A collaborative management method for intelligent highways based on feedback control

Abstract

The present invention discloses a method for collaborative control of intelligent highways based on feedback control, which includes the following steps: constructing an intelligent highway collaborative control system, dividing the entire highway into different control sections according to the location of the control equipment on the entire highway; the control equipment collects the traffic flow status information at each control section of the highway in real time, and restores and predicts the traffic flow operation status of each control section according to the status information of each section; establishes a collaborative control model of variable speed limit and ramp control, obtains the optimal control strategy for each control section and publishes it through the control equipment. The present invention focuses on the upstream and downstream of an entire highway and each entrance and exit ramp, and provides an effective technical method and framework for accurately ensuring the safety and traffic efficiency of traffic flow operation.

Description

Technical Field

[0001] The present invention belongs to the technical field of highway traffic management and control, and in particular to a collaborative management and control method of an intelligent highway based on feedback control. Background Art

[0002] As an important means to improve the efficiency of highway management services, smart highway control has been widely studied and applied in smart highways in recent years. For example, the Wuxi section of the Nanjing-Shanghai Expressway and the Shanghai-Hangzhou-Ningbo Expressway have adopted control measures such as variable speed limits and ramp control. However, the current control measures are mainly aimed at alleviating traffic congestion, and traffic operation safety is not listed as a key issue, so there is still a lack of relevant technical research. Summary of the invention

[0003] The purpose of the present invention is to provide a method for collaborative control of intelligent highways based on feedback control, based on the checkpoint data and radar-vision fusion data of the intelligent highway collaborative control system, to achieve high-precision prediction of the future traffic situation of the entire highway, and to complete the intelligent push of variable speed limits and ramp control measures within the control period, so as to achieve the purpose of smoothing traffic flow, reducing traffic incidents and avoiding traffic accidents.

[0004] The technical solution to achieve the purpose of the present invention is: a method for collaborative management and control of intelligent highways based on feedback control, the method comprising the following steps:

[0005] Step 1: Build a smart highway collaborative control system, and divide the entire highway into different control sections according to the locations of control devices in the smart highway collaborative control system on the entire highway;

[0006] Step 2: The control equipment collects the traffic flow status information at each control section of the expressway in real time, and restores and predicts the traffic flow operation status of each control section based on the traffic flow status information of each control section;

[0007] Step 3: With improving traffic safety as the main goal and road capacity as the secondary goal, a collaborative control model of variable speed limit and ramp control is established to obtain the optimal control strategy for each controlled section and publish it through the control device.

[0008] Furthermore, the control device of the smart highway collaborative management and control system in step 1 includes:

[0009] Traffic flow information perception module, including checkpoint equipment and radar equipment, is used to collect and count vehicle speed, traffic flow and vehicle type information at each control section;

[0010] Online real-time simulation control system, used to restore the traffic operation status of the controlled section in real time, conduct short-term traffic flow prediction and verify the control effect;

[0011] The control strategy release module is used to implement entrance ramp signal lights to control vehicles entering the main line of the expressway, and variable information signs arranged along the main line to release the optimal recommended driving speed to vehicles.

[0012] Furthermore, the control device in step 2 collects the traffic flow status information at each control section of the expressway in real time, and restores and predicts the traffic flow operation status of each control section according to the traffic flow status information of each section. The specific process includes:

[0013] Step 2.1, data collection and preprocessing; the data collection includes: the checkpoint equipment and radar equipment collect and count the vehicle speed, traffic flow and vehicle type information of each control section; the preprocessing includes: intelligent completion and abnormal deletion of the checkpoint data;

[0014] Step 2.2, real-time restoration and prediction of the traffic operation status of the control section; in the online real-time simulation control system, the short-term traffic volume prediction method based on deep learning is used to construct feature engineering using the time characteristics and external characteristics of the pre-processed checkpoint data. Considering the topological relationship between real-time and high-speed space, the traffic flow prediction model is simplified and spatial characteristics are discarded. Through the operation of the long short-term memory network LSTM, the multi-layer perceptron MLP and two fusion mechanisms, and with the set prediction period, the prediction of future short-term traffic flow is realized. The specific process is as follows:

[0015] The time characteristics are divided into proximity characteristics, cycle characteristics and trend characteristics;

[0016] External features include the time slice number and week information corresponding to the time slice to be predicted, and introduce features such as whether it is a weekend or a holiday in the form of one-hot encoding;

[0017] The mechanisms of feature fusion include:

[0018] Fusion 1: The following Hadamard product-based weighted fusion mechanism is adopted:

[0019] M fision =W c ⊙M c +W p ⊙M p +W t ⊙M t (1)

[0020] Where: W c ,W p ,W tRespectively represent the weights of proximity, cycle and trend features in the time features to be learned, M c ,M p ,M t Respectively represent the tensor output of each feature after LSTM calculation;

[0021] Fusion 2: The external features are fused with the output of MLP using a simple horizontal splicing method.

[0022] Furthermore, the preprocessing in step 2.1 is specifically as follows: using an intelligent algorithm that integrates the two types of perception data information and combining radar data to intelligently complete the checkpoint data and delete anomalies.

[0023] Furthermore, in step 3, with the improvement of traffic safety as the main goal and road capacity as the secondary goal, a variable speed limit and ramp control collaborative control model is established to obtain the optimal control strategy for each controlled section and publish it through the control device, specifically including:

[0024] Step 3.1, variable speed limit control strategy, based on the restoration and prediction results of the traffic flow operation status of each controlled section obtained in step 2, taking improving the safety of traffic operation in the controlled section as the optimization goal of the variable speed limit control strategy, establishing a feedback control model and calculating the optimal speed limit value;

[0025] Step 3.2, ramp control strategy, by adjusting the entrance ramp signal lights, controlling the traffic volume entering the main line from the ramp, so that the main line traffic downstream of the entrance ramp is in a stable state;

[0026] Step 3.3, collaborative control strategy, through the coordinated control of variable speed limit control strategy and ramp control strategy, balance the traffic demand upstream and downstream of the entrance ramp, so that the highway as a whole is maintained in a stable state;

[0027] Step 3.4, the optimal control strategy is released. By changing the speed limit value on the variable information sign on the roadside, the optimal speed obtained by the variable speed limit control strategy is released to the vehicle to achieve variable speed limit control; by adjusting the proportion of traffic lights on the entrance ramp, the traffic volume entering the main line section is controlled to achieve ramp control.

[0028] Furthermore, the core control parameters of the variable speed limit control strategy based on the feedback control model in step 3.1 include the following:

[0029] 1) Threshold, i.e. the time when the control is turned on; when the accident risk coefficient during traffic operation in the controlled section exceeds the preset threshold, the variable speed limit control is turned on and the optimal speed limit value calculated by the feedback control model is used; where the accident risk coefficient CRI is defined as:

[0030]

[0031] Among them, R is the number of variables, r represents the sequence number of the variable, ω r is the parameter x r The regression weight coefficient, x r is the rth risk influencing variable, considering the relationship between the speed difference and flow difference between each control point in the research section and the accident risk coefficient, x r The specific selection influencing factors and their definitions are shown in Table 1:

[0032] Table 1 Influencing variables

[0033]

[0034] Where M represents the number of control points in the study section; m represents the sequence number of the control point; Speed m,car 、Speed m,truck Respectively represent the average speed of cars and trucks passing through the mth control point in a control cycle; Speed aver,car 、Speed aver,truck Respectively represent the average speed of all cars and trucks at M control points within a control cycle; Vol m,car 、Vol m,truck Respectively represent the average traffic flow of passenger cars and trucks passing through the mth control point in a control cycle; Vol aver,car 、Vol aver,truck Respectively represent the average traffic flow of all passenger cars and trucks at all M control points within a control period;

[0035] 2) Control cycle, i.e., how often the control strategy is adjusted; the control cycle is the same as the prediction cycle described in step 2.2;

[0036] 3) Speed ​​limit adjustment value: in terms of time, the adjustment value is set to 5km / h; in terms of space, the adjustment value is not greater than 10km / h;

[0037] 4) Optimal speed limit value: Calculate CRI(k+1) and minimize the accident risk coefficient of the entire road to obtain the optimal speed limit value VSL of the i-th control section within the k+1 control period i (k+1):

[0038]

[0039] Among them, CRI i (k+1) represents the accident risk coefficient of the i-th controlled section in the k+1 control cycle;

[0040] When solving the optimal speed limit value, add the following two constraints:

[0041]

[0042] Among them, d s Indicates the upper limit of the speed limit change between two adjacent control sections, d k Indicates the upper limit of the speed limit change between two adjacent control cycles; V VSL,i (k) and V VSL,i+1 (k) represent the optimal speed limit values ​​of the control section i and the downstream control section i+1 in the control period k respectively; V VSL,i (k) and V VSL,i (k+1) represent the optimal speed limit values ​​of control period k and the next control period k+1 on control section i, respectively.

[0043] Furthermore, the ramp control strategy described in step 3.2, the core control parameters include the following:

[0044] Based on the classic closed-loop feedback ALINEA control strategy, the impact of the upstream road on the safety performance of the entrance ramp is considered and an accident risk adjustment item is added:

[0045]

[0046] Where r(k) is the ramp pass rate calculated in the kth control cycle; r(k-1) is the ramp pass rate in the (k-1)th control cycle; K R is a parameter of a calibration nature; is the expected share of the mainline downstream, The value of is set to the optimal occupancy or less than the optimal occupancy; out (k-1) is the actual downstream occupancy measured in (k-1) control cycles; K s is the accident risk correction parameter; CRI k is the accident risk coefficient of the road upstream of the entrance ramp in the kth control period; CRI crit is the accident risk factor threshold.

[0047] Furthermore, the core control parameters of the collaborative control strategy described in step 3.3 include the following:

[0048] When the variable speed limit control strategy and ramp control strategy are used in coordination, the corresponding adjustment term is added to r(k) to obtain r * (k):

[0049]

[0050] Among them, q cap is the road capacity downstream of the entrance ramp after considering the speed limit; i(k) is the arriving traffic volume of the downstream road segment i after the speed limit in the k control period;

[0051] In actual application, the ramp pass rate is controlled by setting the entrance ramp signal light. The pass rate r(k) corresponds to the green light duration g(t) in the signal control cycle, in seconds:

[0052]

[0053] g min <g(t)<g max (8)

[0054] Among them, C is a fixed control period; r sat is the maximum passing rate of the ramp within a control cycle; g min is the shortest green light time; g max is the longest green light time;

[0055] When calculating the traffic light configuration, the green light duration g(t) is adjusted as follows:

[0056]

[0057] Among them, q is the number of queued vehicles on the entrance ramp, and h(t) is the adjusted green light duration.

[0058] Compared with the prior art, the present invention has the following significant advantages:

[0059] 1) The present invention combines real-time checkpoint data (including vehicle desensitized license plates, vehicle speed, vehicle type and vehicle lane, etc.) with radar data of key sections (vehicle ID, vehicle lane changing behavior, vehicle acceleration, vehicle speed, etc.) to restore road traffic flow in the intelligent highway collaborative management and control system, and uses the improved ST-LSTM (space-time long short-term memory neural network) model to achieve high-precision prediction of future traffic conditions in the entire highway area, and thereby dynamically manage traffic management on the highway to ensure the safe operation of the entire highway.

[0060] 2) The present invention selects two strategies, namely variable speed limit and ramp control, and realizes intelligent push of control measures based on real-time simulation and situation prediction, thereby improving the internal mechanism of safety, thereby achieving the purpose of smoothing traffic flow, reducing traffic incidents and avoiding traffic accidents.

[0061] The present invention is further described in detail below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of the process of the present invention.

[0063] Figure 2It is a schematic diagram of the content framework of the present invention.

[0064] Figure 3 It is a specific architecture diagram of the prediction model used in the feature fusion part of the present invention.

[0065] Figure 4 It is a schematic diagram of a divided highway used in an embodiment of the present invention. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0067] In one embodiment, in combination Figure 1 The present invention provides a method for collaborative management of intelligent highways based on feedback control. The specific framework is as follows: Figure 2 As shown, the method comprises the following steps:

[0068] Step 1: Build a smart highway collaborative management and control system, and divide the entire highway into different control sections according to the location of the control equipment in the smart highway collaborative management and control system on the entire highway.

[0069] (1) The control equipment of the smart highway collaborative management and control system includes:

[0070] Traffic flow information perception module, including checkpoint equipment and radar equipment, is used to collect and count vehicle speed, traffic flow and vehicle type information at each section;

[0071] Online real-time simulation control system, real-time restoration of traffic operation status of controlled sections, short-term traffic flow prediction and control effect simulation and verification;

[0072] The control strategy release module, the entrance ramp signal lights control the vehicles entering the main line of the expressway, and the variable information signs arranged along the main line section release the optimal recommended driving speed to the vehicles.

[0073] (2) The entire highway is divided into different control sections according to the location of highway control equipment (checkpoints, radars), the location of entrance ramps, and the geometric characteristics of the road. The divided control sections serve as the minimum control units for the release and implementation of control strategies.

[0074] The control units after the 110 km high-speed segmentation in the embodiment are as follows Figure 4 As shown, each control unit implements different control strategies, but the control strategies of each unit need to take into account the impact of upstream and downstream road traffic flows.

[0075] Step 2: The control equipment collects the traffic flow status information at each control section of the expressway in real time, and restores and predicts the traffic flow operation status of each control section based on the traffic flow status information of each control section. The specific process includes:

[0076] Step 2.1, data collection and preprocessing; the data collection includes: the checkpoint equipment and radar equipment collect and count the vehicle speed, traffic flow and vehicle model information of each control section; the preprocessing includes: intelligent completion and abnormal deletion of the checkpoint data, specifically: using an intelligent algorithm that integrates two types of perception data information, combined with radar data, to intelligently complete and delete the checkpoint data, including vehicle license plate recognition errors, abnormal vehicle data elimination, combined with the location of the main line checkpoint to make a judgment, and eliminate abnormal data such as vehicles whose recorded time and real time do not match.

[0077] Here, the checkpoint data covers a wide range, but the real-time collection of checkpoint data generally has a certain proportion of field missing and errors, especially under extreme weather conditions, such as rain, snow, fog and other adverse weather conditions will cause the accuracy of data acquisition to be further reduced. Radar equipment has a small coverage range and is generally deployed in key sections, but radar data is not affected by natural conditions such as light, visibility, and bad weather, which just makes up for the shortcomings of checkpoint video data. The fusion of the two data has obvious advantages, which can maximize the accuracy of traffic flow status information in key sections (entrance ramps and their downstream sections).

[0078] Step 2.2, real-time restoration and prediction of the traffic operation status of the control section; in the online real-time simulation control system, the short-term traffic volume prediction method based on deep learning is constructed by using the time characteristics and external characteristics of the pre-processed checkpoint data to construct feature engineering, considering the topological relationship between real-time and high-speed space, simplifying the traffic flow prediction model and discarding the spatial characteristics, and through the operation of the long short-term memory network LSTM, the multi-layer perceptron MLP and the two fusion mechanisms, and with a set prediction period (preferably, 10 minutes as a prediction period), the prediction of future short-term traffic flow is realized. The specific process is as follows:

[0079] Time features are the focus of feature engineering. Considering the different effects of short-term, medium-term and long-term features on the prediction of the current time slice, time features are divided into adjacent features (X t-1 ,X t-2 ,X t-3 represents the traffic characteristics of the three time slices closest to the predicted time slice) periodic characteristics (X t-144 ,X t-144*2 ,X t-144*3 represents the traffic characteristics of the same time slice in the previous three days) and trend characteristics (X t-1*(7*144) ,X t-2*(7*144) ,X t-3*(7*144)represents the traffic characteristics of the same time slice in each of the previous three weeks);

[0080] The external features mainly consider the time slice sequence number and week information corresponding to the time slice to be predicted, and introduce features such as whether it is a weekend or a holiday in the form of one-hot encoding;

[0081] The mechanisms of feature fusion include:

[0082] Fusion 1: Considering the different influences of the three time features on the present, the following weighted fusion mechanism based on Hadamard product ⊙ is adopted:

[0083] M fision =W c ⊙M c +W p ⊙M p +W t ⊙M t (10)

[0084] Where: W c ,W p ,W t Respectively represent the weights of proximity, cycle and trend features in the time features to be learned, M c ,M p ,M t Respectively represent the tensor output of each feature after LSTM calculation;

[0085] Fusion 2: A simple horizontal splicing method is used to fuse the external features with the output of MLP. The specific architecture of the prediction model used is as follows: Figure 3 shown.

[0086] The checkpoint data used in the embodiment is divided into traffic flow by passenger cars and large trucks, and then evaluated using two statistical indicators, MAE and RMSE, where y i represents the true value, Represents the predicted value. The model accuracy can be intuitively reflected by comparing the indicators with the actual traffic distribution as shown in Table 2:

[0087]

[0088]

[0089] Table 2 Model prediction accuracy

[0090]

[0091] Step 3: With the improvement of traffic safety as the primary goal and road capacity as the secondary goal, a coordinated control model of variable speed limit and ramp control is established to obtain the optimal control strategy for each controlled section and publish it through the control device, including:

[0092] Step 3.1, variable speed limit control strategy, based on the restoration and prediction results of the traffic flow operation status of each controlled section obtained in step 2, with improving the safety of traffic operation in the controlled section as the optimization goal of the variable speed limit control strategy, a feedback control model is established and the optimal speed limit value is calculated; the core control parameters further include the following:

[0093] 1) Threshold, i.e. the time when the control is turned on; when the accident risk coefficient during traffic operation in the controlled section exceeds the preset threshold, the variable speed limit control is turned on and the optimal speed limit value calculated by the feedback control model is used; where the accident risk coefficient CRI is defined as:

[0094]

[0095] Among them, R is the number of variables, r represents the sequence number of the variable, ω r is the parameter x r The regression weight coefficient, x r is the rth risk influencing variable, considering the relationship between the speed difference and flow difference between each control point in the research section and the accident risk coefficient, x r The specific selection influencing factors and their definitions are shown in Table 3:

[0096] Table 3 Influencing variables

[0097]

[0098] Where M represents the number of control points in the study section; m represents the sequence number of the control point; Speed m,car 、Speed m,truck Respectively represent the average speed of cars and trucks passing through the mth control point in a control cycle; Speed aver,car 、Speed aver,truck Respectively represent the average speed of all cars and trucks at M control points within a control cycle; Vol m,car 、Vol m,truck Respectively represent the average traffic flow of passenger cars and trucks passing through the mth control point in a control cycle; Vol aver,car 、Vol aver,truck Respectively represent the average traffic flow of all passenger cars and trucks at all M control points within a control period;

[0099] 2) Control cycle, that is, how often the control strategy is adjusted; the process of traffic flow operation is not stable. After a period of time after the variable speed limit control is turned on, the operation status changes, so the control strategy needs to be adjusted again. As described in step 2.2, the traffic flow prediction cycle is 10 minutes. In order to ensure that each control strategy modification is based on the latest traffic operation status and considering that it is not appropriate for the driver to adjust the speed multiple times in a short period of time, the control cycle is also uniformly selected as 10 minutes;

[0100] 3) Speed ​​limit adjustment value: In terms of time, in order to ensure smooth traffic flow in the same controlled section, the speed limit value difference between the two control cycles should not be too large. Based on actual experience and literature research, the recommended adjustment value is 5km / h; In terms of space, in the intelligent highway collaborative control system, the entire highway is divided into different controlled sections, so it is necessary to consider the impact of speed change differences between upstream and downstream adjacent sections. Based on actual experience and literature research, the recommended speed difference value should not be greater than 10km / h;

[0101] 4) Optimal speed limit value: In the constructed intelligent highway collaborative control system, the traffic flow operation status in the next control cycle can be predicted. The online real-time simulation control system can verify the impact of different speed limit values ​​on the traffic flow operation status, which is ultimately reflected in the changes in the average vehicle speed and flow at each control point in the study section. The CRI (k+1) is calculated and the accident risk coefficient of the entire road is minimized to obtain the optimal speed limit value VSL of the i-th control section in the k+1 control cycle. i (k+1):

[0102]

[0103] Among them, CRI i (k+1) represents the accident risk coefficient of the i-th controlled section in the k+1 control cycle;

[0104] When solving the optimal speed limit value, add the following two constraints:

[0105] |V VSL,i+1 (k)-V VSL,i (k)|≤d s (15)

[0106] |V VSL,i (k+1)-V VSL,i (k)|≤d k

[0107] Among them, d s Indicates the upper limit of the speed limit change between two adjacent control sections, d k Indicates the upper limit of the speed limit change between two adjacent control cycles; V VSL,i(k) and V VSL,i+1 (k) represent the optimal speed limit values ​​of the control section i and the downstream control section i+1 in the control period k respectively; V VSL,i (k) and V VSL,i (k+1) represent the optimal speed limit values ​​of control period k and the next control period k+1 on control section i, respectively.

[0108] Step 3.2, ramp control strategy, controls the traffic volume entering the main line from the ramp by adjusting the entrance ramp signal lights, ensures that the main line traffic downstream of the entrance ramp is in a stable state, reduces accidents, reduces congestion, and improves overall operation efficiency; the core control parameters include the following:

[0109] Based on the classic closed-loop feedback ALINEA control strategy, the impact of the upstream road on the safety performance of the entrance ramp is considered and an accident risk adjustment item is added:

[0110]

[0111] Where r(k) is the ramp pass rate calculated in the kth control cycle; r(k-1) is the ramp pass rate in the (k-1)th control cycle; K R is a parameter of a calibration nature; is the expected share of the mainline downstream, The value of is set to the optimal occupancy or less than the optimal occupancy; out (k-1) is the actual downstream occupancy measured in (k-1) control cycles; K s is the accident risk correction parameter; CRI k is the accident risk coefficient of the road upstream of the entrance ramp in the kth control period; CRI crit is the accident risk factor threshold

[0112] Step 3.3, collaborative control strategy, through the coordinated control of variable speed limit and ramp control strategy, balances the traffic demand upstream and downstream of the entrance ramp, keeps the highway in a stable state as a whole, effectively reduces the speed difference between the main line and ramp traffic, and thus improves the overall safety level and stability. The core control parameters further include the following:

[0113] After the variable speed limit is implemented on the controlled section, the impact of the change in the traffic capacity of the section, that is, the change in the optimal occupancy rate, when calculating the entrance ramp pass rate, when the variable speed limit control strategy and the ramp control strategy are used in coordination, the corresponding adjustment term is added to r(k) to obtain r * (k):

[0114]

[0115] Among them, qcap is the road capacity downstream of the entrance ramp after considering the speed limit; i (k) is the arriving traffic volume of the downstream road segment i after the speed limit in the k control period;

[0116] In actual application, the ramp pass rate is controlled by setting the entrance ramp signal light. The pass rate r(k) corresponds to the green light duration g(t) in the signal control cycle, in seconds:

[0117]

[0118] g min <g(t)<g max (19)

[0119] Among them, C is a fixed control period; r sat is the maximum passing rate of the ramp within a control cycle; g min is the shortest green light time; g max is the longest green light time;

[0120] When calculating the signal configuration, in order to avoid excessive queue time delays on the entrance ramp, the green light duration g(t) is adjusted as follows:

[0121]

[0122] Among them, q is the number of queued vehicles on the entrance ramp, and h(t) is the adjusted green light duration.

[0123] Step 3.4, the optimal control strategy is released. By changing the speed limit value on the variable information sign on the roadside, the optimal speed obtained by the variable speed limit control strategy is released to the vehicle to achieve variable speed limit control; by adjusting the proportion of traffic lights on the entrance ramp, the traffic volume entering the main line section is controlled to achieve ramp control.

[0124] As a specific example, in one of the embodiments, the present invention is further verified. Using simulation software, the intelligent highway collaborative control system proposed in the present invention is built, and a comparative test is conducted to verify the control effect. In terms of traffic safety, CRI and its reduction percentage, TTC (Time to Collision) and its increase percentage are selected; in terms of traffic efficiency, the average speed in the research section is selected. aver The specific results are shown in Table 4.

[0125] Table 4 Comparison results of control test

[0126]

[0127] From the above, we can see that the application of variable speed limit can effectively improve traffic safety, but it reduces the average speed; however, the control effect of variable speed limit and ramp control is the best, which can improve traffic safety and average speed at the same time, reduce the accident risk coefficient by 10.00%, and increase the average speed by 2.54%.

[0128] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for collaborative management and control of intelligent highways based on feedback control, characterized in that: The method comprises the following steps: Step 1: Build a smart highway collaborative control system, and divide the entire highway into different control sections according to the locations of control devices in the smart highway collaborative control system on the entire highway; Step 2: The control equipment collects the traffic flow status information at each control section of the expressway in real time, and restores and predicts the traffic flow operation status of each control section based on the traffic flow status information of each control section; Step 3: With the improvement of traffic safety as the primary goal and road capacity as the secondary goal, a coordinated control model of variable speed limit and ramp control is established to obtain the optimal control strategy for each controlled section and publish it through the control device; The control equipment of the smart highway collaborative management and control system described in step 1 includes: Traffic flow information perception module, including checkpoint equipment and radar equipment, is used to collect and count vehicle speed, traffic flow and vehicle type information at each control section; Online real-time simulation control system, used to restore the traffic operation status of the controlled section in real time, conduct short-term traffic flow prediction and verify the control effect; The control strategy release module is used to implement the entrance ramp signal lights to control vehicles entering the main line of the expressway, and the variable information signs arranged along the main line to release the optimal recommended driving speed to vehicles; The control device in step 2 collects the traffic flow status information at each control section of the expressway in real time, and restores and predicts the traffic flow operation status of each control section based on the traffic flow status information of each section. The specific process includes: Step 2.1, data collection and preprocessing; the data collection includes: the checkpoint equipment and radar equipment collect and count the vehicle speed, traffic flow and vehicle type information of each control section; the preprocessing includes: intelligent completion and abnormal deletion of the checkpoint data; Step 2.2, real-time restoration and prediction of the traffic operation status of the control section; in the online real-time simulation control system, the short-term traffic volume prediction method based on deep learning is used to construct feature engineering using the time characteristics and external characteristics of the pre-processed checkpoint data. Considering the topological relationship between real-time and high-speed space, the traffic flow prediction model is simplified and spatial characteristics are discarded. Through the operation of the long short-term memory network LSTM, the multi-layer perceptron MLP and two fusion mechanisms, and with the set prediction period, the prediction of future short-term traffic flow is realized. The specific process is as follows: The time characteristics are divided into proximity characteristics, cycle characteristics and trend characteristics; External features include the time slice number and week information corresponding to the time slice to be predicted, and introduce features such as whether it is a weekend or a holiday in the form of one-hot encoding; The mechanisms of feature fusion include: Fusion 1: The following Hadamard product-based weighted fusion mechanism is adopted: M fision =W c ⊙M c +W p ⊙M p +W t ⊙M t (1)W c ,W p ,W t Respectively represent the weights of proximity, cycle and trend features in the time features to be learned, M c ,M p ,M t Respectively represent the tensor output of each feature after LSTM calculation; Fusion 2: The external features are fused with the output of MLP by simple horizontal splicing; Step 3 takes improving traffic safety as the main goal and road capacity as the secondary goal, establishes a coordinated control model of variable speed limit and ramp control, obtains the optimal control strategy for each controlled section and publishes it through the control device, including: Step 3.1, variable speed limit control strategy, based on the restoration and prediction results of the traffic flow operation status of each controlled section obtained in step 2, taking improving the safety of traffic operation in the controlled section as the optimization goal of the variable speed limit control strategy, establishing a feedback control model and calculating the optimal speed limit value; Step 3.2, ramp control strategy, by adjusting the entrance ramp signal lights, controlling the traffic volume entering the main line from the ramp, so that the main line traffic downstream of the entrance ramp is in a stable state; Step 3.3, collaborative control strategy, through the coordinated control of variable speed limit control strategy and ramp control strategy, balance the traffic demand upstream and downstream of the entrance ramp, so that the highway as a whole is maintained in a stable state; Step 3.4, optimal control strategy release, by changing the speed limit value on the variable information sign on the roadside, the optimal speed obtained by the variable speed limit control strategy is released to the vehicle, and variable speed limit control is realized; by adjusting the proportion of traffic lights on the entrance ramp, the traffic volume entering the main line section is controlled to realize ramp control; The core control parameters of the variable speed limit control strategy based on the feedback control model in step 3.1 include the following: 1) Threshold, i.e. the time when the control is turned on; when the accident risk coefficient during traffic operation in the controlled section exceeds the preset threshold, the variable speed limit control is turned on and the optimal speed limit value calculated by the feedback control model is used; where the accident risk coefficient CRI is defined as: Among them, R is the number of variables, r represents the sequence number of the variable, ω r is the parameter x r The regression weight coefficient, x r is the rth risk influencing variable, considering the relationship between the speed difference and flow difference between each control point in the research section and the accident risk coefficient, x r The specific selection factors and their definitions are as follows: Variable: x1 = Speed diff,car ,Definition: Study the speed difference of passenger cars in the road section, Expression: Variable: x2 = Speed diff,truck ,Definition: Study the speed difference of large trucks in the road section, Expression: Variable: x3 = Speed aver ,Definition: The average speed of vehicles in the study section,Expression: Variable: x4 = Vol aver,car ,Definition: Study the difference of passenger car traffic flow within the road section, Expression: Variable: x5 = Vol aver,truck ,Definition: Study the traffic flow differences of large trucks within a road section, Expression: Variable: x6 = Vol aver ,Definition: The average traffic flow in the study section, Expression: Where M represents the number of control points in the study section; m represents the sequence number of the control point; Speed m,car 、Speed m,truck Respectively represent the average speed of cars and trucks passing through the mth control point in a control cycle; Speed aver,car 、Speed aver,truck Respectively represent the average speed of all cars and trucks at M control points within a control cycle; Vol m,car 、Vol m,truck Respectively represent the average traffic flow of passenger cars and trucks passing through the mth control point in a control cycle; Vol aver,car 、Vol aver,truck Respectively represent the average traffic flow of all passenger cars and trucks at all M control points within a control period; 2) Control cycle, i.e., how often the control strategy is adjusted; the control cycle is the same as the prediction cycle described in step 2.2; 3) Speed ​​limit adjustment value: in terms of time, the adjustment value is set to 5km / h; in terms of space, the adjustment value is not greater than 10km / h; 4) Optimal speed limit value: Calculate CRI(k+1) and minimize the accident risk coefficient of the entire road to obtain the optimal speed limit value VSL of the i-th control section within the k+1 control period i (k+1): Among them, CRI i (k+1) represents the accident risk coefficient of the i-th controlled section in the k+1 control cycle; When solving the optimal speed limit value, add the following two constraints: Among them, d s Indicates the upper limit of the speed limit change between two adjacent control sections, d k Indicates the upper limit of the speed limit change between two adjacent control cycles; V VSL,i (k) and V VSL,i+1 (k) represent the optimal speed limit values ​​of the control section i and the downstream control section i+1 in the control period k respectively; V VSL,i (k) and V VSL,i (k+1) represents the optimal speed limit value of control period k and the next control period k+1 on control section i respectively; The ramp control strategy described in step 3.2, the core control parameters include the following: Based on the classic closed-loop feedback ALINEA control strategy, the impact of the upstream road on the safety performance of the entrance ramp is considered and an accident risk adjustment item is added: Where r(k) is the ramp pass rate calculated in the kth control cycle; r(k-1) is the ramp pass rate in the (k-1)th control cycle; K R is a parameter of a calibration nature; is the expected share of the mainline downstream, The value of is set to the optimal occupancy or less than the optimal occupancy; out (k-1) is the actual downstream occupancy measured in (k-1) control cycles; K s is the accident risk correction parameter; CRI k is the accident risk coefficient of the road upstream of the entrance ramp in the kth control period; CRI crit is the accident risk factor threshold; The core control parameters of the collaborative control strategy described in step 3.3 include the following: When the variable speed limit control strategy and ramp control strategy are used in coordination, the corresponding adjustment term is added to r(k) to obtain r * (k): Among them, q cap is the road capacity downstream of the entrance ramp after considering the speed limit; i (k) is the arriving traffic volume of the downstream road segment i after the speed limit in the k control period; In actual application, the ramp pass rate is controlled by setting the entrance ramp signal light. The pass rate r(k) corresponds to the green light duration g(t) in the signal control cycle, in seconds: g min <g(t)<g max (8) Among them, C is a fixed control period; r sat is the maximum passing rate of the ramp within a control cycle; g min is the shortest green light time; g max is the longest green light time; When calculating the traffic light configuration, the green light duration g(t) is adjusted as follows: Among them, q is the number of queued vehicles on the entrance ramp, and h(t) is the adjusted green light duration.

2. The method for collaborative management and control of intelligent highways based on feedback control according to claim 1 is characterized in that: The preprocessing in step 2.1 is as follows: using an intelligent algorithm that integrates the two types of perception data information and combining radar data to intelligently complete the checkpoint data and delete anomalies.

Images