An optimized control method for intersection overflow prevention signals based on connected vehicle data

Abstract

This invention provides an intersection overflow prevention signal optimization control method based on connected vehicle data, belonging to the field of traffic management technology. Addressing situations where downstream road segments at intersections experience a sudden drop in capacity due to unforeseen events, this method, building upon the signal timing of a traditional dual-ring control structure, employs seven steps: traffic flow parameter fitting, traffic flow error fitting, approach lane queue length calculation, approach lane traffic demand calculation, current exit lane remaining capacity calculation, future exit lane remaining capacity prediction, and model-predicted signal control. This achieves the goal of improving intersection efficiency while preventing overflow. This method is applicable to intersection overflow prevention when both traffic flow and bottleneck capacity are unknown, thereby enhancing the ability of intersection signal control methods to cope with overflow risks and improving intersection efficiency.

Description

Technical Field

[0001] This invention belongs to the field of traffic management technology, and in particular relates to an optimized control method for intersection overflow prevention signals based on connected vehicle data. Background Technology

[0002] Signalized intersections are crucial nodes in urban road networks, and signal timing is a vital means of managing them. When sudden events such as traffic accidents occur downstream of the intersection, reducing the road's capacity to below the required level, vehicle queues will increase and overflow into the intersection. When this sudden overflow occurs, the normal operation of the intersection will be severely affected. Not only will the overflow-related flows entering the intersection be blocked, but other non-overflow-related flows will also be obstructed.

[0003] To address overflow risks, current intersection signal timing primarily considers situations where traffic flow or bottleneck capacity is known. However, in the event of an emergency, both traffic flow and bottleneck capacity are unknown, and there is currently a lack of targeted signal control methods. This reduces the intersection's resilience to risks, and its capacity cannot be fully utilized during emergencies.

[0004] A literature review of existing technologies revealed the following main methods for intersection signal timing to address overflow:

[0005] 1. Signal timing methods under known traffic flow and bottleneck capacity. When all data is known, there are many signal timing methods for intersections, which can be categorized by type as timed control, inductive control, and adaptive control; and by scope as single-point control, arterial control, and regional coordinated control. Representative publications include "Global Road Traffic Signal Control Practice—Single-Point Timed Signal Control" and "Simulation of departure flow profile at stop lines for signal approach spillover."

[0006] 2. Signal timing methods for situations where traffic flow is known but bottleneck capacity is unknown. For sudden overflow events, such as congestion caused by accidents, the capacity of the downstream bottleneck is unknown. Therefore, overflow can only be determined by real-time exit lane queuing. Consequently, it is impossible to obtain an optimal signal timing scheme by determining the bottleneck capacity. Overflow is typically prevented by early green light termination. Representative publications include "Adaptive Signal Control for Overflow Prevention at Isolated Intersections Based on Fuzzy Control" and "An Overflow Control Method for Urban Road Intersections Based on Wide-Area Radar Detection" (Patent Application No. CN201811165527.4).

[0007] 3. Signal timing methods for situations where traffic flow is unknown but bottleneck capacity is known. This category primarily focuses on methods where bottleneck capacity is determined, such as signal control methods for conventional congestion at adjacent intersections and regional road networks. In this case, the bottleneck capacity is fixed, allowing for the calculation of the optimal signal timing scheme based on traffic flow theory to address potential overflow risks. Representative publications include "An Adaptive Signal Control Scheme to Prevent Intersection Traffic Blockage" and "A Dual-Detector-Based Anti-Overflow Control Method for Oversaturated Intersections" (Patent Application No. CN201910949272.9).

[0008] Method 1 is based on the assumption that traffic flow and bottleneck capacity are known. In this case, the intersection can adjust its signal timing using real-time and accurate data, employing a carefully designed adaptive signal timing method to prevent overflow.

[0009] Method 2 is based on the premise that traffic flow is known but bottleneck capacity is unknown. Therefore, it is impossible to obtain the optimal signal timing scheme from the bottleneck capacity. In general, overflow prevention can only be achieved by ending green lights earlier.

[0010] Method 3 is based on the premise that traffic flow is known but bottleneck capacity is unknown. Since bottleneck capacity is fixed, overflow risk can be reduced by real-time monitoring of exit lane conditions and coordinating signal timing on arterial roads.

[0011] There are numerous methods for intersection signal timing. However, existing intersection signal control primarily relies on known traffic flow or bottleneck capacity to address sudden overflow risks. For situations where both traffic flow and bottleneck capacity are unknown, there is a lack of overflow prevention signal optimization control methods. For multiple overflow-related flows, and with total travel time limited by exit lane capacity, how to reasonably optimize signal timing schemes using limited data from connected vehicles is a key issue. Summary of the Invention

[0012] The purpose of this invention is to provide an intersection overflow prevention signal optimization control method based on connected vehicle data. This method addresses situations where the downstream road segment of an intersection experiences a sudden drop in capacity due to unforeseen events. Given unknown traffic flow and bottleneck capacity, it uses traffic flow parameters and error fitting, employs a Kalman filter prediction model to determine the remaining capacity of exit lanes, predicts the remaining capacity of exit lanes over a future period, and uses model predictive control to reallocate green light times for each direction of traffic at the intersection, thereby improving traffic efficiency while preventing overflow.

[0013] The present invention provides an intersection overflow prevention signal optimization control method based on connected vehicle data, which includes the following parameters: Let i be the initial flow rate of lane i at step j; The queue length is observed at the connected vehicle of the pth vehicle in lane i with a step length of j. h represents the moment when the p-th vehicle in lane i, with a step length of j, begins to queue; d P represents the average headway in queuing conditions. i ξ represents the total number of connected vehicles queuing in lane i; i Let i be the queue length due to the increase in flow rate; Let g be the flow rate of the lane at step j in the b-th iteration; min is the minimum green light time; m is the number of steps contained in the previous cycle; Let q be the variance of the queue length due to the increase in flow rate in direction i; i,j The estimated queue length for lane i at step j; This represents the observation error value for the queue length at the entrance lane. The error value for estimating the queue length at the entrance lane at the (j-1)th step; v i,j Let g be the effect of vehicles passing through the intersection in lane i before time j on the queue length; i Let t0 be the 0-1 piecewise function of the signal timing discretization for lane i. When i = 0, it represents a red light, and when i = 1, it represents a green light. t0 is the time required for the traffic wave to travel to the end of the queue. Let be the observed queue length of the last connected vehicle in lane i at step j; k1 be the vehicle density in free flow; k2 be the vehicle density in queue; Δt be the single prediction step length; α be the error of the queue of vehicles behind the last connected vehicle; q' i.j+1 β is the predicted queue length for lane i at step j+1; a ω represents the feedback error; K represents the error correction; a For imported Kalman gain; This represents the observation error value for the inlet track. To predict the import channel error value; q' i,j ξ is the predicted queue length of lane i at step j; W is the total number of overflow directions; w is the overflow direction; ξ w w represents the queue length due to the increased flow rate; f represents the overflow direction. j Let be the traffic demand within step j; ε be the average road length occupied by bottleneck vehicles; and λ be the variance of the road length occupied by vehicles. The remaining observation capacity is where the last connected vehicle in lane i of exit road is located at step j. l′ is the exit channel observation error value at the current step size j; j K represents the remaining capacity of the predicted exit lane at step j; e For the Kalman gain of the exit channel; This represents the exit lane prediction error value; β e For the previous step's long exit channel feedback error; L e F(t') is the residual capacity function of exit lane e as a function of t'; F(t') is the total traffic demand at the bottleneck of exit lane e as a function of t'; E(t') is the traffic supply at the bottleneck of exit lane e as a function of t'; w (t') represents the total length of vehicles allowed to pass through the overflow lane before time t', which varies with t'; t' represents a future time; t represents the current time; V represents the objective function; d i The average vehicle delay for lane i; Q' i,j Let x be the flow rate of lane i at step j; T be the simulation duration; C be the cycle length; x i For lane i, saturation; g min The minimum green light time; r w The overflow lane represents the proportion of green light time allocated to it; R is the total proportion of overflow exit lanes, and its value is constant at 1; q' w,j Let w be the predicted queue length of the overflow direction lane at step j; u is the lane number in the same direction; U is the set of lane numbers in the same direction. Let NOL be the flow rate of lane i at step j in the b-th iteration; NOL is the total number of lanes flowing in the same direction.

[0014] A method for optimizing signal control at intersections to prevent overflow based on connected vehicle data, characterized by the following steps:

[0015] S1: Traffic flow parameter fitting;

[0016] S2: Traffic flow error fitting;

[0017] S3: Calculation of import to queue length;

[0018] S4: Calculation of import demand for transportation;

[0019] S5: Calculation of the remaining capacity of the inlet lane at the current moment;

[0020] S6: Future timeline exit lane remaining capacity forecast;

[0021] S7: Model Predictive Signal Control.

[0022] Further, S1 specifically involves estimating the traffic flow of the corresponding lane based on connected vehicle data, using the vehicle spacing and queuing start time of connected vehicles, as shown in Formula 1. Let the Gaussian distribution of the increase in queue length caused by traffic flow be N1~N(ξ,φ), where ξ and φ can be obtained using Formulas 2-3. Then q i,j as well as It can be calculated based on the estimation results in the previous step, as shown in Formula 4-5. Where, v i,j And t0 can be calculated from the traffic wave model, as shown in Formula 6-7.

[0023]

[0024]

[0025]

[0026] q i,j =q′ i.j-1 -v i,j +ξ i (4)

[0027]

[0028]

[0029]

[0030] Furthermore, S2 specifically refers to the following: Since the last connected vehicle is always locked during queue length prediction, there are two possibilities compared to the connected vehicle locked in the previous step: the connected vehicle may change or remain unchanged. If the connected vehicle remains unchanged, only the error caused by the arrival flow within the step length needs to be considered; if the connected vehicle changes, the error α of the queued vehicles behind the last connected vehicle also needs to be considered. The error generated by the queued vehicles behind the last connected vehicle is regarded as a Gaussian distribution N2~N(0,α), as shown in Formula 8. α can be calculated from the actual queue length and the observed queue length in the previous signal period.

[0031] In addition to the error caused by the vehicles queuing behind the last connected vehicle, the feedback error β between the observed and measured values ​​also needs to be considered. a Feedback error refers to the difference between the queue length of the last connected vehicle in the queue and the estimated queue length, as shown in Formula 9. Furthermore, since the error of a Gaussian distribution can be positive or negative, but this paper always locks the last connected vehicle in the queue length prediction, the final predicted queue length can only be further back than the position of the last connected vehicle. To minimize the impact of negative error, ω is used... i The prediction results are corrected as shown in Formula 10. ω is obtained from the average interval between connected vehicles queuing in the same lane.

[0032]

[0033]

[0034]

[0035] Further, S3 specifically involves: calculating the Kalman gain value based on the aforementioned traffic flow parameters and error fitting results, as shown in Formula 11. After obtaining the Kalman gain, the error correlation value of the current step size can be updated based on the Kalman gain, as shown in Formula 12. Finally, the queue length calculation result for the approach lane is obtained based on the Kalman filter prediction model, as shown in Formula 13.

[0036] In congested traffic conditions, drivers in the same direction of traffic tend to queue in the lane with the shortest queue length. Therefore, a dynamic balance mechanism between queue length and traffic flow is established. With the total flow remaining constant, based on the queue length obtained in the previous step, Formula 14 is used to determine whether the queue length distribution is reasonable. If it is not reasonable, the multivariate inverse proportional equation of Formula 15 is used to adjust the traffic flow distribution of each lane in the same direction, the corresponding parameters are refitted, and the Kalman filter prediction model of the entrance lane (Formula 2-13) is used again for calculation, ultimately making the predicted queue length approximately equal.

[0037]

[0038]

[0039] q′ i,j =q i,j +K α β α +ω (13)

[0040] q′ u1,j -q′ u2,j ≤5h d ,u1,u2∈U (14)

[0041]

[0042] Further, S4 specifically involves calculating the number of vehicles passing through the intersection in the overflow direction within this step length, based on the queue length of the approach lane obtained in step 3. By comparing the maximum number of vehicles that can pass through the intersection with the actual traffic demand, the final traffic demand for the exit lane is obtained, as shown in Formula 16.

[0043]

[0044] Furthermore, S5 specifically refers to the error of the Kalman filter prediction model for the exit lane, which consists of the error of the vehicles queuing behind the last connected vehicle and the prediction error of the bottleneck capacity. The error of the vehicles queuing behind the last connected vehicle is the same as in Formula 8. Assuming the bottleneck capacity is unknown and fluctuates, it will affect the prediction result of the remaining capacity of the exit lane. Assuming the current position is at the j-th step, the total length of vehicles passing through the bottleneck within this step is considered a Gaussian distribution N3~N(ε,λ).

[0045] Then l j and error correlation values The result can be calculated based on the estimation results of the previous step, as shown in Equations 19 and 20. The Kalman filter gain of the exit lane, the error correlation value of the current step, and the feedback error are shown in Equations 21-23. The final queue length of the inlet lane is shown in Equation 24.

[0046]

[0047]

[0048] l j =l′ j-1 -f j +ε (19)

[0049]

[0050]

[0051]

[0052]

[0053] l′ j =l j +K e β e -ω (24).

[0054] Furthermore, S6 specifically refers to: based on the obtained real-time exit lane remaining capacity, the exit lane remaining capacity at any future time can be estimated, as shown in formulas 25-28. Formula 26 compares the maximum number of vehicles that can pass through during the green light time with the estimated traffic demand to obtain the actual traffic demand of each overflow direction passing through the intersection exit lane; Formula 27 calculates the total traffic demand of the exit lane; Formula 28 estimates the traffic supply at the exit lane bottleneck in the future period based on the estimated bottleneck capacity.

[0055] L e (t′)=F(t′)-E(t′) (25)

[0056]

[0057]

[0058]

[0059] Furthermore, S7 specifically involves obtaining a signal control scheme by solving the following model. The signal controller is based on, as follows: Figure 2 The traditional dual-loop control structure shown uses Δt as the step size for signal timing. Each control time domain is one cycle. The new timing scheme inherits the existing time of the previous timing scheme and the preceding phase scheme.

[0060] Formula 29 is the objective function of the model, and Formulas 32-34 are the model constraints. The objective function is to minimize the average vehicle delay. Formula 32 is the cycle duration constraint; Formula 33 is the minimum green light time constraint. Formula 34 is the set of green light time constraints for overflow flows related to the remaining capacity of the exit lane. This constraint uses t' as the independent variable, and the value of t' ranges from one cycle duration. Therefore, this constraint is a set of constraints for multiple t' values ​​in a single step length. This constraint set ensures that the available capacity of overflow flows at all times does not exceed the remaining capacity of the exit lane, while distributing the limited green light time to each flow direction according to the queue length of the overflow flows. The weight coefficient of the green light time for each flow direction is inversely proportional to the queue length.

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069] The key to the intersection overflow prevention signal optimization control method based on connected vehicle data lies in determining the remaining capacity of exit lanes and predicting potential future overflow risks when data is limited. The calculation process in this invention fully considers the remaining capacity of exit lanes under intersection overflow conditions at the current moment and over a future period. Through traffic flow parameter fitting and model predictive control, the intersection signal timing scheme is iteratively optimized in real time, thereby improving the rationality and robustness of the intersection signal timing and increasing the intersection's traffic efficiency.

[0070] Compared with the prior art, the beneficial effects of the present invention are mainly reflected in:

[0071] 1. This invention provides an optimized control method for intersection overflow prevention signals based on connected vehicle data, which can improve the ability of intersection signal control methods to cope with overflow risks.

[0072] 2. The method of the present invention is applicable to situations where both traffic flow and bottleneck capacity are unknown.

[0073] 3. The method of the present invention takes into account the remaining capacity of the exit lane at the current moment and in the future period, and realizes the real-time allocation of the green light duration for each flow direction through model predictive control. Attached Figure Description

[0074] Figure 1 This is the control flowchart of the present invention;

[0075] Figure 2 This is a phase sequence diagram of a dual-ring signal implemented in this invention;

[0076] Figure 3 The road geometry and location of the emergency in Embodiment 1 of the present invention;

[0077] Figure 4 This refers to the traffic demand in Embodiment 1 of the present invention. Detailed Implementation

[0078] The following will describe in more detail, with reference to the schematic diagram, a method for optimizing the control of intersection overflow prevention signals based on connected vehicle data, according to the present invention. The diagram illustrates a preferred embodiment of the present invention. It should be understood that those skilled in the art can modify the present invention described herein while still achieving the advantageous effects of the present invention. Therefore, the following description should be understood as being of general knowledge to those skilled in the art and is not intended to limit the present invention.

[0079] like Figure 1 As shown, an intersection overflow prevention signal optimization control method based on connected vehicle data includes the following steps:

[0080] S1: Traffic flow parameter fitting

[0081] Based on connected vehicle data, the flow direction of the corresponding lane is estimated by using the vehicle spacing and the start time of queuing of connected vehicles, as expressed by Formula 1:

[0082]

[0083] In Formula 1, Let i be the initial flow rate of lane i at step j; The queue length is observed at the connected vehicle of the pth vehicle in lane i with a step length of j. h represents the moment when the p-th vehicle in lane i, with a step length of j, begins to queue; d P represents the average headway in queuing conditions. i The total number of connected vehicles queuing in lane i;

[0084] Assume a Gaussian distribution for the increase in queue length caused by flow rate. ξ is the expected value, calculated using Formula 2, which is expressed as:

[0085]

[0086] In Formula 2, ξ i Let i be the queue length due to the increase in flow rate; Let g be the flow rate of the lane at step j in the b-th iteration; min is the minimum green light time; m is the number of steps contained in the previous cycle;

[0087] The variance is calculated using Formula 3, which is expressed as:

[0088]

[0089] In Formula 3, φ i Let i be the variance of the queue length due to the increase in flow rate.

[0090] q i,jThe estimated queue length for lane i at step j is calculated using Formula 4, which is expressed as:

[0091] q i,j =q i.j-1 -v i,j +ξ i

[0092] The error value for the observation of the queue length at the entrance lane is calculated using Formula 5, which is expressed as:

[0093]

[0094] In Formula 5, The error value for estimating the queue length of the inlet lane at the (j-1)th step;

[0095] In Formula 4, v i,j The impact of vehicles passing through the intersection in lane i at time steps j on the queue length is calculated using Formula 6, which is expressed as:

[0096]

[0097] In Formula 6, g i Let be the 0-1 piecewise function of the signal timing for lane i, discretized. When i = 0, it represents a red light; when i = 1, it represents a green light. t0 is the time required for the traffic wave to reach the end of the queue, calculated using Formula 7, which is expressed as:

[0098]

[0099] In Formula 7, Δt is the observed queue length of the last connected vehicle in lane i at step j; k1 is the vehicle density in free flow; k2 is the vehicle density in queue; Δt is the single prediction step length.

[0100] S2: Traffic Flow Error Fitting

[0101] Since the queue length prediction always locks onto the last connected vehicle, there are two possibilities: the connected vehicle may change or remain unchanged compared to the one locked in the previous step. If the connected vehicle remains unchanged, only the error caused by the arrival flow within the step length needs to be considered; if the connected vehicle changes, the error of the queued vehicles behind the last connected vehicle also needs to be considered. Assuming the error caused by the queued vehicles behind the last connected vehicle is a Gaussian distribution N²~N(0,α), the error α of the queued vehicles behind the last connected vehicle is calculated by the actual queue length and the observed queue length in the previous signal period. The calculation formula is expressed by Formula 8:

[0102]

[0103] In Formula 8, q' i.j+1 The predicted queue length for lane i at step j+1; traffic flow error fitting also needs to consider the feedback error β between observations and measurements. a Feedback error β a The difference between the queue length of the last connected vehicle in the queue and the estimated queue length is expressed by Formula Nine as follows:

[0104]

[0105] To avoid the impact of negative errors, the prediction results are corrected using the error correction mean ω, which is expressed by Formula 10:

[0106]

[0107] S3: Calculation of import to queue length

[0108] Includes the following steps:

[0109] S31: Based on the traffic flow parameters and error fitting results obtained in S2, the Kalman gain value is calculated using Formula 11, which is expressed as follows:

[0110]

[0111] In Formula 11, K a For imported Kalman gain; α represents the observation error value of the entrance lane; α represents the error of the vehicles queuing behind the last connected vehicle.

[0112] S32: Update the error correlation value of the current step size based on the obtained Kalman gain value, as expressed by Equation Twelve:

[0113]

[0114] In formula twelve, To predict the import channel error value;

[0115] S33: Based on the Kalman filter prediction model, the queue length of the entrance lane is calculated using Formula Thirteen, which is expressed as:

[0116] q i,j =q i,j +K a β a +ω

[0117] In Formula Thirteen, q' i,j Predicted queue length for lane i at step j

[0118] When roads are congested, drivers in the same direction of traffic tend to queue in the lane with the shortest queue length. A dynamic balance mechanism between queue length and traffic flow is established. When the total traffic flow is constant, the queue length obtained in the previous step is used, and Formula 14 is used to determine whether the queue distribution is reasonable. If the result is unreasonable, the traffic flow distribution of each lane in the same direction is adjusted using the multivariate inverse proportional equation expressed in Formula 15. The corresponding parameters are then refitted, and the Kalman filter prediction model for the entrance lane (Formulas 2 to 13) is used again for calculation, ultimately making the predicted queue lengths approximately equal. Formula 14 is expressed as:

[0119]

[0120] In Formula 14, u represents the lane number in the same direction of flow; U represents the set of lane numbers in the same direction of flow; Formula 15 is expressed as:

[0121]

[0122] In Formula 15, Let NOL be the flow rate of lane i at step j in the b-th iteration; NOL is the total number of lanes flowing in the same direction.

[0123] S4: Import to Transportation Demand Calculation

[0124] Based on the queuing length of the approach lane obtained from S3, the number of vehicles passing through the intersection in the overflow direction within this step length is calculated. By comparing the maximum number of vehicles passing through the intersection with the actual traffic demand, the final traffic demand of the exit lane is obtained, which is expressed by Formula Sixteen:

[0125]

[0126] In Formula Sixteen, W represents the total number of overflow directions; w represents the overflow direction; ξ w w represents the queue length due to the increased flow rate; f represents the overflow direction. j Let J represent the traffic demand within a step size of j.

[0127] S5: Calculation of remaining capacity of the approach lane at the current moment

[0128] The error of the Kalman filter prediction model based on the exit lane consists of the error of the vehicles queuing behind the last connected vehicle and the prediction error of the bottleneck capacity. Assuming the bottleneck capacity is unknown and fluctuates, the prediction result of the remaining capacity of the exit lane is affected by the bottleneck capacity. Assuming the current position is at the j-th step, the total length of vehicles passing through the bottleneck within that step follows a Gaussian distribution N3~N(ε,λ). The average road length ε occupied by vehicles passing through the bottleneck is expressed by Formula 17:

[0129]

[0130] The variance λ of the road length occupied by bottleneck vehicles is expressed by Formula 18 as follows:

[0131]

[0132] In formulas seventeen and eighteen, The remaining observation capacity is where the last connected vehicle in lane i of exit road is located at step j.

[0133] Based on the estimation results from the previous step, the estimated remaining capacity l of the exit channel at the current step j is... j The exit channel observation error value at the current step size j is obtained by formula nineteen. The Kalman filter gain of the exit channel is calculated using Formula 20, expressed by Formula 21; the error correlation value of the current step size j is expressed by Formula 22; the feedback error of the current step size j is expressed by Formula 23; and Formula 19 is expressed as follows:

[0134] l j =l j-1 -f j +ε

[0135] In Formula 19, l′ j Let J be the predicted remaining capacity of the exit channel after step j; Formula 20 is expressed as:

[0136]

[0137] Formula 21 is expressed as:

[0138]

[0139] In formula twenty-one, K e The Kalman gain is the output channel gain; Formula 22 is expressed as:

[0140]

[0141] In formula twenty-two, This represents the exit lane prediction error value; Formula 23 is expressed as:

[0142]

[0143] In formula twenty-three, β e The feedback error of the previous step's exit channel; Formula 24 is expressed as:

[0144] l′ j =l j +K e β e -ω.

[0145] S6: Future Exit Lane Remaining Capacity Forecast

[0146] Based on the real-time remaining capacity of the exit lanes obtained from S5, the remaining capacity of the exit lanes at any future time is estimated, as expressed by Formula 25; the actual traffic demand of each overflow direction passing through the intersection's exit lanes is obtained by comparing the maximum number of vehicles passing through during the green light period with the estimated traffic demand, as expressed by Formula 26; the total traffic demand of the exit lanes is expressed by Formula 27; the traffic supply at the bottleneck of the exit lanes in the future period is estimated based on the estimated bottleneck capacity, as expressed by Formula 28; Formula 25 is expressed as:

[0147] L e (t')=F(t')-E(t')

[0148] In formula twenty-five, L e F(t') is the residual capacity function of exit lane e as a function of t'; F(t') is the total traffic demand at the bottleneck of exit lane as a function of t'; E(t') is the traffic supply at the bottleneck of exit lane as a function of t'; Formula 26 is expressed as:

[0149]

[0150] In formula twenty-six, l w (t') represents the total length of vehicles allowed to pass under the green light before time t' on the overflow lane, which varies with t'; t' represents a future time; t represents the current time; Formula 27 is expressed as:

[0151]

[0152] Formula 28 is expressed as:

[0153]

[0154] S7: Model Predictive Signal Control

[0155] The model predicts signal control as follows: The signal controller performs signal timing in steps of Δt, with each control time domain lasting one cycle. The new timing scheme inherits the existing moments and preceding phase scheme of the previous timing scheme. The signal control scheme is obtained by solving formulas 29 to 34. In formulas 29 to 34, formula 29 is the objective function of the model, expressed as:

[0156]

[0157] In Formula 29, V is the objective function; d i The average vehicle delay for lane i; Q' i,j Let i be the flow rate at step j in lane i; Equations 30 to 34 are model constraints, with Equation 30 expressed as:

[0158]

[0159] In Formula 30, T is the simulation duration; C is the cycle length; Formula 31 is expressed as:

[0160]

[0161] In formula 31, x i Let i be the saturation level of lane i; Formula 32 expresses this as:

[0162]

[0163] In formula 32, g min The minimum green light time; Formula 33 is expressed as:

[0164]

[0165] In formula 33, r w R is the proportion of green light time allocated to the overflow lane; R is the total proportion of overflow exit lanes, and its value is constant at 1; Formula 34 is expressed as:

[0166]

[0167] In formula thirty-four, q' w,j The predicted queue length for overflow flow towards lane at step j+1 is given by Formula 35:

[0168]

[0169] In Formula 34, the overflow flow direction green light time constraint set uses t' as the independent variable, and the value range of t' is one cycle duration. While ensuring that the release capacity of the overflow flow direction does not exceed the remaining capacity of the exit lane at all times, the overflow flow direction green light time is allocated to each flow direction according to the queue length of the overflow flow direction. The weight coefficient of the green light time of each flow direction is inversely proportional to the queue length.

[0170] Example

[0171] The following is combined with Figure 3 and Figure 4 The embodiments shown further illustrate the present invention.

[0172] The embodiments of this invention verify the usability and implementation benefits of the invention based on Vissim simulation. The road geometry conditions of the embodiments are as follows: Figure 3 As shown, traffic flow is as follows Figure 4As shown. The cycle duration C = 120s, the saturation flow rate s = 1800 pcu / h / ln, the vehicle speed in free flow at the exit lane v0 = 50 km / h / ln, and the minimum green light time g for each flow direction. min =5s, average headway h d =8m, single step Δt = 5s. At simulation 0s, the sudden event occurs 200m from the intersection in exit lane 1, reducing its capacity to 300 pcu / h / ln. The impact of the sudden event lasts until simulation 2600s. After 2600s, the sudden event is removed, and exit lane 1 returns to normal capacity. The entire study analysis lasts for 3600s, and the penetration rate of connected vehicles in the study is 50%.

[0173] The overall flowchart of the signal timing scheme for this intersection using the method of the present invention is as follows: Figure 1 As shown, the specific process is briefly described below:

[0174] Step 1: Traffic flow parameter fitting. Based on real-time acquired connected vehicle data, the initial flow rate and queue length of each lane at the current time are calculated using formulas 1-7.

[0175] Step 2: Traffic flow error fitting. Use formula 8-10 to calculate the error of vehicles queuing behind the last connected vehicle, the feedback error of the entrance lane, and the mean of the error correction.

[0176] Step 3: Calculation of queue length at the approach lane. The calculated queue length of the approach lane, obtained using the Kalman filter prediction model according to formulas 11-13, is verified using formula 14. If it does not meet the requirements, the traffic flow is adjusted according to formula 15 until formula 14 is satisfied. The calculation results are shown in Table 1.

[0177]

[0178] Table 1

[0179] Step 4: Calculate the traffic demand for the exit lane. Use Formula 16 to calculate the final traffic demand for the exit lane.

[0180] Step 5: Calculate the remaining capacity of the exit lane at the current moment. The remaining capacity of the exit lane at the current moment is calculated using the Kalman filter prediction model in Equation 17-24.

[0181] Step 6: Predict the remaining capacity of the exit lane at future times. The remaining capacity of the exit lane at any future time is estimated using formulas 25-28. Examples of the calculation results are shown in Table 2.

[0182]

[0183]

[0184] Table 2

[0185] Step 7: Model Predictive Signal Control. The signal control scheme is obtained by solving the model in equations 29-35. The signal controller is based on a traditional dual-loop control structure, and the final signal timing scheme is shown in Table 3.

[0186] overflow direction <![CDATA[Previous step timing plan g i,j-1 (s)]]> <![CDATA[Current step signal timing plan g i,j (s)]]> Turn left at entrance lane 1 20 20 Entrance Lane 1 Straight 31 33 Turn left at entrance lane 2 5 0 Entrance Lane 2 Straight 23 35 Turn left at entrance lane 3 39 30 Entrance Lane 3 Straight 10 8 Turn left at entrance lane 4 27 22 Entrance Lane 4 Straight 26 26 Turn right at entrance lane 4 0 5

[0187] Table 3

[0188] Step 8: Repeat steps 1 to 7 in increments of Δt = 5 seconds until the analysis period ends. Table 4 shows a comparison between this invention and the traditional method (adaptive control without considering the remaining capacity limit of the exit lane). It can be seen that this invention, compared to the traditional method, can prevent overflow and significantly improve intersection throughput. Specifically, overflow-related flow delays are reduced by 27.90%, total delays are reduced by 46.53%, and maximum queue length is reduced by 45.75%.

[0189]

[0190] Table 4

[0191] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.

Claims

1. A method for optimizing signal control at intersections to prevent overflow based on connected vehicle data, characterized in that, Includes the following steps: S1: Traffic flow parameter fitting; S2: Traffic flow error fitting; S3: Calculation of queue length at the entrance lane; S4: Traffic demand calculation for exit routes; S5: Calculation of the remaining capacity of the exit lane at the current moment; S6: Future timeline exit lane remaining capacity forecast; S7: Model Predictive Signal Control; Specifically, S1 involves estimating the flow direction of the corresponding lane based on connected vehicle data, using the vehicle spacing and queuing start time of connected vehicles, as expressed by Formula 1: ； In Formula 1, for Lane 1 Initial flow rate value for the step size; for Lane 1 Step length The queue length is observed at the location of the connected vehicles; for Lane 1 Step length The moment when the queue of vehicles begins to form; This represents the average headway between vehicles in a queue. for The total number of connected vehicles queuing in the lane; Assume a Gaussian distribution for the increase in queue length caused by flow rate. , To achieve the desired result, calculations are performed using Formula 2, which is expressed as follows: ； In the second formula, for The flow direction is affected by the increased queue length due to the increased flow rate; For the first Next iteration lane Step length flow rate; Minimum green light time; This represents the number of step sizes contained in the previous cycle; The variance is calculated using Formula 3, which is expressed as: ； In formula three, for The variance of queue length due to increased flow rate; for Lane 1 The estimated queue length for the step size is calculated using Formula 4, which is expressed as follows: ； The error value for the observation of the queue length at the entrance lane is calculated using Formula 5, which is expressed as: ； In formula five, For the first Error value for estimated queue length at the entrance lane; In formula four, Lane 1 The step size represents the impact of vehicles passing through the intersection at a previous time step on the queue length, calculated using Formula Six, which is expressed as follows: ； In formula six, for The lane signal timing discretization of the 0-1 piecewise function, when When, it indicates a red light, when When the light is green, it indicates a green light. The time required for the traffic wave to travel to the end of the queue is calculated using Formula 7, which is expressed as follows: ； In Formula 7, for Lane 1 The observation queue length of the last connected vehicle in the step; Vehicle density under free-flow conditions; Vehicle density under queuing conditions; This refers to the step size for a single prediction. In step S2, the traffic flow error fitting needs to consider the error of vehicles queuing behind the last connected vehicle. The error generated by vehicles queuing behind the last connected vehicle is assumed to be Gaussian distributed. The error of the queue of vehicles behind the last connected vehicle. It is calculated by comparing the actual queue length and the observed queue length in the previous signal cycle; Specifically, S4 is: based on the queue length of the approach lane obtained in S3, calculate the number of vehicles passing through the overflow direction of the intersection within this step length, and obtain the final traffic demand of the exit lane by comparing the maximum number of vehicles passing through the intersection with the actual traffic demand. The error of the Kalman filter prediction model based on the exit lane consists of the error of the vehicles queuing behind the last connected vehicle and the prediction error of the bottleneck capacity. Assuming the bottleneck capacity is unknown and fluctuates, the prediction result of the remaining capacity of the exit lane is affected by the bottleneck capacity. Let's assume the current vehicle is in the... If the number of steps is given, then the total length of vehicles passing through the bottleneck within that step is a Gaussian distribution. ; Based on the estimation results from the previous step size, the current step size... Estimated remaining capacity of the exit channel ; Specifically, S6 involves: estimating the remaining capacity of the exit lanes at any future time based on the real-time remaining capacity of the exit lanes obtained in S5; comparing the maximum number of vehicles passing through during the green light period with the estimated traffic demand to obtain the actual traffic demand of each overflow direction passing through the intersection's exit lanes; and estimating the traffic supply at the bottleneck of the exit lanes in the future based on the estimated bottleneck capacity. In S7, the model prediction signal control specifically refers to: the signal controller using... Signal timing is performed for the step size. Each control time domain is one cycle. The new timing scheme inherits the existing time of the previous timing scheme and the preceding phase scheme.

2. The intersection overflow prevention signal optimization control method based on connected vehicle data according to claim 1, characterized in that, In S2, the error The calculation formula is expressed by Formula 8 as follows: ； In Formula 8, for Lane 1 The actual queue length is obtained after predicting the step length; traffic flow error fitting also needs to consider the feedback error between the observed and measured values. Feedback error The difference between the queue length of the last connected vehicle in the queue and the estimated queue length is expressed by Formula Nine as follows: ； To avoid the impact of negative errors, an error correction mean is used. The prediction results are corrected, as expressed by Formula 10: 。 3. The intersection overflow prevention signal optimization control method based on connected vehicle data according to claim 1, characterized in that, S3 specifically includes the following steps: S31: Based on the traffic flow parameters and error fitting results obtained in S2, the Kalman gain value is calculated using Formula 11, which is expressed as follows: ； In formula eleven, For imported Kalman gain; This represents the observation error value for the inlet track. The error is due to the number of vehicles queuing behind the last connected vehicle. S32: Update the error correlation value of the current step size based on the obtained Kalman gain value, as expressed by Equation Twelve: ； In formula 12, To predict the import channel error value; S33: Based on the Kalman filter prediction model, the queue length of the entrance lane is calculated using Formula Thirteen, which is expressed as follows: ； In formula thirteen, for Lane 1 Predicted queue length based on step size.

4. The intersection overflow prevention signal optimization control method based on connected vehicle data according to claim 1, characterized in that, In S4, the final exit road traffic demand is expressed by Formula Sixteen as follows: ； In formula sixteen, This represents the total number of overflow flows; The overflow direction; for Overflow flow direction is determined by the increased queue length due to increased flow rate; for Traffic demand within a step length.

5. The intersection overflow prevention signal optimization control method based on connected vehicle data according to claim 1, characterized in that, In S5, the average road length occupied by vehicles passing through the bottleneck is... This can be expressed by Formula 17: ； Variance of road length occupied by bottleneck vehicles This can be expressed by Formula 18 as follows: ； In formulas seventeen and eighteen, For the exit channel Lane 1 The remaining observation capacity where the last connected vehicle in the step is located; Current step size Estimated remaining capacity of the exit channel The current step size is calculated using Formula 19. Exit channel observation error value The Kalman filter gain of the exit channel, calculated using Equation 20, is expressed by Equation 21, with the current step size... The error correlation value is expressed by formula twenty-two, the current step size. The feedback error is represented by Formula 23, and Formula 19 is expressed as follows: ； In formula nineteen, For the first The remaining capacity of the exit channel after the step size prediction; the formula twenty is expressed as: ； Formula 21 is expressed as follows: ； In formula twenty-one, The Kalman gain is the output channel gain; Formula 22 is expressed as: ； In formula twenty-two, The exit lane prediction error value; Formula 23 is expressed as: ； In formula twenty-three, The previous step's long exit channel feedback error; Formula 24 is expressed as: 。 6. The intersection overflow prevention signal optimization control method based on connected vehicle data according to claim 1, characterized in that, In S6, the remaining capacity of the exit lane at any future time is represented by Formula 25; the actual traffic demand of each overflow direction passing through the intersection's exit lane is represented by Formula 26; the total traffic demand of the exit lane is represented by Formula 27; and the traffic supply at the bottleneck of the exit lane in the future period is represented by Formula 28. Formula 25 is expressed as follows: ； In formula twenty-five, For follow changing The remaining capacity function of the exit channel; For follow Changes in total traffic demand at the bottleneck of the exit route; For follow The changing traffic supply at the bottleneck of the exit route; Formula 26 is expressed as: ； In formula twenty-six, For follow changing overflow lane The total length of vehicles allowed to pass under the green light before the designated time. For some point in the future; For the current moment; Formula 27 is expressed as: ； Formula 28 is expressed as follows: 。 7. The intersection overflow prevention signal optimization control method based on connected vehicle data according to claim 1, characterized in that, In step S7, the signal control scheme is obtained by solving formulas 29 to 34. Formula 29 is the objective function of the model, expressed as: ； In formula twenty-nine, The objective function is... for Average vehicle delay per lane; for Lane 1 The flow rate of the step size; Formulas 30 to 34 are model constraints, and Formula 30 is expressed as: ； In formula thirty, For simulation duration; The period duration; Formula 31 is expressed as: ； In formula thirty-one, for Lane saturation; Formula 32 is expressed as: ； In formula thirty-two, The minimum green light time; Formula 33 is expressed as: ； In formula thirty-three, for The proportion of green light time allocated to overflow lanes; The total proportionality coefficient of the overflow outlet channel is constant at 1; Formula 34 is expressed as: ； In formula thirty-four, for Overflow flows into the lane The predicted queue length based on the step size; Formula 35 is expressed as: ； In Formula 34, the overflow flow direction green light time constraint is a concentrated constraint. As the independent variable, The value range is one cycle duration; the overflow flow direction green light time constraint set ensures that the release capacity of the overflow flow direction does not exceed the remaining capacity of the exit lane at all times, while distributing the green light time to each flow direction according to the queue length of the overflow flow direction, and the weight coefficient of the green light time of each flow direction is inversely proportional to the queue length.

8. The intersection overflow prevention signal optimization control method based on connected vehicle data according to claim 3, characterized in that, In step S3, based on the principle that when roads are congested, drivers of traffic flowing in the same direction tend to queue in the lane with the shortest queue length, a dynamic balance mechanism between queue length and traffic flow is established. When the total traffic flow is constant, the queue length obtained in the previous step is used, and formula fourteen is used to determine whether the queue length distribution is reasonable. If the result is unreasonable, the traffic flow distribution of each lane in the same direction is adjusted by using the multivariate inverse proportional equation expressed in formula fifteen, and the corresponding parameters are refitted and the entrance lane Kalman filter prediction model (i.e., formulas two to thirteen) is reused for calculation, ultimately making the predicted queue lengths approximately equal. Formula fourteen is expressed as: , ； In formula fourteen, Lane numbers for lanes flowing in the same direction; This refers to the set of lane numbers for lanes flowing in the same direction; Formula 15 is expressed as: ； In formula fifteen, For the first Second iteration Lane 1 Step length flow rate; This represents the total number of lanes flowing in the same direction.

Images