Intelligent emergency lane opening system based on traffic flow prediction and implementation method thereof

Abstract

The invention discloses an intelligent emergency lane opening system based on traffic flow prediction and an implementation method thereof. The system comprises a data acquisition module, a data processing module, a central control module, a power supply module and a display screen module. The method comprises the following steps: collecting and processing traffic flow in a period of time, and constructing a spatial-temporal characteristic matrix of the traffic flow; a CNN-GRU-Attention prediction model is constructed, and the CNN-GRU- Optimizing parameters of the CNN-GRU-Attention prediction model by using an improved big sugarcane mouse algorithm IGCRA, wherein the parameters comprise a weight matrix and an offset vector; obtaining a traffic flow prediction result based on the optimized model; and judging whether an emergency lane needs to be opened based on the obtained traffic flow prediction result. The method can accurately predict the traffic flow, timely judge and open the emergency lane before congestion, provide extra passing space for vehicles, and avoid excessive concentration of the traffic flow, so that the traffic congestion condition is effectively relieved, the waiting time of the vehicles is shortened, and the overall passing efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of traffic flow prediction and algorithm optimization, and particularly to an intelligent emergency lane opening system based on traffic flow prediction and its implementation method. Background Art

[0002] With the acceleration of the urbanization process and the continuous increase in the number of motor vehicles, the problem of traffic congestion has become increasingly severe. As a key resource to ensure smooth traffic and respond to emergencies, the scientific and effective management of the emergency lane has become increasingly important. However, the traditional emergency lane management mode has many drawbacks, which greatly limits the full play of the role of the emergency lane.

[0003] The traditional management mode mostly relies on manual experience or fixed-time opening strategies. The primary problem is poor dynamic adaptability. Urban traffic flow is constantly changing dynamically, and the traffic flow on different sections and at different times varies greatly. However, the traditional mode lacks a real-time and comprehensive traffic flow monitoring system and cannot adjust the use status of the emergency lane in a timely manner according to the changes in traffic flow, often resulting in waste or over-occupation of emergency lane resources and reducing traffic operation efficiency. In addition, the insufficient accuracy of traffic flow prediction is also a major problem in the traditional mode. In the absence of an advanced prediction model, it is difficult to accurately predict short-term fluctuations and sudden congestions in traffic flow, resulting in the traffic management department being unable to plan and deploy emergency lane resources in advance and often being lagged in response to sudden traffic events, leading to aggravated traffic congestion.

[0004] Although some existing systems attempt to introduce sensors to monitor traffic flow, they fail to deeply integrate the spatio-temporal feature analysis of traffic data with deep learning prediction technology, resulting in inaccurate judgment of the emergency lane opening time and poor timeliness of opening, and it is difficult to meet the actual needs of modern traffic management. In this context, it is urgent to construct an intelligent emergency lane opening system based on traffic flow prediction. Summary of the Invention

[0005] Object of the Invention: The object of the present invention is to provide an intelligent emergency lane opening system based on traffic flow prediction and its implementation method.

[0006] Technical Solution: The intelligent emergency lane opening system based on traffic flow prediction described in the present invention includes a data acquisition module, a data processing module, a central control module, a power supply module, and a display screen module; the data acquisition module detects traffic flow, inputs the detected traffic flow data into the data processing module and predicts the traffic flow within a future period of time, inputs the prediction result into the central control module, and the central control module determines whether to open the emergency lane according to the prediction result and displays the judgment result through the display screen module.

[0007] Further, the power supply module includes a solar array panel, a wind turbine, and a storage battery. The solar array panel is used to implement photovoltaic power generation, the wind turbine is used to implement wind power generation, and the storage battery is used to achieve complementarity and mutual backup between different types of power supplies.

[0008] The implementation method of the intelligent emergency lane opening system based on traffic flow prediction according to the present invention includes the following steps:

[0009] Step 1: Collect and process the traffic flow within a period of time, and construct a spatio-temporal feature matrix of the traffic flow;

[0010] Step 2: Based on the traffic flow data collected in Step 1, construct a CNN-GRU-Attention prediction model;

[0011] Step 3: Based on the model obtained in Step 2, use the improved giant rat algorithm IGCRA to optimize the parameters of the CNN-GRU-Attention prediction model, including the weight matrix and the bias vector;

[0012] Step 4: Obtain the traffic flow prediction result based on the model optimized in Step 3;

[0013] Step 5: Based on the traffic flow prediction result obtained in Step 4, determine whether it is necessary to open the emergency lane.

[0014] Further, in Step 1, the traffic flow data includes time features and spatial features, and the spatio-temporal feature matrix is as follows:

[0015]

[0016] Among them, H is the spatio-temporal feature matrix of the traffic flow, which is composed of the feature vectors h k (y); k is the number of nodes; [x i (1), x i (2), … x i (m)] is the time feature of node i at the m-th time point.

[0017] Further, in Step 2, the establishment process of the CNN-GRU-Attention prediction model is as follows:

[0018] Step 2.1: Use the convolutional layer in the convolutional neural network CNN to extract the time and spatial features of the traffic flow data, and its output O1 can be expressed as:

[0019] O1 = fW1 * I + b1

[0020] Among them, W1 and b1 respectively represent the weights and biases of the I-th convolutional layer; * represents the convolution operation; f is the activation function, and the rectified linear unit is usually selected;

[0021] Step 2.2: The GRU neural network unit consists of two parts, an update gate and a reset gate, and its operating mechanism is as follows:

[0022] z t = σ(W z [h t-1 , x t +b z )

[0023] r t = σ(W r [h t-1 , x t +b r )

[0024] h t = tanh(W h [h t-1 , x t +b h )

[0025] h t = z t *h t-1 +(z t -1)*h tt

[0026] Among them, x t is the traffic flow information input at the current moment; h t-1 is the hidden state at the previous moment; h t is the hidden state passed to the next moment; h tt is the candidate hidden state; r t , z t are the reset gate and the update gate respectively; σ is the sigmoid activation function; tanh is the activation function that maps data to [-1, 1]; W and b are the weight and bias matrices of each control gate respectively;

[0027] Step 2.3: Introduce an attention mechanism based on the CNN-GRU model to perform reallocation calculation of weights and generate a feature sequence, and its expression is as follows:

[0028]

[0029] Among them, n is the time; a is the attention distribution function; s(x n , q) is the attention scoring function, and the scaled dot product model is usually adopted; a n(x, q) is the weighted sum of the input information aggregated using the information selection mechanism.

[0030] Furthermore, in step 3, the improved cane rat algorithm is used to optimize the weight matrices W z 、W r 、W h and the bias constants b z 、b r 、b h The process is as follows:

[0031] Step 3.1: Initialize the population and parameters. The initial positions of the cane rats in the search space are randomly assigned, and its population matrix is as follows:

[0032]

[0033] where X is the cane rat population; x ij is the value of the jth problem variable proposed by the ith cane rat; n is the total number of zebra population members; d is the number of decision variables;

[0034] Step 3.2: Substitute the randomly generated cane rat positions into the objective function for evaluation. Each cane rat represents a candidate solution to the optimization problem, and the objective function formula is as follows:

[0035]

[0036] where n is the number of samples; y i is the true value; is the predicted value;

[0037] Step 3.3: Update the population members. Determine the new positions of the remaining rat groups in the search space according to the positions of the dominant male rats, that is, update the optimal weight matrices W z 、W r 、W h and the bias constants b z 、b r 、b h , and the position formula is as follows:

[0038]

[0039] α = 2 × r × rand - r

[0040] β = 2 × r × μ - r

[0041] where X i is the new position of the ith cane rat, is the value of its jth dimension; x i，j is the value of the current position of the cane rat; xk，j is the position of the dominant male mouse; x k，j is the position of the randomly selected female mouse; F i new is the latest objective function value; F i is the current objective function value; is the objective function value of the dominant male mouse; C is a random number defined within the boundary of the problem space; r is to simulate the influence of rich food sources; C iter is the current iteration number; MaxIter is the maximum iteration number; α is a coefficient to simulate the reduction of food sources; β is a coefficient to prompt the greater cane rat to transfer to other available rich food sources within the breeding area; μ is a random number between [1, 4];

[0042] Step 3.4: Update the search space, develop areas with rich food sources according to the female mouse position, that is, update the optimal weight matrix W z 、W r 、W h and the bias constant b z 、b r 、b h , and the position formula is as follows:

[0043]

[0044] Step 3.5: Determine whether the termination condition is satisfied. If the termination condition is not satisfied, return to Step 3.3; if the constraint condition is satisfied, go to Step 3.6;

[0045] Step 3.6: Output the obtained optimal weight matrix W z 、W r 、W h and the bias constant b z 、b r 、b h 。

[0046] Furthermore, in the said Step 3.4, an adaptive weight factor is introduced to improve the position of the newly generated particle and enhance the local optimization ability of IGCRA. The calculation formula is as follows:

[0047]

[0048] where ω is the adaptive weight factor.

[0049] Furthermore, in the said Step 5, the decision-making model for judging whether to open the emergency lane is as follows:

[0050]

[0051] where γ pThe traffic congestion index for the corresponding period at time t at location p is (t); occ p The road time occupancy rate at time t at location p is (t); Q p The traffic flow at time t at location p is (t)); v p The location speed at time t at location p is (t).

[0052] Furthermore, in the said step 5, the judgment model for determining whether to open the emergency lane is as follows:

[0053]

[0054] Wherein, is the traffic congestion degree evaluation value at time t at location p, 1 represents the unobstructed state, and 2 represents the congested state; I p The single traffic state degree evaluation value at time t at location p is (t); g is the set number of detections.

[0055] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: The present invention utilizes the CNN-GRU-Attention prediction model, which can effectively process time series data, capture the dynamic characteristics of traffic flow changing with time, thereby improving the accuracy of traffic flow prediction. Compared with only using a single prediction method, the present invention uses a combined prediction model, and the prediction result is more accurate and reliable; The GCRA optimization algorithm is used to optimize the prediction model, and the error is further reduced by precisely controlling the model parameters. In addition, by introducing an adaptive weight factor to improve the position of the newly generated particles, the local optimization ability of the IGCRA algorithm is improved; It can accurately predict the traffic flow, timely judge and open the emergency lane before congestion occurs, provide additional passing space for vehicles, avoid excessive concentration of traffic flow, thereby effectively alleviating the traffic congestion condition, reducing the vehicle waiting time, and improving the overall passing efficiency. Description of the Drawings

[0056] Figure 1 is the framework diagram of the intelligent emergency lane opening system based on traffic flow prediction;

[0057] Figure 2 is the flow chart of the improved cane rat algorithm. Detailed Embodiments

[0058] The technical solution of the present invention will be further described below with reference to the drawings.

[0059] As Figure 1 shown, the intelligent emergency lane opening system based on traffic flow prediction described in the present invention includes a data acquisition module, a data processing module, a central control module, a power supply module, and a display screen module.

[0060] The data acquisition module includes a loop detector for detecting traffic flow, that is, detecting the number of vehicles passing through a certain section of the road per unit time;

[0061] The data processing module contains a prediction model and algorithms. The prediction model predicts the traffic flow in the future for a period of time based on the traffic flow data measured by the loop detector; the algorithms are used to optimize the model parameters;

[0062] The central control module determines whether to open the emergency lane according to the prediction result;

[0063] The power supply module includes a solar array panel, a wind turbine and a storage battery. The solar array panel is used for photovoltaic power generation, the wind turbine is used for wind power generation, and the storage battery is used for complementarity and mutual backup between different types of power supplies; the display screen module displays the judgment result of the central control module to provide the traffic police and drivers with the opening situation of the emergency lane.

[0064] The implementation method of the intelligent emergency lane opening system based on traffic flow prediction described in the present invention includes the following steps:

[0065] Step 1: Collect and process the traffic flow within a period of time to construct a spatio-temporal feature matrix of the traffic flow;

[0066] The traffic flow data includes time features and space features. The spatio-temporal feature matrix is as follows:

[0067]

[0068] Among them, H is the spatio-temporal feature matrix of the traffic flow, which is composed of the feature vectors h k (t) of each node in space; k is the number of nodes; [x i (1), x i (2),... x i (m)] are the time features of node i at the m-th time point.

[0069] Step 2: Based on the traffic flow data collected in Step 1, construct a CNN-GRU-Attention prediction model;

[0070] As Figure 2 shown, the establishment process of the CNN-GRU-Attention prediction model is as follows:

[0071] Step 2.1: Use the convolutional layer in the convolutional neural network CNN to extract the time and space features of the traffic flow data, and its output O11 can be expressed as:

[0072] O1 = fW1 * I + b1

[0073] Among them, W1 and b1 respectively represent the weights and biases of the I-th convolutional layer; * represents the convolution operation; f is the activation function, and the rectified linear unit is usually selected;

[0074] Step 2.2: The GRU neural network unit consists of an update gate and a reset gate, and its operation mechanism is as follows:

[0075] z t = σ(W z [h t-1 , x t +b z )

[0076] r t = σ(W r [h t-1 , x t +b r )

[0077] h t = tanh(W h [h t-1 , x t +b h )

[0078] h t = z t *h t-1 +(z t -1)*h tt

[0079] Among them, x t is the traffic flow information input at the current moment; h t-1 is the hidden state at the previous moment; h t is the hidden state passed to the next moment; h tt is the candidate hidden state; r t , z t are the reset gate and the update gate respectively; σ is the sigmoid activation function; tanh is the activation function that maps data to [-1, 1]; W and b are the weight and bias matrices of each control gate respectively;

[0080] Step 2.3: Introduce the attention mechanism based on the CNN-GRU model to perform the reallocation calculation of weights and generate the feature sequence, and its expression is as follows:

[0081]

[0082] Among them, n is the moment; a is the attention distribution function; s(x n , q) is the attention scoring function, and the scaled dot product model is usually adopted; a n(x, q) is the weighted sum of aggregating the input information using the information selection mechanism.

[0083] Step 3: Based on the model obtained in Step 2, use the improved giant cane rat algorithm IGCRA to optimize the parameters of the CNN-GRU-Attention prediction model, including the weight matrix and the bias vector;

[0084] Use the improved giant cane rat algorithm to optimize the weight matrices W z 、W r 、W h and the bias constant b z 、b r 、b h The process is as follows:

[0085] Step 3.1: Initialize the population and parameters. The initial positions of the giant cane rats in the search space are randomly assigned, and its population matrix is as follows:

[0086]

[0087] where X is the giant cane rat population; x ij is the value of the j-th problem variable proposed by the i-th giant cane rat; n is the total number of zebra population members; d is the number of decision variables;

[0088] Step 3.2: Substitute the randomly generated positions of the giant cane rats into the objective function for evaluation. Each giant cane rat represents a candidate solution to an optimization problem, and the objective function formula is as follows:

[0089]

[0090] where n is the number of samples; y i is the true value; is the predicted value;

[0091] Step 3.3: Update the population members. Determine the new positions of the remaining rats in the search space according to the position of the dominant male rat, that is, update the optimal weight matrices W z 、W r 、W h and the bias constant b z 、b r 、b h The position formula is as follows:

[0092]

[0093] α = 2 × r × rand - r

[0094] β = 2 × r × μ - r

[0095] where Xi is the new position of the i-th giant cane rat, is the value of its j-th dimension; x i，j is the value of the current position of the giant cane rat; x k，j is the position of the dominant male rat; x k，j is the position of a randomly selected female rat; F i new is the latest objective function value; F i is the current objective function value; is the objective function value of the dominant male rat; C is a random number defined within the boundary of the problem space; r is to simulate the influence of abundant food sources; C iter is the current iteration number; MaxIter is the maximum iteration number; α is a coefficient to simulate the reduction of food sources; β is a coefficient to prompt the giant cane rat to transfer to other available abundant food sources within the breeding area; μ is a random number between [1, 4];

[0096] Step 3.4: Update the search space, develop areas with abundant food sources according to the positions of female rats, that is, update the optimal weight matrices W z 、W r 、W h and the bias constant b z 、b r 、b h , and the position formula is as follows:

[0097]

[0098] Step 3.5: Determine whether the termination condition is satisfied. If the termination condition is not satisfied, return to Step 3.3; if the constraint condition is satisfied, go to Step 3.6;

[0099] Step 3.6: Output the obtained optimal weight matrices W z 、W r 、W h and the bias constant b z 、b r 、b h .

[0100] The improvement in the above Step 3.4 is to introduce an adaptive weight factor to improve the position of the newly generated particle and enhance the local optimization ability of IGCRA. The calculation formula is as follows:

[0101]

[0102] where ω is the adaptive weight factor.

[0103] Step 4: Obtain the traffic flow prediction result based on the model optimized in Step 3;

[0104] Step 5: Based on the traffic flow prediction result obtained in Step 4, determine whether to open the emergency lane.

[0105] The decision-making model for determining whether to open the emergency lane is as follows:

[0106]

[0107] where γ p (t) is the traffic congestion index corresponding to the time period at location p at time y; occ p (t) is the road time occupancy rate at location p at time t; Q p (y)) is the traffic flow at location p at time t; v p (t) is the location speed at location p at time t.

[0108] The judgment model for determining whether to open the emergency lane is as follows:

[0109]

[0110] where is the traffic congestion degree evaluation value at location p at time t, 1 represents the unobstructed state, and 2 represents the congested state; I p (y) is the single traffic state degree evaluation value at location p at time t; g is the set number of detections.

Claims

1. An intelligent emergency lane opening system based on traffic flow prediction, characterized in that, It includes a data acquisition module, a data processing module, a central control module, a power supply module, and a display screen module; the data acquisition module detects traffic flow, inputs the detected traffic flow data into the data processing module, predicts the traffic flow within a certain period in the future, inputs the prediction result into the central control module, and the central control module determines whether to open the emergency lane according to the prediction result and displays the judgment result through the display screen module.

2. The intelligent emergency lane opening system based on traffic flow prediction according to claim 1, wherein The power supply module includes a solar array panel, a wind turbine, and a storage battery. The solar array panel is used for photovoltaic power generation, the wind turbine is used for wind power generation, and the storage battery is used for complementarity and mutual backup between different types of power supplies.

3. An implementation method of an intelligent emergency lane opening system based on traffic flow prediction, characterized in that, It includes the following steps: Step 1: Collect and process the traffic flow within a certain period, and construct a spatio-temporal feature matrix of the traffic flow; Step 2: Based on the traffic flow data collected in Step 1, construct a CNN-GRU-Attention prediction model; Step 3: Based on the model obtained in Step 2, use the improved giant rat algorithm IGCRA to optimize the parameters of the CNN-GRU-Attention prediction model, including the weight matrix and the bias vector; Step 4: Obtain the traffic flow prediction result based on the model optimized in Step 3; Step 5: Based on the traffic flow prediction result obtained in Step 4, determine whether to open the emergency lane.

4. The implementation method of the intelligent emergency lane opening system based on traffic flow prediction according to claim 3, characterized in that In Step 1, the traffic flow data includes time features and space features, and the spatio-temporal feature matrix is as follows: Among them, H is the spatio-temporal feature matrix of traffic flow, which is composed of the feature vectors h k (t) of each node in space; k is the number of nodes; [x i (1), x i (2), … x i (m)] is the time feature of node i at the m-th time point.

5. The implementation method of the intelligent emergency lane opening system based on traffic flow prediction according to claim 3, characterized in that, In Step 2, the establishment process of the CNN-GRU-Attention prediction model is as follows: Step 2.1: Use the convolutional layer in the convolutional neural network CNN to extract the time and space features of the traffic flow data, and its output O1 can be expressed as: O1 = fW1 * I + b1 Where, W1 and b1 respectively represent the weight and bias of the I-th convolutional layer; * represents the convolutional operation; f is the activation function, and usually the rectified linear unit is selected; Step 2.2: The GRU neural network unit consists of an update gate and a reset gate, and its operation mechanism is as follows: z t = σ(W z [h t-1 , x t + b z ) r t = σ(W r [h t-1 , x t + b r ) h t = tanh(W h [h t-1 , x t + b h ) h t = z t * h t-1 +(z t - 1) * h tt where x t is the traffic flow information input at the current moment; h t-1 is the hidden state at the previous moment; h t is the hidden state passed to the next moment; h tt is the candidate hidden state; r t , z t are the reset gate and the update gate respectively; σ is the sigmoid activation function; tanh is the activation function that maps data to [-1, 1]; W and b are the weight and bias matrices of each control gate respectively; Step 2.3: Introduce the attention mechanism on the basis of the CNN-GRU model, perform the calculation of weight redistribution and generate the feature sequence, and its expression is as follows: Among them, n is the time; a is the attention distribution function; s(x n , q) is the attention scoring function, and the scaled dot product model is often adopted; a n (x, q) is the weighted sum that aggregates the input information using the information selection mechanism.

6. The implementation method of the intelligent emergency lane opening system based on traffic flow prediction according to claim 3, characterized in that, In step 3, the weight matrices W z , W r , W h and the bias constants b z , b r , b h of the CNN-GRU-Attention model are optimized by using the improved cane rat algorithm as follows: Step 3.1: Initialize the population and parameters. The initial positions of the giant rats in the search space are randomly assigned, and its population matrix is as follows: Among them, X is the cane rat population; x ij is the value of the j-th problem variable proposed for the i-th cane rat; n is the total number of all members of the zebra population; d is the number of decision variables; Step 3.2: Substitute the randomly generated positions of the giant rats into the objective function for evaluation. Each giant rat represents a candidate solution to an optimization problem, and the objective function formula is as follows: Among them, n is the number of samples; y i is the true value; is the predicted value; Step 3.3: Update the population members, and determine the new positions of the remaining rat populations in the search space according to the position of the dominant male rat, that is, update the optimal weight matrix W z , W r , W h and the bias constant b z , b r , b h , and the position formula is as follows: α = 2×r×rand - r β = 2×r×μ - r Among them, X i is the new position of the i-th giant cane rat, and its value at the j-th dimension; x i，j is the value of the current position of the giant cane rat; x k，j is the position of the dominant male rat; x k，j is the position of the randomly selected female rat; F i new is the latest objective function value; F i is the current objective function value; is the objective function value of the dominant male rat; C is a random number defined within the boundary of the problem space; r is to simulate the influence of abundant food sources; C iter is the current iteration number; MaxIter is the maximum number of iterations; α is the coefficient to simulate the reduction of food sources; β is the coefficient to prompt the giant cane rat to transfer to other available abundant food sources within the breeding area; μ is a random number between [1, 4]; Step 3.4: Update the search space, develop areas with rich food sources according to the positions of female mice, that is, update the optimal weight matrices W z , W r , W h and the bias constant b z , b r , b h , and the position formula is as follows: Step 3.5: Determine whether the termination condition is satisfied. If the termination condition is not satisfied, return to Step 3.3; if the constraint condition is satisfied, go to Step 3.6; Step 3.6: Output the optimal weight matrix W z , W r , W h and the bias constant b z , b r , b h .

7. The implementation method of the intelligent emergency lane opening system based on traffic flow prediction according to claim 6, characterized in that, In Step 3.4, introduce an adaptive weight factor to improve the position of the newly generated particle and enhance the local optimization ability of IGCRA. The calculation formula is as follows: Where, ω is the adaptive weight factor.

8. The implementation method of the intelligent emergency lane opening system based on traffic flow prediction according to claim 1, characterized in that, In Step 5, the decision model for determining whether to open the emergency lane is as follows: Among them, γ p (y) is the traffic congestion index corresponding to the time period at time t at location p; occ p (t) is the road time occupancy rate at time t at location p; Q p (y)) is the traffic flow at time y at location p; v p (t) is the location speed at time t at location p.

9. The implementation method of the intelligent emergency lane opening system based on traffic flow prediction according to claim 1, characterized in that, In Step 5, the judgment model for determining whether to open the emergency lane is as follows: Among them, is the traffic congestion evaluation value at time t at location p, where 1 represents the unobstructed state and 2 represents the congested state; I p (t) is the single traffic state evaluation value at time t at location p; g is the set number of detections.

Images