A collaborative optimization control method for multi-lane traffic oscillation in low-penetration environment

Abstract

The present invention discloses a multi-lane traffic shock collaborative optimization control method in a low-penetration environment, including: S1, vehicle information collection; S2, traffic shock wave tracking and identification; S3, scope delimitation of control strategy execution; S4, shock prediction of control strategy execution; S5, strategy optimization of control strategy execution; S6, vehicle group coordination of control strategy execution; S7, shock mitigation of control strategy execution; S8, after the wave-absorbing vehicle group passes through the shock wave occurrence section, the free flow speed is restored for free driving, and the control strategy returns to the first step to continuously monitor the emergence of new shock waves. The present invention completes the longitudinal merger on a multi-lane highway through the optimally selected networked automatic driving vehicle combination to form a wave-absorbing vehicle group, and guides the upstream vehicles to travel with the wave-absorbing speed control that can effectively alleviate the propagation of shock waves, alleviates the propagation of traffic shock waves in each lane from downstream to upstream, and solves the problems of frequent congestion and vehicle collisions on multi-lane highways.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation technology, and specifically relates to a multi-lane traffic shock collaborative optimization control method in a low-permeability environment. Background Art

[0002] Traffic shock waves on highways refer to the start-stop waves of alternating vehicle deceleration and acceleration in congested traffic conditions. They usually propagate upstream continuously, even for tens of kilometers. The propagation of traffic shock waves leads to a higher probability of collision, traffic flow interference and reduced traffic capacity. Once a traffic shock wave is formed and propagates backward, it cannot dissipate automatically in a local traffic environment with high traffic volume. Instead, it will continue to propagate and spread upstream, eventually leading to local traffic congestion and paralysis. With the in-depth study of the traffic shock wave phenomenon, its transmission characteristics have gradually become clear. Therefore, how to reduce the amplitude of the shock wave to slow down its propagation upstream has become a key research issue.

[0003] In recent years, connected autonomous vehicles (CAVs) have shown their core advantages and important research value in improving traffic efficiency, promoting traffic safety, and achieving energy conservation and emission reduction due to their excellent controllability, high operating efficiency, and highly intelligent characteristics. In the long run, the popularization of connected autonomous vehicles will go through a continuous development and evolution phase, which will inevitably involve mixed traffic flows of manually driven vehicles and autonomous vehicles of different levels. In order to give full play to the advantages of connected autonomous vehicles in energy conservation and emission reduction, efficient execution, and networked coordination, and effectively use the driving characteristics of connected autonomous vehicles to solve current traffic problems, the academic community has gradually attached importance to research contents such as model construction based on connected autonomous vehicles, performance analysis of mixed traffic flows, optimization of connected autonomous vehicles, and traffic management methods of linkage control.

[0004] With the deepening of research on connected autonomous vehicles, new solutions have been found to alleviate highway traffic congestion and traffic shock waves. Adaptive cruise control (ACC) is used to maintain a constant distance from the vehicle in front to alleviate traffic shock waves. Cooperative adaptive cruise control (CACC) integrates vehicle-to-vehicle (V2V) communication into adaptive cruise control to provide more information about surrounding vehicles, thereby more effectively controlling the headway. Advanced driver assistance systems (ADAS) recommend reducing speed to redistribute traffic and thus prevent shock waves from propagating. However, due to various overtaking or lane-changing behaviors of normal vehicles, in the real world with low CAV penetration, it is difficult to maintain the larger distance recommended by the system between the controlled vehicle and the vehicle in front of it, as human-driven vehicles tend to make lane changes to fill any possible gaps in the traffic flow. To solve the above problems, the framework of Jam Absorption Driving (JAD) was proposed and successfully applied to single-lane scenarios. The basic idea is to select a connected autonomous driving vehicle upstream of the shock wave to slow down in advance when the shock wave is triggered to avoid more upstream vehicles being affected by the downstream shock wave, thereby achieving the purpose of blocking the propagation of the shock wave. The key to this strategy lies in the selection of the strategy implementation vehicle and the determination of the absorption speed after deceleration. However, the strategy implementation vehicle and the absorption speed in the current congestion absorption strategy are determined through a limited number of trial and error experiments, and there is a lack of optimal selection scheme. At the same time, in multi-lane scenarios, since the gaps on the lanes are usually filled by other vehicles by changing lanes and overtaking the slow strategy implementation vehicle, the congestion absorption strategy in the single-lane scenario is not effective on multi-lane roads. Therefore, in multi-lane intelligent connected scenarios with low penetration rates of connected autonomous driving vehicles, an effective collaborative optimization control strategy is still needed to alleviate the propagation and impact of traffic shock waves on each lane, improve highway traffic safety performance, and avoid congestion. Summary of the invention

[0005] Technical problem solved: The present invention provides a method for collaborative optimization control of multi-lane traffic shock waves in a low-penetration environment, which can solve the technical problem that traffic shock waves are difficult to optimize and control in a mixed-flow traffic environment with low penetration of connected autonomous driving vehicles, effectively alleviate the propagation of traffic shock waves on multi-lane roads, improve traffic safety, and reduce traffic interference.

[0006] Technical solution:

[0007] A collaborative optimization control method for multi-lane traffic oscillation in a low-permeability environment, the collaborative optimization control method comprising the following steps:

[0008] S1. For multi-lane highway scenarios where an intelligent network environment is deployed, the penetration rate ratio within the road section is returned to the intelligent network system in real time, and a real-time multi-dimensional information matrix of vehicles is constructed. The real-time multi-dimensional information matrix is ​​monitored and corrected in real time using a radar-visual fusion correction algorithm.

[0009] S2. Integrate dynamic threshold monitoring technology to build an improved multi-scale wave packet cascade model. When the permeability ratio in a road section is lower than the preset permeability ratio, the real-time multi-dimensional information of vehicles obtained in a multi-lane highway scenario under a low-permeability intelligent network environment is input into the multi-scale wave packet cascade model to track the wave packet energy at multiple scales between vehicles and dynamically monitor the vehicle speed to determine the starting time and starting position of the traffic shock wave.

[0010] S3. In a low-penetration intelligent network environment, when a traffic shock wave occurs, a multi-dimensional DBSCAN vehicle group information clustering algorithm is used to construct a vehicle group for executing a strategy to be selected in a dynamically delineated section upstream from the start of the shock wave, and a set of strategies to be selected is constructed based on the speed gradient;

[0011] S4. Construct a Bayesian harmonic moving bottleneck correction shock prediction model to predict the propagation of shock waves during the strategy execution process, and output the prediction results of the shock wave termination time and shock wave termination position;

[0012] S5. Based on fuzzy theory, an improved multi-index comprehensive optimization evaluation model is constructed. Based on the fuzzy theory weight value, a fuzzy comprehensive evaluation function that takes both efficiency and safety into consideration is calculated. The optimal strategy is selected based on the maximization of the improved fuzzy comprehensive evaluation function as the optimization condition, and the optimal strategy execution vehicle group, the optimal strategy execution merging speed, and the corresponding optimal strategy execution wave absorption speed are obtained.

[0013] S6: Control the optimal strategy execution vehicle group to change lanes to the target lane nearby, and decelerate from downstream to upstream to the optimal strategy execution merging speed to complete the merging in sequence, and arrive at the same longitudinal position, forming a wave-absorbing vehicle group that travels side by side and covers all lanes, and starts the coordinated control of the rear traffic at the optimal strategy execution wave-absorbing speed;

[0014] S7. After the coordinated control is turned on, a dynamic nonlinear PID vehicle group control strategy based on the BP neural network is constructed based on the optimal strategy execution wave absorption speed to calculate the dynamic control speed of the wave absorption vehicle group, and the forward end point of the control vehicle group is completely coincident with the time and space point where the shock wave ends, so as to avoid the shock of the vehicle flow and the road ahead;

[0015] S8. After the wave-absorbing vehicle group passes the road section where the shock wave occurs, it resumes the free flow speed and travels freely. The control strategy returns to the first step to continuously monitor the emergence of new shock waves.

[0016] Furthermore, the specific method in step S1 is as follows:

[0017] For multi-lane highway scenarios where an intelligent connected environment is deployed, the penetration rate ratio of the road section is returned to the intelligent connected system in real time. At the same time, a real-time multi-dimensional information matrix of the vehicle is constructed, and a radar-visual fusion correction algorithm is proposed to monitor and correct the real-time information. The specific steps are as follows:

[0018] S11. Update the penetration rate of connected autonomous vehicles (CAV) in real time on the n-lane highway where the intelligent connected environment is deployed, and construct a real-time multi-dimensional information matrix of vehicles as follows:

[0019] M i = {lo i ,la i ,v i ,a i ,h i}

[0020] Among them, lo i is the longitude coordinate of the i-th vehicle, la i is the latitude coordinate of the i-th vehicle, v i is the real-time speed of the i-th vehicle, a i is the real-time acceleration of the i-th vehicle, h i is the headway time of the i-th vehicle.

[0021] S12. A radar-visual fusion correction algorithm is proposed to monitor and correct real-time information. The formula of the radar-visual fusion correction algorithm is as follows:

[0022]

[0023] Among them, M i (t) is the modified multidimensional information matrix of the i-th vehicle at time t, CAV(M i (t)) is the precise multi-dimensional information matrix data of the i-th intelligent connected vehicle at time t, L(CAV i ) is the multi-dimensional information matrix data collected by the radar equipment at time t for the i-th intelligent connected vehicle, VS(CAV i ) is the multi-dimensional information matrix data collected by the video surveillance equipment at time t for the i-th intelligent connected vehicle, is the initial multi-dimensional information matrix data collected by the radar equipment at time t for the i-th vehicle, is the initial multi-dimensional information matrix data collected by the video surveillance equipment at the i-th vehicle at time t, ε is the accuracy coefficient of the radar equipment, ζ is the accuracy coefficient of the video surveillance equipment, and n is the total number of intelligent connected vehicles in the detection section at the current moment.

[0024] Furthermore, the specific method in step S2 is as follows:

[0025] The real-time multi-dimensional information of vehicles obtained in the n-lane highway scenario under a low-penetration intelligent network environment is input into a multi-scale wave packet cascade model improved by fusion dynamic threshold monitoring technology to achieve multi-scale wave packet energy tracking and dynamic vehicle speed monitoring between vehicles, thereby determining the starting time and starting position of the traffic shock wave.

[0026] The multi-scale wave packet cascade model formula is as follows:

[0027]

[0028] In the formula, E ms The cascade wave packet energy values ​​δ, ε, and ζ respectively represent the cascade coefficients of wave packets of different scales after multi-scale wave packet cascade analysis. v (a,b),WPT a (a,b),WPT h (α, β) represent the continuous signals v of different scales in the wave packet dimension respectively. i (t), a i (t), h i (t), max(a) represents the maximum scaling factor.

[0029] After calculating the multi-scale wave packet energy value at each time through the above model, the energy peak time T maxE .

[0030] Integrate dynamic threshold monitoring technology algorithm to determine the energy peak time T maxE Whether the corresponding oscillating head vehicle speed V1 is below the dynamic threshold, the dynamic threshold calculation formula is as follows:

[0031]

[0032] is the dynamic speed threshold, is the average speed of traffic, is the average acceleration of the traffic flow, represents the average headway of traffic flow, and k is a dynamic adjustment factor used to control the influence of average acceleration and average headway on congestion judgment.

[0033] The starting time and starting position of the shock wave under the integration of dynamic threshold monitoring technology are calculated as follows:

[0034]

[0035] Furthermore, the specific method in step S3 is as follows:

[0036] The strategy execution range is defined as L upstream from the starting position of the shock wave, and a multi-dimensional DBSCAN vehicle group information clustering algorithm is proposed to select n vehicles as a combination to form a candidate strategy execution vehicle group K from the limited m (m>n) CAVs within this range. The specific steps are as follows:

[0037] S31, according to the multi-dimensional information matrix M of different CAV i Calculate the weighted vehicle information distance D ij , the calculation formula is as follows:

[0038]

[0039] Among them, r a 、r b is the information distance index weight, A and B are the information distance weight matrices, α i , β i is the dimension weight;

[0040] S32, introduce D as the weighted vehicle information distance threshold, when D ij When ≤D, it is determined that the CAVs satisfy the weighted vehicle information distance proximity principle, and the CAVs that satisfy the weighted vehicle information distance proximity principle are combined and constructed by the multi-dimensional DBSCAN vehicle group information clustering algorithm to obtain the candidate strategy execution vehicle group K. The multi-dimensional DBSCAN vehicle group information clustering algorithm includes the following steps:

[0041] S321, calculate the optimal parameters of the multi-dimensional DBSCAN vehicle group information clustering algorithm, which are the vehicle group information sample area radius ε, the vehicle group information sample size Pts and the minimum vehicle group information sample size Min corresponding to the specific time. Pts The mutation value is calculated as follows:

[0042]

[0043] Min Pts =point(Dist ij ≤D ij )

[0044] Where D(k) is the kernel density function in k dimension, w(k) is the kernel density estimation function in k dimension, σ is the density parameter, λ and μ are the dimension parameters, δ is the sample domain radius parameter; ρ i , j are the densities of data objects i and j, Dist ijis the Euclidean distance between target points i and j.

[0045] S322, perform one-dimensional DBSCAN clustering in the position feature dimension of the vehicle group information matrix to obtain an initial clustering cluster. The calculation process is as follows:

[0046] SC1[S(t,n,lo k ,la k ),ε,Pts,Min Pts ]=K1(t,n,lo pk ,la pk )

[0047] M pi =[X ik ,Y ik ,V ik ,a ik ],i=1,2,...,n

[0048] Where SC1[S(t,n,lo k ,la k ),ε,Pts,Min Pts ] represents the DBSCAN clustering process of the vehicle group information matrix in the latitude and longitude dimensions, S(t,n,lo k ,la k ) represents the set of vehicles in the latitude and longitude dimensions under the kth time window, K1(t,n,lo pk ,la pk ) represents the initial clustering cluster of the pth cluster vehicle group set in the next dimension of the kth time window, M pi Represents the vehicle information matrix of the i-th CAV in each cluster.

[0049] S323, perform two-dimensional DBSCAN clustering under the motion feature dimension of the vehicle group information matrix to obtain secondary target clustering clusters. The calculation process is as follows:

[0050] SC2[S p (t,n,,v i ,a i ),ε,Pts,Min Pts ]=K2(t,n,lo qk ,la qk )

[0051] M qj =[X jk ,Y jk ,V jk ,a jk ],j=1,2,...,n

[0052] Among them SC2[S p (t,n,,vi ,a i ),ε,Pts,Min Pts ] represents the DBSCAN clustering process of the vehicle group information matrix under the motion feature dimension, S p (t,n,,v i ,a i ) represents the vehicle group set in the latitude and longitude dimension under the kth time window, M qj Represents the vehicle information matrix of the j-th CAV in each cluster.

[0053] S33. For different candidate strategy execution vehicle groups, refer to the principle of setting the speed limit value with speed gradient in the variable speed limit design, set the candidate strategy execution merging speed V1 with a gradient of 1km / h from the free flow speed u downward. This speed is used to control the strategy execution head vehicle to drive slowly to reach the same longitudinal position as other strategy execution vehicles to realize the strategy execution vehicle combination and form a wave absorbing vehicle group. The candidate strategy execution vehicle group K and the candidate strategy execution merging speed V1 together constitute the candidate strategy set {(K i ,V 1j )}.

[0054] Furthermore, the specific method in step S4 is as follows:

[0055] The Bayesian harmonic moving bottleneck corrected shock wave prediction model is used to accurately predict the shock wave propagation during the strategy execution process, and the predicted results of the shock wave termination time and shock wave termination position are output.

[0056] S41. Consider the impact of some vehicles overtaking on the propagation of shock waves during the merging process of n strategy execution vehicles, and predict the propagation of shock waves during the strategy execution process. start , the longitudinal position S of the upstream vehicle of the shock wave at this moment start The selected strategy execution merge speed V1 is used as the mobile bottleneck speed input model to predict the position and speed of the upstream vehicle at each time after the start of the shock wave. The calculation formula of the shock wave prediction model is as follows:

[0057]

[0058] in represents the predicted longitudinal position of the vehicle in free flow state, represents the predicted longitudinal position of the vehicle in the congested state, v max Indicates the maximum speed, ξ follows a normal distribution with mean E and variance V2, and Where β is the inverse of the relaxation time, σ 2 is the coefficient of dispersion.

[0059] S42. The prediction of shock waves also needs to consider the impact of some vehicles that overtake the strategy executing vehicle during the strategy execution process on the propagation of shock waves. The strategy executing vehicle during the deceleration and merging process is regarded as a mobile bottleneck. The Bayesian harmonic mobile bottleneck theory is used to calculate the number of overtaking vehicles to make the shock wave prediction more accurate. The Bayesian harmonic mobile bottleneck correction formula is as follows:

[0060]

[0061] where q r represents the traffic volume beyond the mobile bottleneck, q represents the traffic flow rate, and u c represents the critical speed under maximum traffic capacity, u f represents the free flow speed, u represents the spatial average speed, q c represents the maximum road capacity, u b Represents the speed of the mobile bottleneck. k is the prediction weight after Bayesian reconciliation.

[0062] The above Bayesian harmonic mobile bottleneck correction formula can be used to derive the number of vehicles N that surpass the strategy execution vehicle during the strategy execution vehicle group merging process, so that the spatiotemporal information of N vehicles can be input into the prediction model to further correct the original shock wave propagation prediction. The calculation formula for the number of vehicles N that surpass the strategy execution vehicle is as follows:

[0063] N=t1q r1 +t2q r2 +…+t n-1 q r(n-1)

[0064]

[0065] where t n-1 represents the time it takes for the first n-1 strategy execution vehicles to merge with the nth strategy execution vehicle, q r(n-1) It indicates the number of vehicles that are overtaken during the process of merging with the nth strategy execution vehicle after the first n-1 strategy execution vehicles merge.

[0066] When the speed of the last overtaking vehicle before the strategy execution vehicle group merges returns to the free flow speed u, the shock wave is determined to be terminated and the shock wave termination time T is output. end and shock wave end position E n .

[0067] S43, according to the prediction result calculation strategy, the wave absorbing speed V2 is executed, and the speed is used to control the driving of the wave absorbing vehicle group to guide the upstream vehicle to maintain a larger gap with the traffic shock wave. The calculation formula is as follows:

[0068]

[0069] in It represents the longitudinal position of the leading vehicle in strategy execution at the beginning of the shock wave, E n Indicates the longitudinal position of the last overtaking vehicle at the end of the shock wave, L n-1 Indicates the longitudinal distance between the strategy execution lead vehicle and the nth strategy execution vehicle.

[0070] Furthermore, the specific method in step S5 is as follows:

[0071] On the basis of oscillation prediction, a multi-index comprehensive optimization evaluation model based on fuzzy theory is proposed. The oscillation flow rate and cumulative collision time of multiple candidate vehicle groups are calculated, and different fuzzy weight values ​​are assigned to calculate the fuzzy comprehensive evaluation function. The fuzzy comprehensive evaluation function is maximized as the optimization condition to achieve the optimal selection of candidate strategies, and the best strategy execution vehicle group, the best strategy execution merging speed and the corresponding best strategy execution wave absorption speed are obtained. The specific steps are as follows:

[0072] S51. Calculate the traffic flow rate Q at the start section of the shock wave within a specified time period T. The specified time period is generally a certain time range after the shock wave propagation period. The calculation formula is as follows:

[0073]

[0074] in represents the time period T defined under the corresponding candidate strategy, Denotes the corresponding candidate strategy (K i ,V 1j ) The traffic volume at the starting section of the lower shock wave within the specified time period T.

[0075] S52, calculating the cumulative collision time TIT of all vehicles in the road section within the designated time period T, the calculation formula is as follows:

[0076]

[0077] The collision time TTC i (k) represents the time required for a collision when the vehicle i moves faster than the preceding vehicle i-1, where k is the time step and l is the length of the vehicle i-1. The cumulative collision time TIT represents the time below the threshold TTC. * The TTC accumulated over time.

[0078] S53, convert the two evaluation indicators into fuzzy sets to take into account the fuzziness and uncertainty of the evaluation indicators. Assign fuzzy weight values ​​Tra(i) to the two evaluation indicators (oscillation flow rate) to reflect the relative importance in the process of optimizing the candidate strategy. Since if the strategy execution wave absorption speed V2 is less than the congestion speed v0, a new shock wave may be triggered upstream of the wave absorption vehicle group, so V2>v0 is used as the constraint condition for the optimal strategy execution vehicle group and the strategy execution speed. Under this constraint, the fuzzy sets of the two evaluation indicators are synthesized using fuzzy addition to obtain the fuzzy comprehensive evaluation function as shown below:

[0079]

[0080] Tra(Q) and Tra(T) are the fuzzy weight values ​​assigned to the two evaluation indicators of oscillation flow rate and cumulative collision time, respectively. are the fuzzy balance coefficients of the two evaluation indicators, oscillation flow rate and cumulative collision time, λC={λC1,λC1,…,λC K} is the coincidence characteristic root of the two evaluation indicators of oscillation flow rate and cumulative collision time, and Nm represents the multi-dimensional information amount of the vehicle of the mth candidate strategy in the clustering result.

[0081] According to the comprehensive evaluation results of each candidate vehicle group, fuzzy optimization is performed to select the candidate vehicle group with the highest comprehensive evaluation as the optimal strategy, including the best strategy execution vehicle group K m , optimal strategy execution merge speed V 1m And the corresponding optimal strategy execution absorption speed V 2m .

[0082] Furthermore, the specific method in step S6 is as follows:

[0083] The best strategy execution vehicle group changes lanes and decelerates to the best strategy execution merging speed in turn to reach the same longitudinal position, forming a wave-absorbing vehicle group that travels side by side and covers all lanes to prevent the rear vehicle from overtaking, and then uses the best strategy execution wave-absorbing speed collaborative control to guide the upstream vehicle to maintain a large gap with the shock wave. Specifically, the following steps are included:

[0084] S61, the best strategy execution vehicle group merges longitudinally to form a wave-absorbing vehicle group. When the shock wave is detected, the best strategy execution vehicle group selected in step S2 is composed of n connected autonomous driving vehicles CAV1, CAV2, ..., CAVn (arranged in order from downstream to upstream of the traffic flow) that are adjacent to each other in the longitudinal position. Each strategy execution vehicle first changes lanes to different n lanes nearby, and then merges longitudinally in n-1 stages. First, CAV1 decelerates to the best strategy execution merging speed V 1mControl driving to reach the same longitudinal position as CAV2 to complete the merger; after CAV1 and CAV2 complete the merger, the two vehicles move at the same time in V 1m Control driving and reach the same longitudinal position with CAV3 to complete the merger, and so on, until n strategy execution vehicles complete the merger to form a wave-absorbing vehicle group CAV* covering all lanes.

[0085] S62, the wave absorbing vehicle group executes the wave absorbing speed control to guide the vehicle group with the best strategy. After the wave absorbing vehicle group CAV* is formed, the speeds of the strategy executing vehicles CAV1, CAV2, ..., CAVn are simultaneously reduced from V 1m Adjust to the optimal strategy execution absorption speed V 2m .

[0086] Furthermore, the specific method in step S7 is as follows:

[0087] After the cooperative control is turned on, the optimal strategy is executed at the wave absorption speed V 2m Based on this, a dynamic nonlinear PID vehicle group control strategy based on BP neural network is constructed to calculate the dynamic control speed of the wave-absorbing vehicle group, and the forward end point of the vehicle group is controlled to completely coincide with the time and space point where the shock wave ends, so as to avoid the shock of the vehicle flow and the road ahead. The wave-absorbing speed optimization scheme in the dynamic nonlinear PID vehicle group control strategy based on BP neural network specifically includes the following steps:

[0088] S71. Determine the basic structure of the BP neural network, including the number of input nodes, hidden layer nodes and output nodes. Initialize the weighted coefficient w with the initial value 1 ij(0) and w 2 ij(0). Select the learning rate and inertia coefficient, and set the time step Δt.

[0089] S72, sampling the input and output of the current time step And calculate the control error e(t) of this time step as follows:

[0090]

[0091] S73, calculate the input and output of each neuron in the BP neural network. The output of the BP neural network provides the online adjustable parameter k of the PID controller. p ,k i ,k d

[0092] S74. Calculate the wave absorption velocity in the PID control strategy:

[0093]

[0094] Among them, V2(t) is the dynamically adjusted forward speed at time t, is the speed gain proportional coefficient at time t, V2(t-1) is the optimal strategy execution absorption speed recursively forward one time step, v0(t) is the congestion speed corresponding to the congested traffic flow at time t, δ is the speed balance adjustment parameter, and k p is the proportional gain coefficient, k i is the integral gain coefficient, k d is the differential gain coefficient.

[0095] S75, real-time learning of BP neural network, adjusting weighting coefficient w 1 ij(t) and w 2 ij(t), realizes adaptive adjustment of PID control parameters.

[0096] Beneficial effects:

[0097] First, the multi-lane traffic shock collaborative optimization control method in a low-penetration environment of the present invention completes the longitudinal merger on the multi-lane highway through the optimally selected combination of connected autonomous driving vehicles to form a wave-absorbing vehicle group, and guides the upstream vehicles to travel at a wave-absorbing speed that can effectively alleviate the propagation of shock waves, thereby better alleviating the propagation of traffic shock waves in each lane of the multi-lane highway from downstream to upstream.

[0098] Second, the multi-lane traffic shock collaborative optimization control method in a low penetration environment of the present invention can effectively avoid the reduction in driving efficiency and safety of manually driven vehicles caused by the propagation of traffic shock waves when the penetration rate of networked autonomous driving vehicles is low, and expand the traditional single-vehicle control strategy to multi-lane collaborative control based on an intelligent networked communication environment, effectively solving the problem of frequent congestion and vehicle collisions on multi-lane highways. The present invention provides a safe and efficient driving environment for driving participants in an intelligent networked communication environment, and also provides a better theoretical reference for traffic decision makers when conducting traffic control methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 It is a flow chart of the multi-lane traffic oscillation collaborative optimization control method in a low-permeability environment of the present invention;

[0100] Figure 2 It is a schematic diagram of the control range demarcation result in the present invention;

[0101] Figure 3 It is a schematic diagram of the shock wave prediction model in the present invention;

[0102] Figure 4 It is a schematic diagram of the spatiotemporal trajectory of the vehicle in the scene of the present invention;

[0103] Figure 5 It is a schematic diagram of the control steps of the strategy execution vehicle in the present invention. DETAILED DESCRIPTION

[0104] The following examples will enable those skilled in the art to more fully understand the present invention, but are not intended to limit the present invention in any way.

[0105] The present invention discloses a multi-lane traffic oscillation coordinated optimization control method in a low permeability environment, the specific steps are as follows: Figure 1 shown.

[0106] This embodiment takes a three-lane main line highway section with a length of 5km as an example. The free flow speed of the section is 93km / h. The section is covered by an intelligent network system, and multi-dimensional information of vehicles is collected by means of deployed radar equipment and video surveillance equipment. The penetration rate of networked autonomous driving vehicles on the section fluctuates between 0-20%, and the low penetration detection threshold is set to 15%. The communication conditions between the networked autonomous driving vehicles and the intelligent network system are good.

[0107] Combination Figure 1 The specific steps of the embodiment are as follows:

[0108] Step 1: Vehicle information collection. In this embodiment, in a 5km long highway main line three-lane road section, the intelligent network system updates the penetration rate of connected autonomous vehicles (CAV) in the road section in real time. When the penetration rate is lower than 15%, the real-time multi-dimensional information matrix of the vehicle is constructed as follows:

[0109] M i = {lo i ,la i ,v i ,a i ,h i}

[0110] Among them, lo i is the longitude coordinate of the i-th vehicle, la i is the latitude coordinate of the i-th vehicle, v i is the real-time speed of the i-th vehicle, a i is the real-time acceleration of the i-th vehicle, h i is the headway time of the i-th vehicle.

[0111] Next, a radar-visual fusion correction algorithm is proposed to monitor and correct real-time information. The formula of the radar-visual fusion correction algorithm is as follows:

[0112]

[0113] Among them, M i (t) is the modified multidimensional information matrix of the i-th vehicle at time t, CAV(Mi (t)) is the precise multi-dimensional information matrix data of the i-th intelligent connected vehicle at time t, L(CAV i ) is the multi-dimensional information matrix data collected by the radar equipment at time t for the i-th intelligent connected vehicle, VS(CAV i ) is the multi-dimensional information matrix data collected by the video surveillance equipment at time t for the i-th intelligent connected vehicle, is the initial multi-dimensional information matrix data collected by the radar equipment at time t for the i-th vehicle, is the initial multi-dimensional information matrix data collected by the video surveillance equipment at the i-th vehicle at time t, ε is the accuracy coefficient of the radar equipment, ζ is the accuracy coefficient of the video surveillance equipment, and n is the total number of intelligent connected vehicles in the detection section at the current moment.

[0114] Step 2: Traffic shock wave tracking and identification. The real-time multi-dimensional information of vehicles obtained in a multi-lane highway scenario under a low-penetration intelligent network environment is input into a multi-scale wave packet cascade model improved by integrating dynamic threshold monitoring technology to achieve multi-scale wave packet energy tracking and dynamic vehicle speed monitoring between vehicles, thereby determining the starting time and starting position of the traffic shock wave;

[0115] First, the multi-scale wave packet cascade model is used to obtain the cascade wave packet energy value as follows:

[0116]

[0117] In the formula, E ms The cascade wave packet energy values ​​δ, ε, and ζ respectively represent the cascade coefficients of wave packets of different scales after multi-scale wave packet cascade analysis. v (a,b),WPT a (a,b),WPT h (α, β) represent the continuous signals v of different scales in the wave packet dimension respectively. i (t), a i (t), h i (t), max(a) represents the maximum scaling factor.

[0118] After calculating the multi-scale cascade wave packet energy value at each time through the above model, the energy peak time T maxE .

[0119] Next, the dynamic threshold monitoring technology algorithm is integrated to determine the energy peak time T maxE Whether the corresponding oscillating head vehicle speed V1 is below the dynamic threshold, the dynamic threshold calculation formula is as follows:

[0120]

[0121] is the dynamic speed threshold, is the average speed of traffic, is the average acceleration of the traffic flow, represents the average headway of traffic flow, and k is a dynamic adjustment factor used to control the influence of average acceleration and average headway on congestion judgment.

[0122] Then the starting time and starting position of the shock wave under the fusion dynamic threshold monitoring technology are obtained as follows:

[0123]

[0124] Step 3: Define the scope of control strategy execution. In a low penetration environment, when a traffic shock wave is detected, a strategy execution scope of 1 km upstream from the shock wave starting position is defined, and a multi-dimensional DBSCAN vehicle group information clustering algorithm is proposed to select 3 vehicles from the limited m (m>3) CAVs within this range to form a candidate strategy execution vehicle group K.

[0125] First, according to the multidimensional information matrix M of different CAV i Calculate the weighted vehicle information distance D ij , the calculation formula is as follows:

[0126]

[0127] Among them, r a 、r b is the information distance index weight, A and B are the information distance weight matrices, α i , β i is the dimension weight;

[0128] Next, we introduce D as the weighted vehicle information distance threshold. ij When ≤D, it is determined that the CAVs satisfy the weighted vehicle information distance proximity principle, and the CAVs that satisfy the weighted vehicle information distance proximity principle are combined and constructed by the multi-dimensional DBSCAN vehicle group information clustering algorithm to obtain the candidate strategy execution vehicle group K. The multi-dimensional DBSCAN vehicle group information clustering algorithm includes the following steps:

[0129] First, we need to calculate the optimal parameters of the multi-dimensional DBSCAN vehicle group information clustering algorithm, which are the vehicle group information sample area radius ε, vehicle group information sample size Pts and minimum vehicle group information sample size Min corresponding to the specific time. Pts The mutation value is calculated as follows:

[0130]

[0131] Min Pts =point(Distij ≤D ij )

[0132] Where D(k) is the kernel density function in k dimension, w(k) is the kernel density estimation function in k dimension, σ is the density parameter, λ and μ are the dimension parameters, δ is the sample domain radius parameter; ρ i , j are the densities of data objects i and j, Dist ij is the Euclidean distance between target points i and j.

[0133] Then, one-dimensional DBSCAN clustering is performed under the position feature dimension of the vehicle group information matrix to obtain an initial clustering cluster. The calculation process is as follows:

[0134] SC1[S(t,n,lo k ,la k ),ε,Pts,Min Pts ]=K1(t,n,lo pk ,la pk )

[0135] M pi =[X ik ,Y ik ,V ik ,a ik ],i=1,2,...,n

[0136] Where SC1[S(t,n,lo k ,la k ),ε,Pts,Min Pts ] represents the DBSCAN clustering process of the vehicle group information matrix in the latitude and longitude dimensions, S(t,n,lo k ,la k ) represents the set of vehicles in the latitude and longitude dimensions under the kth time window, K1(t,n,lo pk ,la pk ) represents the initial clustering cluster of the pth cluster vehicle group set in the next dimension of the kth time window, M pi Represents the vehicle information matrix of the i-th CAV in each cluster.

[0137] Then, two-dimensional DBSCAN clustering is performed under the motion feature dimension of the vehicle group information matrix to obtain the secondary target clustering cluster. The calculation process is as follows:

[0138] SC2[S p (t,n,,v i ,a i ),ε,Pts,Min Pts ]=K2(t,n,lo qk ,la qk )

[0139] M qj =[X jk ,Y jk ,V jk ,a jk ],j=1,2,...,n

[0140] Among them SC2[S p (t,n,,v i ,a i ),ε,Pts,Min Pts ] represents the DBSCAN clustering process of the vehicle group information matrix under the motion feature dimension, S p (t,n,,v i ,a i ) represents the vehicle group set in the latitude and longitude dimension under the kth time window, M qj Represents the vehicle information matrix of the j-th CAV in each cluster.

[0141] In this embodiment, a strategy execution range of 1 km is defined upstream from the starting position of the shock wave, and multi-dimensional DBSCAN clustering is performed among the limited 5 connected autonomous driving vehicles within this range to obtain 3 vehicles as a combination to form a candidate strategy execution vehicle group K, totaling 3 groups K1, K2, K3, as shown in FIG. Figure 2 shown.

[0142] For different candidate strategy execution vehicle groups, the candidate strategy execution merging speed V1 is set from the free flow speed of 93km / h to a gradient of 1km / h. This speed is used to control the strategy execution head vehicle to drive slowly to reach the same longitudinal position as other strategy execution vehicles to realize the strategy execution vehicle combination and form a wave absorbing vehicle group. The candidate strategy execution vehicle group K and the candidate strategy execution merging speed V1 together constitute the candidate strategy set {(K i ,V 1j )}.

[0143] Step 4: Predict shocks during the execution of the control strategy. Figure 3 As shown in the figure, a Bayesian harmonic moving bottleneck corrected shock wave prediction model is proposed to accurately predict the propagation of shock waves during the strategy execution process, and output the predicted results of the shock wave termination time and shock wave termination position.

[0144] First, the shock wave starting time T start , the longitudinal position S of the upstream vehicle of the shock wave at this moment start The selected strategy execution merge speed V1 is used as the mobile bottleneck speed input model to predict the position and speed of the upstream vehicle at each time after the start of the shock wave. The calculation formula of the shock wave prediction model is as follows:

[0145]

[0146] in represents the predicted longitudinal position of the vehicle in free flow state, represents the predicted longitudinal position of the vehicle in the congested state, v max Indicates the maximum speed, ξ follows a normal distribution with mean E and variance V2, and Where β is the inverse of the relaxation time, σ 2 is the dispersion coefficient. In this embodiment, β=0.03s-1, δ=0.03m / s2, Lveh=5m, and τ=1s are selected.

[0147] At the same time, the prediction of shock waves also needs to consider the impact of some vehicles that overtake the strategy execution vehicle during the strategy execution process on the propagation of shock waves. The strategy execution vehicle in the deceleration and merging process is regarded as a mobile bottleneck. The Bayesian harmonic mobile bottleneck theory is used to calculate the number of overtaking vehicles to make the shock wave prediction more accurate. The Bayesian harmonic mobile bottleneck correction formula is as follows:

[0148]

[0149] where q r represents the traffic volume beyond the mobile bottleneck, q represents the traffic flow rate, and u c represents the critical speed under maximum traffic capacity, u f represents the free flow speed, u represents the spatial average speed, q c represents the maximum road capacity, u b Represents the speed of the mobile bottleneck. k is the prediction weight after Bayesian reconciliation.

[0150] The above Bayesian harmonic mobile bottleneck correction formula can be used to derive the number of vehicles N that surpass the strategy execution vehicle during the strategy execution vehicle group merging process, so that the spatiotemporal information of N vehicles can be input into the prediction model to further correct the original shock wave propagation prediction. The calculation formula for the number of vehicles N that surpass the strategy execution vehicle is as follows:

[0151] N=t1q r1 +t2q r2 +…+t n-1 q r(n-1)

[0152]

[0153] where t n-1 represents the time it takes for the first n-1 strategy execution vehicles to merge with the nth strategy execution vehicle, q r(n-1)It indicates the number of vehicles that are overtaken during the process of merging with the nth strategy execution vehicle after the first n-1 strategy execution vehicles merge.

[0154] When the speed of the last overtaking vehicle before the strategy execution vehicle group merges returns to the free flow speed of 93km / h, the shock wave is determined to be terminated and the shock wave termination time T is output. end and shock wave end position E n , and calculate the candidate strategy (K i ,V 1j ) corresponds to the strategy execution absorption speed V2, V2 solution process refers to Figure 4 As shown in the figure, the simplified shock wave prediction model modified by Bayesian harmonic moving bottleneck theory predicts the vehicle trajectory position-time diagram in the lane where the strategy execution lead vehicle is located. V2 is the slope of the straight line BB' in the figure, which can be obtained by x BB′ / t BB′ calculate.

[0155] The driving trajectories of the two upstream policy execution vehicles are represented by dotted lines. start , the longitudinal distance between the first strategy execution vehicle and the second strategy execution vehicle is L1, the longitudinal distance between the second strategy execution vehicle and the third strategy execution vehicle is L2, the longitudinal distance between the first strategy execution vehicle and the third strategy execution vehicle is L1+L2, and the time when the strategy execution vehicle group completes the merger is T mid , so the duration T1 of the strategy execution vehicle group merging can be calculated as follows:

[0156]

[0157] Further calculation of t BB′ as follows:

[0158]

[0159] x B′ That is, the shock wave termination position E predicted by the shock wave prediction model n , x A That is, the longitudinal position of the strategy execution leader at the start of the shock wave So calculate x BB′ as follows:

[0160] x BB′ =x B′ -x A +T1V 1j ;

[0161] Therefore, the calculation of the wave absorption velocity V2 is as follows:

[0162]

[0163] Step 5, strategy optimization for control strategy execution. Based on the shock prediction, a multi-index comprehensive optimization evaluation model based on fuzzy theory is proposed. The fuzzy comprehensive evaluation function that considers both efficiency and safety is calculated based on the fuzzy theory weight value. The optimization condition is used to maximize the fuzzy comprehensive evaluation function to achieve the optimization of the candidate strategy, and the best strategy execution vehicle group, the best strategy execution combined speed, and the corresponding best strategy execution wave absorption speed are obtained.

[0164] First, calculate the traffic flow rate Q at the starting section of the shock wave within the designated time period T. The designated time period is generally a certain time range after the shock propagation period. The calculation formula is as follows:

[0165]

[0166] in represents the time period T defined under the corresponding candidate strategy, Denotes the corresponding candidate strategy (K i ,V 1j ) The traffic volume at the starting section of the lower shock wave within the specified time period T.

[0167] Then calculate the cumulative collision time TIT of all vehicles in the road section within the designated time period T, and the calculation formula is as follows:

[0168]

[0169] The collision time TTC i (k) represents the time required for a collision when the vehicle i moves faster than the preceding vehicle i-1, where k is the time step and l is the length of the vehicle i-1. The cumulative collision time TIT represents the time below the threshold TTC. * The TTC accumulated over time.

[0170] The next step is to convert the two evaluation indicators into fuzzy sets to take into account the fuzziness and uncertainty of the evaluation indicators. Fuzzy weight values ​​Tra(i) are assigned to the two evaluation indicators (oscillation flow rate) to reflect their relative importance in the process of optimizing the candidate strategy. Since the strategy execution wave absorption speed V2 is less than the congestion speed v0, it may trigger a new shock wave upstream of the wave absorption vehicle group. Therefore, V2>v0 is used as the constraint condition for the optimal strategy execution vehicle group and the strategy execution speed. Under this constraint, the fuzzy sets of the two evaluation indicators are synthesized using fuzzy addition to obtain the fuzzy comprehensive evaluation function as shown below:

[0171]

[0172] Tra(Q) and Tra(T) are the fuzzy weight values ​​assigned to the two evaluation indicators of oscillation flow rate and cumulative collision time, respectively. are the fuzzy balance coefficients of the two evaluation indicators, oscillation flow rate and cumulative collision time, λC={λC1,λC1,…,λC K} is the coincidence characteristic root of the two evaluation indicators of oscillation flow rate and cumulative collision time, and Nm represents the multi-dimensional information amount of the vehicle of the mth candidate strategy in the clustering result.

[0173] Then, according to the comprehensive evaluation results of each candidate vehicle group, fuzzy optimization is performed to select the candidate vehicle group with the highest comprehensive evaluation as the optimal strategy, including the best strategy execution vehicle group K m , optimal strategy execution merge speed V 1m And the corresponding optimal strategy execution absorption speed V 2m .

[0174] Step 6: Control the vehicle group coordination of strategy execution. By controlling the optimal strategy execution vehicle group to change lanes to the target lane when necessary, and decelerate from downstream to upstream to the optimal strategy execution merging speed to complete the sequential merging, reaching the same longitudinal position, forming a wave-absorbing vehicle group that travels side by side and covers all lanes, and starts the coordinated control of the rear vehicle flow at the optimal strategy execution wave-absorbing speed. Figure 5 As shown;

[0175] First, the best strategy execution vehicle group merges longitudinally to form a wave-absorbing vehicle group. When the shock wave is detected, the best strategy execution vehicle group selected through the above steps consists of n connected autonomous driving vehicles CAV1, CAV2, ..., CAVn (arranged in order from downstream to upstream of the traffic flow) that are adjacent to each other in the longitudinal position. Each strategy execution vehicle first changes lanes to different n lanes nearby, and then merges longitudinally in n-1 stages. First, CAV1 decelerates to the best strategy execution merging speed V 1m Control driving to reach the same longitudinal position as CAV2 to complete the merger; after CAV1 and CAV2 complete the merger, the two vehicles move at the same time in V 1m Control driving and reach the same longitudinal position with CAV3 to complete the merger, and so on, until n strategy execution vehicles complete the merger to form a wave-absorbing vehicle group CAV* covering all lanes.

[0176] Next, the wave-absorbing vehicle group executes the wave-absorbing speed control with the best strategy to guide the convoy. After the wave-absorbing vehicle group CAV* is formed, the speeds of the strategy-executing vehicles CAV1, CAV2, ..., CAVn are simultaneously increased from V 1m Adjust to the optimal strategy execution absorption speed V 2m .

[0177] Step 7, control strategy execution shock relief. After the collaborative control is turned on, the dynamic nonlinear PID vehicle group control strategy based on the BP neural network is constructed based on the optimal strategy execution shock absorption speed to calculate the dynamic control speed of the shock absorption vehicle group, and the forward end point of the control vehicle group is completely coincident with the time and space point where the shock wave ends, so as to avoid shocks between the vehicle flow and the road ahead. The shock absorption speed optimization scheme in the dynamic nonlinear PID vehicle group control strategy based on the BP neural network is as follows:

[0178] First, determine the basic structure of the BP neural network, including the number of input nodes, hidden nodes, and output nodes. Initialize the weight coefficient w with the initial value 1 ij(0) and w 2 ij(0). Select the learning rate and inertia coefficient, and set the time step Δt.

[0179] Then sample the input and output of the current time step And calculate the control error e(t) of this time step as follows:

[0180]

[0181] Then the input and output of each neuron in the BP neural network are calculated. The output of the BP neural network provides the online adjustable parameter k of the PID controller. p ,k i ,k d

[0182] Then calculate the wave absorption velocity in the PID control strategy:

[0183]

[0184] Among them, V2(t) is the dynamically adjusted forward speed at time t, is the speed gain proportional coefficient at time t, V2(t-1) is the optimal strategy execution absorption speed recursively forward one time step, v0(t) is the congestion speed corresponding to the congested traffic flow at time t, δ is the speed balance adjustment parameter, and k p is the proportional gain coefficient, k i is the integral gain coefficient, k d is the differential gain coefficient.

[0185] At the same time, the BP neural network is learned in real time and the weighting coefficient w is adjusted. 1 ij(t) and w 2 ij(t), realizes adaptive adjustment of PID control parameters.

[0186] Step 8: After the wave-absorbing vehicle group passes the section where the shock wave occurs, it resumes the free flow speed of 93 km / h for free driving, and the control strategy returns to the first step to continuously monitor the occurrence of new shock waves.

[0187] The invention provides a collaborative optimization control strategy for alleviating multi-lane traffic shock waves in a low-penetration intelligent network environment. The optimally selected networked autonomous driving vehicle combination completes the longitudinal merger on the multi-lane highway to form a wave-absorbing vehicle group, and controls and guides the upstream vehicles to travel at a wave-absorbing speed that can effectively alleviate the propagation of shock waves, thereby effectively alleviating the propagation of traffic shock waves from downstream to upstream in each lane of the multi-lane highway. This strategy can effectively avoid the reduction in driving efficiency and safety of manually driven vehicles caused by the propagation of traffic shock waves when the penetration rate of networked autonomous driving vehicles is low. It expands the traditional single-vehicle control strategy to multi-lane collaborative control based on the intelligent network communication environment, effectively solving the problems of frequent congestion and vehicle collisions on multi-lane highways. The invention provides a safe and efficient driving environment for driving participants in the intelligent network communication environment, and also provides a better theoretical reference for traffic decision makers when conducting traffic control methods.

[0188] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A multi-lane traffic oscillation collaborative optimization control method in a low-permeability environment, characterized in that: The collaborative optimization control method comprises the following steps: S1. For multi-lane highway scenarios where an intelligent network environment is deployed, the penetration rate ratio within the road section is returned to the intelligent network system in real time, and a real-time multi-dimensional information matrix of vehicles is constructed. The real-time multi-dimensional information matrix is ​​monitored and corrected in real time using a radar-visual fusion correction algorithm. S2. Integrate dynamic threshold monitoring technology to build an improved multi-scale wave packet cascade model. When the permeability ratio in a road section is lower than the preset permeability ratio, the real-time multi-dimensional information of vehicles obtained in a multi-lane highway scenario under a low-permeability intelligent network environment is input into the multi-scale wave packet cascade model to track the wave packet energy at multiple scales between vehicles and dynamically monitor the vehicle speed to determine the starting time and starting position of the traffic shock wave. S3. In a low-penetration intelligent network environment, when a traffic shock wave occurs, a multi-dimensional DBSCAN vehicle group information clustering algorithm is used to construct a vehicle group for executing a strategy to be selected in a dynamically delineated section upstream from the start of the shock wave, and a set of strategies to be selected is constructed based on the speed gradient; S4. Construct a Bayesian harmonic moving bottleneck correction shock prediction model to predict the propagation of shock waves during the strategy execution process, and output the prediction results of the shock wave termination time and shock wave termination position; S5. Based on fuzzy theory, an improved multi-index comprehensive optimization evaluation model is constructed. Based on the fuzzy theory weight value, a fuzzy comprehensive evaluation function that takes both efficiency and safety into consideration is calculated. The optimal strategy is selected based on the maximization of the improved fuzzy comprehensive evaluation function as the optimization condition, and the optimal strategy execution vehicle group, the optimal strategy execution merging speed, and the corresponding optimal strategy execution wave absorption speed are obtained. S6: Control the optimal strategy execution vehicle group to change lanes to the target lane nearby, and decelerate from downstream to upstream to the optimal strategy execution merging speed to complete the merging in sequence, and arrive at the same longitudinal position, forming a wave-absorbing vehicle group that travels side by side and covers all lanes, and starts the coordinated control of the rear traffic at the optimal strategy execution wave-absorbing speed; S7. After the coordinated control is turned on, a dynamic nonlinear PID vehicle group control strategy based on the BP neural network is constructed based on the optimal strategy execution wave absorption speed to calculate the dynamic control speed of the wave absorption vehicle group, and the forward end point of the control vehicle group is completely coincident with the time and space point where the shock wave ends, so as to avoid the shock of the vehicle flow and the road ahead; S8, after the wave-absorbing vehicle group passes the road section where the shock wave occurs, it resumes the free flow speed and travels freely, and the control strategy returns to the first step to continuously monitor the emergence of new shock waves; Step S2 further comprises: S21. Calculate the multi-scale wave packet energy value at each moment through the multi-scale wave packet cascade model: In the formula, E ms represents the cascade wave packet energy value obtained through multi-scale wave packet cascade analysis, δ, ε, ζ represent the cascade coefficients of wave packets of different scales, WPT v (α,β), WPT a (α,β), WPT h (α, β) represent the continuous signals v of different scales in the wave packet dimension. i (t), a i (t), h i (t) is the continuous wave packet coefficient, max(a) represents the maximum scaling factor; Locking energy peak time T maxE , integrating dynamic threshold monitoring technology algorithm to determine the energy peak time T maxE Whether the corresponding oscillating head vehicle speed V1 is below the dynamic threshold, the dynamic threshold calculation formula is as follows: In the formula, is the dynamic speed threshold, is the average speed of traffic, is the average acceleration of the traffic flow, represents the average headway of traffic flow, k is a dynamic adjustment factor, which is used to control the influence of average acceleration and average headway on congestion judgment; Integrating dynamic threshold monitoring technology, the starting time and starting position of the shock wave are calculated:

2. The method for collaborative optimization control of multi-lane traffic shock in a low-permeability environment according to claim 1 is characterized in that: Step S1 further comprises: S11. For n-lane highways with intelligent connected environments, the penetration rate of connected autonomous driving vehicles (CAVs) in the road section is updated in real time, and a real-time multi-dimensional information matrix of vehicles is constructed as follows: M i ={it i ,the i ,v i ,to i ,h i } Among them, lo i is the longitude coordinate of the i-th vehicle, la i is the latitude coordinate of the i-th vehicle, v i is the real-time speed of the i-th vehicle, a i is the real-time acceleration of the i-th vehicle, h i is the headway time of the i-th vehicle; S12. Use the radar-visual fusion correction algorithm to monitor and correct the real-time multi-dimensional information matrix: Among them, M i (t) is the modified multidimensional information matrix of the i-th vehicle at time t, CAV(M i (t)) is the precise multi-dimensional information matrix data of the i-th intelligent connected vehicle at time t, L(CAV i ) is the multi-dimensional information matrix data collected by the radar equipment at time t for the i-th intelligent connected vehicle, VS(CAV i ) is the multi-dimensional information matrix data collected by the video surveillance equipment at time t for the i-th intelligent connected vehicle, is the initial multi-dimensional information matrix data collected by the radar equipment at time t for the i-th vehicle, is the initial multi-dimensional information matrix data collected by the video surveillance equipment at the i-th vehicle at time t, ε is the accuracy coefficient of the radar equipment, ζ is the accuracy coefficient of the video surveillance equipment, and n is the total number of intelligent connected vehicles in the detection section at the current moment.

3. The method for collaborative optimization control of multi-lane traffic shock in a low-permeability environment according to claim 1 is characterized in that: In step S3, a strategy execution range of distance L is defined upstream from the starting position of the shock wave, and n vehicles are selected from the limited m CAVs within the strategy execution range using the multi-dimensional DBSCAN vehicle group information clustering algorithm to form a candidate strategy execution vehicle group K, where m>n. The specific steps are as follows: S31, according to the multi-dimensional information matrix M of different CAV i Calculate the weighted vehicle information distance D ij , the calculation formula is as follows: Among them, r a 、r b is the information distance index weight, A and B are the information distance weight matrices, α i , β i is the dimension weight; S32, introduce D as the weighted vehicle information distance threshold, when D ij When ≤D, determine that the CAVs satisfy the weighted vehicle information distance proximity principle, and use the multi-dimensional DBSCAN vehicle group information clustering algorithm to combine and construct the CAVs that satisfy the weighted vehicle information distance proximity principle to obtain the candidate strategy execution vehicle group K; S33. For different candidate strategy execution vehicle groups, the candidate strategy execution merging speed V1 is set from the free flow speed u downward with a gradient of 1 km / h. The candidate strategy execution merging speed is used to control the strategy execution head vehicle to drive slowly to reach the same longitudinal position as other strategy execution vehicles to realize the strategy execution vehicle combination and form a wave-absorbing vehicle group; the candidate strategy execution vehicle group K and the candidate strategy execution merging speed V1 together constitute the candidate strategy set {(K i ,V 1j )}.

4. The method for collaborative optimization control of multi-lane traffic shock in a low-permeability environment according to claim 3 is characterized in that: Step S32 further includes: S321, calculate the optimal parameters of the multi-dimensional DBSCAN vehicle group information clustering algorithm, which are the vehicle group information sample area radius ε, the vehicle group information sample size Pts and the minimum vehicle group information sample size Min corresponding to the specific time. Pts The mutation value is calculated as follows: Min Pt =point(Dist ij ≤D ij ) Where D(k) is the kernel density function in k dimension, w(k) is the kernel density estimation function in k dimension, σ is the density parameter, λ and μ are the dimension parameters, δ is the sample domain radius parameter; ρ i , j are the densities of data objects i and j, Dist ij is the Euclidean distance between target points i and j; S322, perform one-dimensional DBSCAN clustering in the position feature dimension of the vehicle group information matrix to obtain an initial clustering cluster. The calculation process is as follows: SC1[S(t,n,lo k ,old k ),ε,Pts,Min Pts ]=K q (t,n,lo pk ,old pk ) M pi =[X ik ,Y ik ,V ik ,a ik ],i=1,2,...,n In the formula, SC1[S(t,n,lo k ,la k ),ε,Pts,Min Pts ] represents the DBSCAN clustering process of the vehicle group information matrix in the latitude and longitude dimensions, S(t,n,lo k ,la k ) represents the set of vehicles in the latitude and longitude dimensions under the kth time window, K1(t,n,lo pk ,la pk ) represents the initial clustering cluster of the pth cluster vehicle group set in the next dimension of the kth time window, M pi represents the vehicle information matrix of the i-th CAV in each cluster; S323, perform two-dimensional DBSCAN clustering under the motion feature dimension of the vehicle group information matrix to obtain secondary target clustering clusters. The calculation process is as follows: SC2[S p (t,n,,v i ,a i ),ε,Pts,Min Pts ]=K2(t,n,lo qk ,la qk ) M qj =[X jk ,Y jk ,V jk ,a jk ],j=1,2,...,n Among them SC2[S p (t,n,,v i ,a i ),ε,Pts,Min Pts ] represents the DBSCAN clustering process of the vehicle group information matrix under the motion feature dimension, S p (t,n,,v i ,a i ) represents the vehicle group set in the latitude and longitude dimension under the kth time window, M qj Represents the vehicle information matrix of the j-th CAV in each cluster.

5. The method for collaborative optimization control of multi-lane traffic shock in a low-permeability environment according to claim 1 is characterized in that: Step S4 further comprises: S41. Consider the impact of some vehicles overtaking on the propagation of shock waves during the merging process of n strategy execution vehicles, and predict the propagation of shock waves during the strategy execution process; set the shock wave starting time T start , the longitudinal position S of the upstream vehicle of the shock wave corresponding to the starting time of the shock wave start The merge speed V1 of the candidate strategy is used as the mobile bottleneck speed input model to predict the position and speed of the upstream vehicle at each time after the start of the shock wave. The calculation formula of the shock wave prediction model is as follows: In the formula, represents the predicted longitudinal position of the vehicle in free flow state, represents the predicted longitudinal position of the vehicle in the congested state, v max Indicates the maximum speed, ξ follows a normal distribution with mean E and variance V2, and β is the inverse of the relaxation time, σ 2 is the coefficient of dispersion; S42. The strategy executing vehicle in the deceleration and merging process is regarded as a moving bottleneck, and the number of overtaking vehicles is calculated using the Bayesian harmonic moving bottleneck theory, wherein the Bayesian harmonic moving bottleneck correction formula is as follows: where q r represents the traffic volume beyond the mobile bottleneck, q represents the traffic flow rate, and u c represents the critical speed under maximum traffic capacity, u f represents the free flow speed, u represents the spatial average speed, q c represents the maximum road capacity, u b represents the speed of the mobile bottleneck, λ k is the prediction weight after Bayesian reconciliation; The number of vehicles N that surpass the strategy execution vehicle during the strategy execution vehicle group merging process is derived, and the spatiotemporal information of N vehicles is input into the prediction model to further correct the original shock wave propagation prediction. The calculation formula for the number of vehicles N that surpass the strategy execution vehicle is as follows: N=t1q r1 +t2q r2 +…+t n-1 q r(n-1) where t n-1 represents the time it takes for the first n-1 strategy execution vehicles to merge with the nth strategy execution vehicle, q r(n-1) It indicates the number of vehicles that are overtaken by the first n-1 strategy execution vehicles after merging with the nth strategy execution vehicle; When the speed of the last overtaking vehicle before the strategy execution vehicle group merges returns to the free flow speed u, the shock wave is determined to be terminated, and the shock wave termination time T is output. end and shock wave end position E n ; S43: Calculate the strategy execution wave absorption speed V2 according to the prediction result. The strategy execution wave absorption speed is used to control the driving of the wave absorption vehicle group to guide the upstream vehicle to maintain a larger gap with the traffic shock wave. The calculation formula is as follows: in, It represents the longitudinal position of the leading vehicle in strategy execution at the start of the shock wave, E n represents the longitudinal position of the last overtaking vehicle at the end of the shock wave, L n-1 Indicates the longitudinal distance between the strategy execution lead vehicle and the nth strategy execution vehicle.

6. The method for collaborative optimization control of multi-lane traffic shock in a low-permeability environment according to claim 1 is characterized in that: Step S5 further comprises: S51. Calculate the traffic flow rate Q of the shock wave starting section within the designated time period T: in represents the time period T defined under the corresponding candidate strategy, Denotes the corresponding candidate strategy (K i ,V 1j ) The traffic volume at the starting section of the lower shock wave within the specified time period T; S52, calculating the cumulative collision time TIT of all vehicles in the road section within the designated time period T, the calculation formula is as follows: The collision time TTC i (k) represents the time required for a collision when the vehicle i moves faster than the preceding vehicle i-1 and the two vehicles continue to travel at the current speed, where k is the time step and l is the length of vehicle i-1; the cumulative collision time TIT represents the time below the threshold TTC * The TTC accumulated over time; S53, convert the two evaluation indicators of traffic flow rate Q and cumulative collision time TIT into fuzzy sets, assign fuzzy weight values ​​Tra(i) to the two evaluation indicators to reflect their relative importance in the process of optimizing the candidate strategy; take V2>v0 as the constraint condition for the optimal strategy execution vehicle group and strategy execution speed, and use fuzzy addition to synthesize the fuzzy sets of the two evaluation indicators to obtain a fuzzy comprehensive evaluation function: In the formula, Tra(Q) and Tra(T) are the fuzzy weight values ​​assigned to the two evaluation indicators of oscillation flow rate and cumulative collision time, respectively. are the fuzzy balance coefficients of the two evaluation indicators, oscillation flow rate and cumulative collision time, respectively, λC={λC1,λC2,…,λC K } is the coincidence characteristic root of the two evaluation indicators of oscillation flow rate and cumulative collision time, Nm represents the multi-dimensional information of the vehicle of the mth candidate strategy in the clustering result; According to the comprehensive evaluation results of each candidate vehicle group, fuzzy optimization is performed to select the candidate vehicle group with the highest comprehensive evaluation as the optimal strategy, including the best strategy execution vehicle group K m , optimal strategy execution merge speed V 1m And the corresponding optimal strategy execution absorption speed V 2m .

7. The method for collaborative optimization control of multi-lane traffic shock in a low-permeability environment according to claim 1 is characterized in that: Step S6 further comprises: S61. When the shock wave is detected, the best strategy execution vehicle group is selected, which is composed of n connected autonomous driving vehicles CAV1, CAV2, ..., CAVn that are adjacent to each other in the longitudinal position. Each strategy execution vehicle first changes lanes to a different n lanes, and then merges longitudinally in n-1 stages until the n strategy execution vehicles merge to form a wave-absorbing vehicle group CAV* covering all lanes; S62, after the wave absorbing vehicle group CAV* is formed, the speeds of the strategy executing vehicles CAV1, CAV2, ..., CAVn are simultaneously reduced from V 1m Adjust to the optimal strategy execution absorption speed V 2m .

8. The method for collaborative optimization control of multi-lane traffic shock in a low-permeability environment according to claim 1 is characterized in that: Step S7 further comprises: S71. Determine the basic structure of the BP neural network, including the number of input nodes, hidden layer nodes and output nodes; initialize the weight coefficient w with the initial value 1 ij(0) and w 2 ij(0), select the learning rate and inertia coefficient, and set the time step Δt; S72, sampling the input and output of the current time step And calculate the control error e(t) of this time step as follows: S73, calculating the input and output of each neuron in the BP neural network, and obtaining the online adjustable parameter k of the PID controller according to the output of the BP neural network p ,k i ,k d ; S74. Calculate the wave absorption velocity in the PID control strategy: Among them, V2(t) is the dynamically adjusted forward speed at time t, is the speed gain proportional coefficient at time t, V2(t-1) is the optimal strategy execution absorption speed recursively forward one time step, v0(t) is the congestion speed corresponding to the congested traffic flow at time t, δ is the speed balance adjustment parameter, and k p is the proportional gain coefficient, k i is the integral gain coefficient, k d is the differential gain coefficient; S75, real-time learning of BP neural network, adjusting weighting coefficient w 1 ij(t) and w 2 ij(t), adaptively adjust the PID control parameters.

Images