exponentially with the increase in the number of vehicles, making traditional algorithms face huge time challenges when dealing with multi-vehicle trajectory planning. The Adaptive Coevolutionary Algorithm (ACA) can effectively cope with high-dimensional complex optimization problems by executing evolutionary algorithms in parallel on multiple subsystems and exchanging information to co-optimize the population. ACA not only improves the search efficiency and accuracy, but also can make full use of the computing resources of intelligent connected vehicles to optimize the vehicle group trajectories of the local road network within a limited time.

[0005] Intelligent connected vehicle cluster decision-making in the urban local road network has become a hot and difficult issue in the field of intelligent transportation. How to enable vehicles to drive safely, quickly, energy-efficiently and comfortably in the urban local road network has become an important research direction in the research of intelligent transportation systems.

[0006] For example, Chinese Patent Publication No.: CN117116043A discloses a method for macroscopic path planning of urban road network vehicles facing frequent congestion, belonging to the field of traffic path planning. This method adopts a two-layer planning model, including macroscopic path planning and local path planning. In macroscopic path planning, the existing traffic sub-region division method and macroscopic traffic flow distribution strategy are used to achieve effective management of large-scale urban road networks. Local path planning, based on the traffic sub-region passing sequence given by the macroscopic planning, uses the A* algorithm within each traffic sub-region to plan a shortest-time path for the vehicle from the boundary of the previous sub-region to the boundary of the next sub-region. In order to cope with frequent traffic congestion in urban road networks, the present invention applies the flow equilibrium theory to conduct traffic flow distribution at the macroscopic level from a global perspective. The invention performs flow scheduling between the current traffic sub-regions according to the macroscopic path planning algorithm, so as to make the urban traffic flow reach the global optimum.

[0007] However, in the prior art, the following problems still exist:

[0008] 1. In the traffic road network scenario, currently a single vehicle mainly relies on vehicle intelligence for autonomous decision-making, lacking cooperation with other vehicles and roadside facilities, and it is difficult to comprehensively obtain complex traffic environment information to achieve global optimal decision-making. In order to overcome the limitations of single-vehicle decision-making, the information exchange and sharing, collaborative perception and cooperation mechanism among intelligent connected vehicle clusters enable intelligent connected vehicle clusters to obtain more comprehensive traffic information, and thus make better decisions.

[0009] 2. The centralized vehicle group collaborative decision-making method will encounter extremely high computational complexity when facing the traffic road network scenario, and the causal loop in the road network scenario exacerbates the problem-solving difficulty. In this case, the distributed vehicle group collaborative decision-making mechanism in the road network scenario is expected to achieve a balance between coordination performance and computational efficiency.

[0010] 3. Currently, the relevant research on trajectory optimization in urban roads all has some idealized assumptions, and there is still room for improvement in the solutions that take into account comfort, traffic efficiency, energy conservation and safety.

[0011] 4. For the dynamic topology and high mobility of intelligent connected vehicles, traditional formation methods mainly focus on the following behavior of vehicles, but in urban roads, there are a large number of other driving behaviors such as lane changing and turning. Traditional formation methods are difficult to meet the requirements, so a macro-micro integrated coalition decision-making method for intelligent connected vehicle clusters is proposed. Summary of the Invention

[0012] To this end, the present invention provides a macro-micro integrated coalition decision-making method for intelligent connected vehicle clusters to overcome the problems in the prior art that for the dynamic topology and high mobility of intelligent connected vehicles, traditional formation methods mainly focus on the following behavior of vehicles, but in urban roads, there are a large number of other driving behaviors such as lane changes and turns, and traditional formation methods are difficult to meet the requirements;

[0013] To achieve the above object, the present invention provides a macro-micro integrated coalition decision-making method for intelligent connected vehicle clusters, including:

[0014] Construct an initial path for the intelligent connected vehicle cluster;

[0015] Construct a coalition set for the intelligent connected vehicle cluster, and construct a cost function for a single intelligent connected vehicle in the coalition set, including a driving safety cost function, a riding comfort cost function, a driving efficiency cost function, and an energy consumption cost function;

[0016] Construct a coalition game model for the coalition set, and set the optimization goal to minimize the cost function;

[0017] Determine the constraint conditions for game decision-making, including kinematic constraints and comfort constraints;

[0018] In the process of solving the coalition model, each intelligent connected vehicle is determined as an independent population for evolution. Every fixed number of evolution times, the intelligent connected vehicles exchange information through V2X communication technology to achieve global optimization;

[0019] Repeat the solving process until the termination condition is met to obtain the optimal trajectory solution.

[0020] Further, in the step S1, the process of constructing the initial path of the intelligent connected vehicle cluster includes,

[0021] Pre-construct road network data, including dividing nodes and constructing a directed graph for the road network;

[0022] Each intelligent connected vehicle loads the road network data to determine the starting point and the ending point and independently runs the A* algorithm to generate an initial path.

[0023] Further, in the step S2, the driving safety cost function is the sum of the longitudinal safety cost and the lateral safety cost;

[0024] The riding comfort cost function is determined based on the lateral jerk of the intelligent connected vehicle;

[0025] The driving efficiency cost function is determined based on the longitudinal speed;

[0026] The energy consumption cost function is determined based on air density, vehicle frontal area, air drag coefficient, rolling resistance coefficient, vehicle mass, gravitational acceleration, vehicle comprehensive acceleration, fuel consumption rate, and engine efficiency;

[0027] Among them, the longitudinal safety cost is related to the longitudinal gap and the relative speed with respect to the vehicle in front, and the lateral safety cost is related to the relative distance and relative speed between the connected and autonomous vehicle and adjacent vehicles.

[0028] Furthermore, the process of constructing a coalition game model for the coalition set includes,

[0029] Performing coalition initialization, dividing coalitions based on the direction labels of vehicles, including a straight-ahead coalition, a left-turn coalition, and a right-turn coalition;

[0030] Determining the preference order of the coalition game process;

[0031] Determining the behavioral conditions for coalition merging, coalition splitting, and coalition swapping.

[0032] Furthermore, it also includes determining whether the coalition is in a stable state, where,

[0033] If all connected and autonomous vehicles adopt the preference order, and each of the connected and autonomous vehicles cannot reduce its own cost by joining or exiting the coalition, then the coalition is in a stable state.

[0034] Furthermore, in step S4, the kinematic constraints are determined based on the acceleration value, speed value, front wheel steering angle, yaw angle, maximum acceleration, minimum steering angle, maximum steering angle, minimum yaw angle, and maximum yaw angle of the connected and autonomous vehicle;

[0035] The comfort constraints are constructed based on the front wheel steering angle, front wheel steering angle change rate, acceleration increment, and jerk.

[0036] Furthermore, each connected and autonomous vehicle is determined as an independent population for evolution, where,

[0037] Performing gene coding on the connected and autonomous vehicle to obtain a chromosome for the connected and autonomous vehicle;

[0038] Allocating populations to each connected and autonomous vehicle, and evaluating the fitness of the chromosomes that meet the constraint conditions to determine the evaluation value for each population;

[0039] Adjusting the proportion of global exploration and local optimization of individuals in the population during the evolution process according to the evaluation value;

[0040] Among them, the evaluation value is determined based on the Euclidean distance between the chromosomes corresponding to the connected and autonomous vehicles.

[0041] Further, for local optimization, an individual mutation rate of the intelligent connected vehicle is determined based on traffic flow.

[0042] Further, during the population evolution process, the number of chromosomes for the crossover operation of the intelligent connected vehicle is determined, and the crossover operation is performed by the chromosome single-point crossover method.

[0043] Further, in step S5, the termination condition is that the number of cycles of global optimization reaches a set maximum value or the total fitness of the intelligent connected vehicle shows no improvement under a predetermined number of sub-global optimizations.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows: by constructing an initial path of an intelligent connected vehicle cluster, considering the goals of safety, comfort, efficiency, and energy consumption, a cluster coalition decision-making cost function is constructed. Under the condition of satisfying the kinematic constraints and comfort conditions of the intelligent connected vehicle, an adaptive co-evolution algorithm is used for solving. In the adaptive co-evolution algorithm, a search and allocation mechanism based on diversity is adopted for population diversity allocation and mutation of traffic flow guidance. The global exploration and local optimization modules are respectively optimized, and periodic global optimization and other operations are performed to generate an optimal intelligent connected vehicle cluster decision, so that the vehicle can reduce energy consumption and improve comfort on the premise of maximizing safety and improving efficiency.

[0045] In particular, when constructing the cost function, the present invention comprehensively considers factors such as driving safety, riding comfort, driving efficiency, and energy consumption cost, and then obtains the decision of the cluster subsequently, so that the intelligent connected vehicle can improve driving safety, riding comfort, driving efficiency while reducing energy consumption.

[0046] In particular, the present invention constructs a coalition game model for the coalition set. Through coalition merger, coalition split, and the preference order in the coalition game process, with the goal of finding the optimal solution of the combined state among intelligent connected vehicles, the stable state of the coalition is obtained. The problem of minimizing the sum cost based on game decision-making is finally transformed into a closed-loop iterative optimization process with multiple constraint conditions, and an efficient adaptive co-evolution algorithm is used for solving. It can make more efficient, more accurate, and more reliable decisions.

[0047] In particular, the present invention determines each intelligent connected vehicle as an independent population for evolution, encodes the genes of the intelligent connected vehicle to obtain chromosomes for the intelligent connected vehicle. Secondly, each intelligent connected vehicle is assigned a population, and the fitness of the chromosomes that meet the constraint conditions is evaluated to determine the evaluation value for each population; according to the evaluation value, the proportion of individuals in the population for global exploration and local optimization during the evolution process is adjusted. DSAM aims to solve the problem of rapid decline in population diversity by dynamically balancing global exploration and local optimization, thereby effectively avoiding the algorithm from falling into a local optimal solution. Brief Description of the Drawings

[0048] Figure 1 It is a step diagram of the macro-micro integrated coalition decision-making method for intelligent connected vehicle clusters based on the present invention;

[0049] Figure 2 It is a schematic diagram of the node of the urban local road network in the embodiment of the present invention;

[0050] Figure 3 It is a schematic diagram of the coordinate system of the kinematic model of the intelligent connected vehicle in the embodiment of the present invention;

[0051] Figure 4 It is a schematic diagram of the gene encoding of the co-evolution algorithm in the embodiment of the present invention;

[0052] Figure 5 It is a schematic diagram of the search and allocation mechanism based on diversity in the embodiment of the present invention;

[0053] Figure 6 It is a schematic diagram of the mutation guided by traffic flow in the embodiment of the present invention. Detailed Embodiments

[0054] In order to make the objectives and advantages of the present invention clearer and more understandable, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0055] The preferred embodiments of the present invention will be described below with reference to the drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principle of the present invention and do not limit the protection scope of the present invention.

[0056] Please refer to Figure 1 as shown, which is a step diagram of the macro-micro integrated coalition decision-making method for intelligent connected vehicle clusters based on the present invention. The macro-micro integrated coalition decision-making method for intelligent connected vehicle clusters in this embodiment includes:

[0057] Step S1, constructing an initial path for the intelligent connected vehicle cluster;

[0058] Step S2, constructing a coalition set for the intelligent connected vehicle cluster, and constructing a cost function for a single intelligent connected vehicle in the coalition set, including a driving safety cost function, a riding comfort cost function, a driving efficiency cost function, and an energy consumption cost function;

[0059] Step S3, constructing a coalition game model for the coalition set, and setting the optimization objective to minimize the cost function;

[0060] Step S4, determining the constraint conditions for the game decision-making, including kinematic constraints and comfort constraints;

[0061] Step S5, during the solution process of the coalition model, each intelligent connected vehicle is determined as an independent population for evolution. Every fixed number of evolution times, each of the intelligent connected vehicles exchanges information through V2X communication technology to achieve global optimization;

[0062] Step S6, repeat Step S5 until the termination condition is met to obtain the optimal trajectory solution.

[0063] Specifically, in the said Step S1, the process of constructing the initial path of the intelligent connected vehicle cluster includes,

[0064] Pre-construct road network data, including dividing nodes and constructing a directed graph for the road network;

[0065] Please refer to Figure 2 as shown, which is the schematic diagram of the urban local road network nodes in the embodiment of the present invention. In implementation, those skilled in the art can, based on the urban local road network, divide the starting point and the ending point of each road section as nodes. These nodes represent the connection points of the roads, including intersections, starting and ending positions of road sections, etc., which will not be elaborated here.

[0066] Specifically, to achieve the efficiency and accuracy of path planning, a directed graph for the road network can be constructed. Among them, the entire road network is represented as a directed graph G=(V, E), where V is the set of nodes and E is the set of edges. For each road section, the connection relationship between its starting point and ending point is represented as a directed edge. The attributes of the edge include information such as the length and speed limit of the road section. At the same time, according to the connectivity of the actual road, the reachable relationship between nodes is determined.

[0067] Specifically, first rasterize the road network. The length and width of each grid are both l grid , and the center of each grid is called a node. Each intelligent connected vehicle loads the road network data to determine the starting point and the ending point and independently runs the A* algorithm to generate the initial path.

[0068] It can be understood that, based on the traditional A* algorithm, a heuristic function is designed for the characteristics of the urban road network. Combining the Manhattan distance (applicable to grid roads) and the direction-corrected Euclidean distance (for dealing with curved road sections), the heuristic function is automatically switched according to the topological characteristics of the road section. The cost estimation function f(n) of the intelligent connected vehicle from the starting point to the ending point is shown in Equation (1)

[0069]

[0070] Among them, g(n) is the actual cost from the starting point to the current node. h(n) is the estimated cost from the current node to the ending point, also called the heuristic function. x n 、y n are respectively the horizontal and vertical coordinates of the current node. x goal, y goal are the horizontal and vertical coordinates of the end point respectively. α is the direction angle correction coefficient, which can be dynamically calculated according to the historical path curvature and can ensure that the heuristic function is close to the actual road conditions.

[0071] The initial path consists of discrete path points According to the channelization signs of the lanes where the path points are located, direction labels are assigned to each vehicle. Straight ahead corresponds to "S", left turn corresponds to "L", and right turn corresponds to "R".

[0072] Specifically, in step S2, when constructing the coalition set for the intelligent connected vehicle cluster, the set of intelligent connected vehicles is represented as M == {1, 2,..., M}, where M is the number of intelligent connected vehicles in the intelligent connected vehicle cluster. The intelligent connected vehicles can cooperate with each other to form G different coalitions, and the coalition set is represented as G = {1, 2,..., G}.

[0073] Specifically, in the said step S2, the driving safety cost function is the sum of the longitudinal safety cost and the lateral safety cost;

[0074] In some possible implementations, the driving safety cost function is represented by formula (2),

[0075]

[0076] In formula (2), represents the longitudinal safety cost. represents the lateral safety cost.

[0077] Longitudinal safety cost is related to the longitudinal gap and the relative speed with respect to the vehicle in front, and can be shown by formulas (3) and (4).

[0078]

[0079] In formula (3) and formula (4), represents the longitudinal speed of intelligent connected vehicle i - 1, that is, the longitudinal speed of the vehicle in front. represents the longitudinal speed of intelligent connected vehicle i. X i-1 represents the longitudinal position of intelligent connected vehicle i - 1, X i represents the longitudinal position of intelligent connected vehicle i. represents the first weighting coefficient, represents the second weighting coefficient. ε represents the design parameter, aiming to avoid the situation of zero denominator. L i is the vehicle length. is the switching function. If the vehicle in front is faster than the following vehicle, that is then the switching function Then the cost of longitudinal safety is only related to the relative distance. Otherwise, longitudinal safety is related to both the relative distance and the relative speed.

[0080] Lateral safety cost is related to the relative distance and relative speed with respect to the adjacent vehicle (neighbor vehicle, NV), and can be expressed by Equations (5) and (6):

[0081]

[0082] In Equations (5) and (6), are the longitudinal speeds of the connected and automated vehicle i and its adjacent vehicle respectively. (X NV , Y NV ) is the position of the adjacent vehicle. represents the third weighting coefficient, represents the fourth weighting coefficient. If i.e., then the cost of lateral safety is only related to the relative distance. Otherwise, it is related to both the relative distance and the relative speed.

[0083] The ride comfort cost function is determined based on the lateral and longitudinal jerks of the connected and automated vehicle;

[0084] In some possible implementations, it is determined according to Equation (7),

[0085]

[0086] where and are the longitudinal jerk and lateral jerk of the connected and automated vehicle i respectively, represents the fifth weighting coefficient, represents the sixth weighting coefficient.

[0087] The driving efficiency cost function is determined based on the longitudinal speed;

[0088] In some possible implementations,

[0089] Driving efficiency cost is a function of the longitudinal speed of the connected and automated vehicle i, as shown in Equations (8) and (9):

[0090]

[0091] In Equations (8) and (9), represent the longitudinal speed and the longitudinal desired speed of the connected and automated vehicle i respectively. represents the speed limit of the lane, represents the speed of the vehicle in front of the connected and automated vehicle i, and is the seventh weighting coefficient.

[0092] The energy consumption cost function is determined based on air density, vehicle frontal area, air drag coefficient, rolling resistance coefficient, vehicle mass, gravitational acceleration, vehicle comprehensive acceleration, fuel consumption rate, and engine efficiency;

[0093] In some possible implementations,

[0094] The energy consumption of a fuel vehicle is mainly related to factors such as engine efficiency, vehicle mass, air resistance, and rolling resistance. The following is the cost expression for the energy consumption of an intelligent connected vehicle:

[0095]

[0096] In formula (10), ρ is the air density, A is the vehicle frontal area, C d is the air drag coefficient, C g is the rolling resistance coefficient, m is the vehicle mass, g is the gravitational acceleration, a is the vehicle comprehensive acceleration, b is the fuel consumption rate, and η is the engine efficiency.

[0097] The goal of each intelligent connected vehicle is to maximize the driving safety, comfort, and passing efficiency of the intelligent connected vehicle while minimizing the energy consumption of the vehicle, that is, minimizing the decision cost function. To achieve this optimization goal, the decision cost function of each intelligent connected vehicle i is expressed as in Equation (11).

[0098]

[0099] In formula (11), represents the driving safety cost function, represents the ride comfort cost function, represents the driving efficiency cost function and represents the energy consumption cost function, represents the first driving characteristic weighting coefficient, represents the second driving characteristic weighting coefficient, represents the third driving characteristic weighting coefficient, represents the fourth driving characteristic weighting coefficient.

[0100] The optimization goal of the intelligent connected vehicle cluster is to minimize the total cost of the coalition, as shown in Equation (12).

[0101]

[0102] In formula (12), the key to minimizing the optimization objective lies in dealing with the cooperation relationship among intelligent connected vehicles from the perspective of coalition games. This is because the total cost in (12) consists of driving safety, ride comfort, driving efficiency, and energy consumption costs, which depends on the coalition formation result of intelligent connected vehicles. The optimization objective in formula (12) is also transformed into the problem of minimizing the cost function in (13) from the perspective of coalition games.

[0103] Specifically, in step S3, the process of constructing a coalition game model for the coalition set includes

[0104] Performing coalition initialization, dividing coalitions based on the direction labels of vehicles, including a straight coalition, a left-turn coalition, and a right-turn coalition;

[0105] Determining the preference order of the coalition game process;

[0106] Determining the action conditions for coalition merging, coalition splitting, and coalition swapping.

[0107] Specifically, in the coalition formation game model, the path direction of intelligent connected vehicles, the cost of intelligent connected vehicle coalitions, and optional strategies need to be considered. In the coalition initialization stage, vehicles with the direction label "S" are automatically assigned to the straight coalition (S-Coalition), vehicles with the direction label "L" form the left-turn coalition (L-Coalition), and vehicles with the direction label "R" are grouped into the right-turn coalition (R-Coalition) to perform coalition initialization.

[0108] The game model of intelligent connected vehicles can be expressed as {β i , G g , S i , J i , r J}, where the decision of the i-th intelligent connected vehicle is β i . If β i = g, it means that intelligent connected vehicle i joins coalition g; G g is the set of intelligent connected vehicles within the g-th coalition; S i is the strategy set of intelligent connected vehicle i. The optional decision space of intelligent connected vehicle i includes all coalitions G = {1, 2,..., G}; J i is the cost function of the i-th intelligent connected vehicle, and its definition is shown in formula (11). The cost function of the g-th coalition is represented by r J (G g ) and can be defined as:

[0109]

[0110] Formula (13), the cost function of the g-th coalition is determined by the entire set of intelligent connected vehicles G within the coalitiong is determined by the total cost. In addition, it is defined that the alliances are non - overlapping, and an intelligent connected vehicle can only join one alliance at a time.

[0111] Specifically, for the i - th intelligent connected vehicle \(i\in M\), for the i - th intelligent connected vehicle, when time, the \(g\) - th alliance \(G\) g is superior to the \(k\) - th alliance \(G\) k , which means that the i - th intelligent connected vehicle is more inclined to join the alliance \(G\) g rather than \(G\) k . \(M\) represents the set of all intelligent connected vehicles. represents the priority ranking of the intelligent connected vehicle \(i\) for potential alliances.

[0112] It can be understood that there are various types of preference orders, and different preference orders lead to different alliance results. In this embodiment, the cooperation order is adopted.

[0113]

[0114] In Equation (14), \(r\) J (\(G\) g \(\cup i\)) represents the cost after the intelligent connected vehicle \(i\) joins the alliance \(G\) g . \(r\) J (\(G\) g ) represents the cost of the alliance \(G\) g . \(r\) J (\(G\) k ) represents the cost of the alliance \(G\) k . \(r\) J (\(G\) k \(\cup i\)) represents the total cost after the intelligent connected vehicle \(i\) joins the alliance \(G\) k . \(M\) represents the set of all intelligent connected vehicles. \(G\) is the set of all intelligent connected vehicle alliances. If the total cost becomes smaller when the i - th intelligent connected vehicle joins the alliance, then the intelligent connected vehicle will choose to stay in the new alliance and withdraw from the old alliance. The cooperation order aims to minimize the cost of the alliance, rather than minimizing the cost of a single intelligent connected vehicle. The behavior of alliance members has two types: joining or withdrawing, which is the way for the alliance to change its state, and at the alliance level, it is manifested as the merger or split of the alliance.

[0115] Determine the behavioral conditions for alliance merger, alliance split, and alliance exchange, where,

[0116] Regarding alliance merger,

[0117] If the new alliance (excluding the i - th intelligent connected vehicle) takes a merger action, and after the i - th intelligent connected vehicle joins the new alliance, the cost function of the original alliance does not increase, then the merger is allowed:

[0118]

[0119] As can be seen from (15), when the \(i\)-th connected and autonomous vehicle (CAV) is about to withdraw from the previous coalition and join a new coalition, the overall cost of all CAVs in the coalition will not increase, and the performance of the coalition can be improved.

[0120] Regarding coalition splitting,

[0121] If the coalition containing the \(i\)-th CAV undergoes a splitting action and the \(i\)-th CAV withdraws from the coalition, and the cost of its original coalition does not increase, then the splitting is allowed. As shown in Equation (16)

[0122]

[0123] It can be seen that the basic actions of the coalition are brought about by the joining and withdrawal actions of CAVs. However, when a CAV joins or withdraws, a temporary increase in the total cost of all CAVs may occur.

[0124] Regarding coalition swapping,

[0125] A swapping behavior occurs between coalitions when the cost of the new coalition that the \(i\)-th CAV joins is less than the cost of the old coalition that the CAV withdraws from.

[0126]

[0127] Equation (17) indicates that the total cost after swapping is not higher than that before swapping to ensure that the global performance of the system does not degrade.

[0128] Specifically, when all CAVs adopt a preference order and each CAV cannot reduce its own cost by joining or withdrawing from a coalition, the coalition is in a stable state, that is, it reaches the Nash equilibrium state, as shown in Equation (18):

[0129] \(i\in M, G\) g , \(r\) J \((G\) g )\leq r\) J \((G\) g \cup i)(18)

[0130] Since the set of CAVs and the number of formed coalitions are finite, and at the same time the combined states of all CAVs are finite. Therefore, the goal is to find the optimal solution of the combined states among CAVs, converge to a stable coalition partition, and finally reach a stable state.

[0131] Specifically, in the step S4, the kinematic constraint is determined based on the acceleration value, speed value, front wheel steering angle, yaw angle, maximum acceleration, minimum steering angle, maximum steering angle, minimum yaw angle, and maximum yaw angle of the intelligent connected vehicle;

[0132] The comfort constraint is constructed based on the front wheel steering angle, front wheel steering angle change rate, acceleration increment, and jerk.

[0133] Specifically, please refer to Figure 3 As shown, it is a schematic diagram of the coordinate system of the kinematic model of the intelligent connected vehicle according to an embodiment of the present invention. The present invention selects a two-degree-of-freedom kinematic model with higher accuracy and lower computational resource occupancy as the kinematic model of the intelligent connected vehicle. The kinematic model of the intelligent connected vehicle with the center of the rear axle as the origin of the vehicle body coordinate system is shown in Equation (19).

[0134] Regarding the kinematic constraint, the kinematic constraint equation can be expressed by Equation (20).

[0135]

[0136] In Equation (20), a i is the acceleration value of the intelligent connected vehicle i; v i is the speed value of the intelligent connected vehicle; δ i is the front wheel steering angle of the intelligent connected vehicle; is the yaw angle of the intelligent connected vehicle; a min and a max are the minimum acceleration and maximum acceleration of the intelligent connected vehicle respectively; δ min and δ min are the minimum and maximum steering angles of the intelligent connected vehicle respectively; and are the minimum and maximum yaw angles of the intelligent connected vehicle respectively.

[0137] In addition to the motion constraints and its own physical limitations of the intelligent connected vehicle, comfort constraint conditions also need to be considered in the decision-making process. The comfort constraint covers the limitations of the front wheel steering angle and its change rate, and also includes the limitations of acceleration change and jerk, so as to ensure the comfort of the vehicle.

[0138] The constraints of the front wheel steering angle and its increment are shown in Equation (21):

[0139]

[0140] In Equation 21, is the front wheel steering angle of the intelligent connected vehicle i, is the maximum front wheel steering angle of the intelligent connected vehicle i. and respectively represent the front wheel angle and the maximum front wheel angle of the intelligent connected vehicle i within a unit step length.

[0141] Acceleration increment The constraint is defined as shown in Equation (22).

[0142]

[0143] In Equation 22, and respectively represent the longitudinal acceleration and the maximum acceleration of the intelligent connected vehicle i within a unit step length.

[0144] The jerk constraint condition is as shown in Equation (23):

[0145]

[0146] In Equation 23, and respectively represent the longitudinal jerk and the maximum longitudinal jerk of the intelligent connected vehicle i, and respectively represent the lateral jerk and the maximum lateral jerk of the intelligent connected vehicle i.

[0147] Specifically, in step S5, during the solution process of the coalition model, each intelligent connected vehicle is determined as an independent population for evolution. Every fixed number of evolution times, each of the intelligent connected vehicles exchanges information through V2X communication technology to achieve global optimization;

[0148] Genetic coding is performed on the intelligent connected vehicle to obtain a chromosome for the intelligent connected vehicle;

[0149] A population is assigned to each intelligent connected vehicle, and the fitness of the chromosomes that meet the constraint conditions is evaluated to determine the evaluation value for each population;

[0150] According to the evaluation value, adjust the ratio of global exploration and local optimization of individuals in the population during the evolution process;

[0151] Among them, the evaluation value is determined based on the Euclidean distance between the chromosomes corresponding to the intelligent connected vehicles.

[0152] For local optimization, the individual mutation rate of the intelligent connected vehicle is determined based on the traffic flow.

[0153] During the population evolution process, determine the number of chromosomes for crossover operation of the intelligent connected vehicle, and perform crossover operation by means of single-point crossover of chromosomes.

[0154] Regarding genetic coding,

[0155] Please refer to Figure 4As shown, it is a schematic diagram of gene coding for the co-evolution algorithm in an embodiment of the present invention. According to the preliminary discrete driving path points of the intelligent connected vehicle cluster, individual-level gene coding can be constructed. In the gene coding of intelligent connected vehicle i, represents the path point coding value of intelligent connected vehicle i, and l represents the number of rows in the path point coding area, corresponding to the number of lanes; the chromosome is composed of a total of N W genes.

[0156] The solution of the cluster decision contains the node set of the driving path points obtained by the A* algorithm for each intelligent connected vehicle Acceleration decision set Front wheel steering angle decision set Alliance decision set Steering decision set S represents going straight, L represents turning left, R represents turning right, and n represents the number of the path point. Among them represents the acceleration of intelligent connected vehicle i at path point n, represents the front wheel steering angle of intelligent connected vehicle i at path point n, represents the alliance to which intelligent connected vehicle i belongs at path point n, represents the direction label of intelligent connected vehicle i at path point n.

[0157] Regarding the fitness evaluation of population initialization,

[0158] To initialize the adaptive co-evolution algorithm, according to the known lane-level path, a population containing P0 individuals is assigned to each intelligent connected vehicle. An initial solution that meets the constraint conditions is generated.

[0159] The present invention performs fitness evaluation on the chromosomes that meet the constraint conditions. The fitness evaluation function includes comfort, energy consumption, and intersection passing time, and a penalty weight P is set for the results that violate the constraint conditions pun . At the same time, in order to reasonably construct the fitness function of intelligent connected vehicles, after transforming Equation (11), Equation (24) can be obtained.

[0160]

[0161] Therefore, the fitness Si of the i-th intelligent connected vehicle in the optimization process of the distributed genetic algorithm can be defined i as shown in Equation (25).

[0162]

[0163] In the formula represents driving safety, represents riding comfort, represents driving efficiency, Indicates the energy consumption cost, Indicates the eighth weighting coefficient, Indicates the ninth weighting coefficient, Indicates the tenth weighting coefficient and Indicates the eleventh weighting coefficient.

[0164] Regarding the dynamic balance of global exploration and local optimization,

[0165] The present invention adopts a search allocation mechanism based on population diversity (Diversity-based Search Allocation Mechanism, DSAM). DSAM aims to solve the problem of rapid decline in population diversity by dynamically balancing global exploration and local optimization, thereby effectively avoiding the algorithm from falling into local optimal solutions.

[0166] The main idea of this mechanism is to divide individuals into two independent parts according to population diversity: global exploration and local optimization parts. The individuals for global exploration are used to search for unknown regions, while local optimization focuses on finding better solutions in the current optimal region. After measuring the population diversity, individuals will be assigned to the global exploration or local optimization part. The proportionality factor r is used to adjust the relative sizes of the two parts.

[0167] The present invention uses the Euclidean distance as an index to measure population diversity. The Euclidean distance between chromosomes is defined as the square root of the sum of the squares of the differences of the corresponding gene parameters. In the present invention, the individual with the highest fitness in the population is selected as the benchmark for evaluating diversity, and the Euclidean distance between each individual and the optimal individual is calculated.

[0168] The standard population diversity (SPD) describes the degree of difference in the population. The standard population diversity is denoted as D, which is the average value of the D i values of P0 individuals in the population, as shown in Equation (26). D i is denoted as the contribution of individual i with chromosome C i to the standard population diversity. Specifically, D i is calculated through the Euclidean distance between the chromosome C best with the highest fitness and the individual chromosome C i , as shown in Equations (26) and (27), where L is the coding length.

[0169]

[0170] It can be seen from Equation (27) that if there are more individuals similar to the optimal individual C best , the population diversity will decrease.

[0171] Please refer to Figure 5As shown in the figure, it is a schematic diagram of the diversity-based search allocation mechanism according to an embodiment of the present invention. According to the diversity of the population, the present invention divides the parental population into global exploration and local optimization, and evolves independently. The ratio of global exploration to local optimization is adjusted by a ratio factor r, as shown in Equation (28). The first preset parameter h and the second preset parameter l limit the value range of r within [l, h]. The number of individuals in the global exploration part S1 and the local optimization part S2 are given by Equations (28) and (29), where P0 represents the total number of individuals in the population.

[0172]

[0173] S1 = P0r, S2 = P0(1 - r) (29)

[0174] To ensure that the algorithm has both global search ability and local optimization ability, the ratio factor r is adaptively adjusted within the range of [0.5, 0.9]. This can ensure that at least half of the individuals perform local search, and no less than 10% of the individuals execute global exploration. When the population diversity decreases, the value of r is reduced to increase the proportion of global exploration, prompting the population to explore unknown regions. Conversely, if the diversity increases and the distribution range of the population is larger, the algorithm will focus more on the local area around the individual with the highest fitness in the current population, and perform local optimization search within this area.

[0175] Please refer to Figure 6 As shown in the figure, it is a schematic diagram of traffic flow-guided mutation according to an embodiment of the present invention. For local optimization, the individual mutation rate of the intelligent connected vehicle is determined based on the traffic flow;

[0176] Regarding traffic flow-guided mutation, in the diversity-based search allocation mechanism (DSAM), the local optimization part focuses on improving the quality of the solution through traffic flow-guided mutation. The probability P of the mutation operation mut is comprehensively affected by the individual fitness, the contribution to the population diversity, and the real-time traffic flow.

[0177] The specific calculation method of the mutation probability is shown in Equation (30), where f par represents the parental individual, f max represents the maximum fitness of the population, and f min represents the minimum fitness of the population.

[0178]

[0179] Design the individual mutation rate from the perspective of population diversity as shown in Equation (31). In the sub-population of the local optimization part, the main purpose is to find better solutions around the current optimal individual. When the standard population diversity D is equal to 0, all individuals have the same chromosome, When it reaches the maximum value, the algorithm will encourage the population to mutate and generate more new individuals.

[0180]

[0181] To enable the cluster to dynamically adjust the individual mutation rate based on the real-time traffic flow T and ensure the robustness of the algorithm under different traffic conditions, the calculation is as shown in Equation (32). The schematic diagram of traffic flow-guided mutation is shown in (32).

[0182]

[0183] where the real-time traffic flow T refers to the number of vehicles passing through a specific section within a unit time. Define the maximum traffic flow T max as the maximum traffic volume that can be achieved on the passing section. Introduce an adjustment coefficient μ to regulate the impact of traffic flow on the mutation rate. As T increases, the mutation rate grows linearly, which helps to introduce genetic diversity and avoid premature convergence. By dynamically adjusting the mutation rate according to the real-time traffic flow, the genetic algorithm can adapt to different traffic conditions: increasing the mutation rate when the traffic volume is large to increase population diversity, and decreasing the mutation rate when the traffic volume is small to maintain the stability of excellent genes.

[0184] The probability P of the mutation operation mut is calculated as shown in Equation (33). Determine the baseline mutation rate p0 under normal traffic conditions. This is the standard rate at which genetic mutations occur in the algorithm, ensuring a balance between exploration and exploitation in the search space. σ is the weight parameter of the individual fitness and the mutation rate of population diversity, which is the weight parameter of the traffic flow mutation rate.

[0185]

[0186] According to the above formula, it can be known that when the individual fitness is low and the population diversity is insufficient, the mutation rate will increase significantly to promote the algorithm to explore new optimal solutions. Considering the traffic flow as an independent factor enhances the adaptability and flexibility of the model to different traffic conditions. This dynamic adjustment mechanism enables the algorithm to adjust the search strategy according to real-time traffic data and achieve more effective optimization.

[0187] Regarding chromosome crossover and mutation,

[0188] During the population evolution process, set the number of chromosomes for the intelligent connected vehicle crossover operation to N cros , and perform the crossover operation through the single-point chromosome crossover method. Through the crossover operation, N crosIndividual chromosomes proceed to the next generation. During the iterative evolution of the population, chromosomal mutations enable the algorithm to approach the global optimal solution. The number of chromosomes participating in the mutation operation is N p -N cros , and the probability of mutation occurring is P mut . During the mutation process, the mutation operation mainly includes three categories:

[0189] ① The change in the front wheel steering angle of the intelligent connected vehicle;

[0190] ② The change in the acceleration of the intelligent connected vehicle;

[0191] ③ The change in the alliance to which the intelligent connected vehicle belongs.

[0192] Regarding the periodic global optimization of the population,

[0193] In the adaptive co-evolution algorithm, N p initial populations are assigned P0 individuals, and the optimization step size is set to T choose . Each intelligent connected vehicle independently completes the evolutionary process of the population. During the evolutionary process, every T choose times of evolution, information is exchanged with the current populations of other intelligent connected vehicles, and thus N cho excellent chromosomes are output. The setting of the optimization step size can reduce the amount of real-time information communication, thereby effectively reducing the communication load of the V2X network and accelerating the algorithm convergence speed.

[0194] Specifically, in step S6, step S5 is repeated until the termination condition is met, and the optimal trajectory solution is obtained. The termination condition is that the number of cycles of global optimization reaches the set maximum value or the total fitness of the intelligent connected vehicles shows no improvement in a predetermined number of sub-global optimizations.

[0195] It can be understood that when the number of cycles of global optimization during the above chromosomal iteration process reaches the set maximum value N global or the total fitness of the intelligent connected vehicles shows no improvement within N window times of global optimization, the loop can be terminated. The optimal decision is output.

[0196] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

[0197] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention; for those skilled in the art, the present invention may have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A macro-micro integrated coalition decision-making method based on intelligent connected vehicle clusters, characterized in that Including: Step S1, constructing an initial path for the intelligent connected vehicle cluster; Step S2, constructing a coalition set for the intelligent connected vehicle cluster, and constructing a cost function for each individual intelligent connected vehicle in the coalition set, including a driving safety cost function, a riding comfort cost function, a driving efficiency cost function, and an energy consumption cost function; Step S3, constructing a coalition game model for the coalition set, and setting the optimization goal as minimizing the cost function; Step S4, determining the constraint conditions for the game decision-making, including kinematic constraints and comfort constraints; Step S5, during the solution process of the coalition model, each intelligent connected vehicle is determined as an independent population for evolution. Every fixed number of evolution times, each of the intelligent connected vehicles exchanges information through V2X communication technology to achieve global optimization; Step S6, repeating Step S5 until the termination condition is met to obtain the optimal trajectory solution.

2. The macro-micro integrated coalition decision-making method for intelligent connected vehicle clusters according to claim 1, characterized in that In the said Step S1, the process of constructing the initial path of the intelligent connected vehicle cluster includes pre-constructing road network data, including dividing nodes and constructing a directed graph for the road network; each intelligent connected vehicle loads the road network data to determine the starting point and the ending point and independently runs the A* algorithm to generate an initial path.

3. The macro-micro integrated coalition decision-making method based on intelligent connected vehicle clusters according to claim 1, wherein In the said Step S2, the driving safety cost function is the sum of the longitudinal safety cost and the lateral safety cost; the riding comfort cost function is determined based on the lateral and longitudinal jerks of the intelligent connected vehicle; the driving efficiency cost function is determined based on the longitudinal speed; the energy consumption cost function is determined based on the air density, the vehicle frontal area, the air drag coefficient, the rolling resistance coefficient, the vehicle mass, the gravitational acceleration, the vehicle comprehensive acceleration, the fuel consumption rate, and the engine efficiency; wherein, the longitudinal safety cost is related to the longitudinal gap and the relative speed with respect to the vehicle in front, and the lateral safety cost is related to the relative distance and the relative speed between the intelligent connected vehicle and the adjacent vehicle.

4. The macro-micro integrated coalition decision-making method based on intelligent connected vehicle clusters according to claim 1, wherein In Step S3, the process of constructing the coalition game model for the coalition set includes performing coalition initialization, dividing the coalition based on the direction labels of the vehicles, including a straight-ahead coalition, a left-turn coalition, and a right-turn coalition; determining the preference order of the coalition game process; determining the action conditions for coalition merger, coalition split, and coalition exchange.

5. The macro-micro integrated coalition decision-making method for intelligent connected vehicle clusters according to claim 4, wherein It also includes determining whether the coalition is in a stable state, where if all intelligent connected vehicles adopt the preference order and each of the intelligent connected vehicles cannot reduce its own cost by joining or exiting the coalition, then the coalition is in a stable state.

6. The macro-micro integrated coalition decision-making method based on intelligent networked vehicle clusters according to claim 1, wherein In the said Step S4, the kinematic constraints are determined based on the acceleration value, the speed value, the front wheel steering angle, the yaw angle, the maximum acceleration, the minimum steering angle, the maximum steering angle, the minimum yaw angle, and the maximum yaw angle of the intelligent connected vehicle; the comfort constraints are constructed based on the front wheel steering angle, the front wheel steering angle change rate, the acceleration increment, and the jerk.

7. The macro-micro integrated coalition decision-making method based on intelligent connected vehicle clusters according to claim 1, characterized in that Each intelligent connected vehicle is determined as an independent population for evolution, where performing gene coding on the intelligent connected vehicle to obtain a chromosome for the intelligent connected vehicle; Assign a population to each intelligent connected vehicle, evaluate the fitness of the chromosomes that meet the constraint conditions to determine the evaluation value for each population; Adjust the proportion of global exploration and local optimization of individuals in the population during the evolution process according to the evaluation value; Among them, the evaluation value is determined based on the Euclidean distance between the chromosomes corresponding to the intelligent connected vehicles.

8. The macro-micro integrated coalition decision-making method for intelligent connected vehicle clusters according to claim 7, wherein, For local optimization, determine the individual mutation rate of the intelligent connected vehicle based on the traffic flow.

9. The macro-micro integrated coalition decision-making method based on intelligent connected vehicle clusters according to claim 1, characterized in that During the population evolution process, determine the number of chromosomes for crossover operation of the intelligent connected vehicle, and perform crossover operation through the single-point crossover method of chromosomes.

10. The macro-micro integrated coalition decision-making method based on intelligent connected vehicle clusters according to claim 1, characterized in that In the step S5, the termination condition is that the number of cycles of global optimization reaches the set maximum value or the total fitness of the intelligent connected vehicle has no improvement under a predetermined number of sub-global optimizations.

Images