JP2003067883A - Method for predicting traffic situation in wide road traffic space - Google Patents

Abstract

PROBLEM TO BE SOLVED: To predict a traffic situation of traffic space being a control object. SOLUTION: This method for predicting a traffic situation in wide road traffic space being a control object is provided with a first step for fetching traffic data in the traffic space, a second step for determining the positions of the whole vehicles existing in the traffic space after a sufficiently short time by checking traffic signal instruction information, a straightly advancing ratio, a right turning ratio and a left turning ratio defined foreseeingly, and space on straightly advancing road, space on a road to a left turn and space on a road to a right turn information, and a third step for controlling the second step so as to be repeated for a desired time. For this reason, a traffic situation of the whole traffic routs included in the traffic space of the object is predicted with high accuracy to be able to configure a control idea that eliminates and reduce congestion over the whole area.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a traffic condition prediction method for providing an optimum pattern of instruction patterns for a huge number of traffic signals installed in a road traffic space forming a complicated network.

[0002]

2. Description of the Related Art Traffic jams on roads occur in various places, but as a frequent case, the traffic processing amount is low such as from two lanes to one lane, or even if the number of lanes is the same, This is the case when it becomes narrow. In addition, a place with a lot of curves such as a mountain road, a steep slope with a sharp curve, and a road with a steep slope also cause a reduction in the traffic processing amount.

The traffic congestion caused by these causes the road itself, but another major factor is the intersection. An intersection is a place where roads intersect, and if an intersection with a large volume of traffic is left unattended, traffic accidents will frequently occur. In reality, about 60% of traffic accidents occur at intersections. At intersections with a certain amount of traffic, traffic lights are installed for traffic control. When the traffic space is regarded as one network, the traffic light plays a role of a control effector, which allocates the traffic volume on the network to the transit time to the route connected to each intersection.

The optimization of signal control, which is the object of the present invention, is to eliminate waste of green time, and to cope with traffic conditions without time delay to maximize the allowable range of the target traffic space. That is. Achieving signal optimization is more economical in that it essentially solves the problem than, for example, widening the road where congestion occurs due to symptomatic treatment or constructing a new bypass tunnel. It is also excellent in actual effect.

At present, not only in Japan but also in various places around the world, the number of automobiles is rapidly increasing. However, the expansion of the traffic capacity cannot keep up with this, resulting in serious road conditions. Therefore, if optimal signal control is realized over a wide area, the problem can be expected to be greatly improved, and there is a strong demand for such advanced control. On the other hand, various methods and devices have been proposed and put to practical use in the past, but the present situation is that a sufficient effect is not obtained in terms of optimality over a wide area.

Signal control methods are roughly classified into two types: (1) a control method for optimizing the traffic signal operation at one intersection (2) a control method for considering the target area as a network and optimizing the whole area. . Clearly, the method of (2) includes the method of (1), and if the method of (2) can be achieved, the method of (1) will also be achieved at the same time, but in reality, the method of (2) Since the technical level of the method is not sufficient, the set of methods (1) often gives better results than the method (2).

As a recent technique relating to the method (1), there is an open patent "Signal Controller" (Japanese Patent Laid-Open No. 28770/1990).
0) can be given. In this method, the traffic situation at the target intersection is captured as sensor information,
By referring to these temporal transitions, the wave information of the day of the week is further added, and the pattern of the traffic signal is determined by fuzzy reasoning.

The method (2) which is prominent is "SCOO" which is used for traffic control in the central area of London, England.
It is an online wide area signal control method called "T". This method is described in the document "The SCOOT on-line traffi
c signal optimization technique ”(TRAFFIC ENGINEER
ING & CONTROL magazine, April 1982) or "The devel
opment of Urban Traffic Control in London: fixed-t
For more information on "ime and 'SCOOT'" (the same magazine, October 1981). In this method, the traffic information transition at the intersection where a plurality of related traffic lights are installed is used as sensor information, and the time when the signal pattern makes one round, that is, a cycle, a split indicating the time of blue or red in the cycle, or the intersection This is a method for optimally determining parameters such as offset, which is the phase difference of the cycle determined by the physical positional relationship and the group velocity of the traffic flow, in real time. There are many other methods, but the problems are currently summarized above.

[0009]

The method of (1) above is
It is a control for each traffic light and is effective for a specific intersection, but since, for example, the situation of an adjacent intersection is not considered, when looking at the entire control target area, optimization is not achieved, Instead, the problem will spread to intersections other than the target intersection.

The above method (2) can be understood from the conventional concept of signal control, in order to dynamically determine possible parameters for a wide area while considering the degree of influence between mutual intersections. It was thought to be effective. However, in reality, many problems to be improved are included. According to this method, firstly, it is the direction or system in which the traffic volume is large that is emphasized in determining each parameter, and the directions and systems in which the traffic volume is not so large have not been sufficiently examined. This is also clear by defining an offset that is a parameter. This is because the offset is defined by the flow and the positional relationship of a certain series of intersections, so that the route that does not correspond within the target area will have no regularity. In this way, by setting the offset, the traffic flow is locally improved, but it does not give good results when viewed in the wide traffic space as a whole. Furthermore, since the transition of traffic volume is also based on past results, good control is achieved when reproducibility is high, but good results are not obtained when this is not the case.

It is an object of the present invention to accurately predict traffic conditions for all traffic routes included in a traffic space to be controlled.

[0012]

The present invention is a method for predicting the traffic situation in a wide area road traffic space to be controlled, wherein the traffic situation at that time is input from a traffic sensor in the traffic space, Provided is a method of predicting a traffic situation in a wide area road traffic space, which predicts the behavior of all vehicles existing in the traffic space for a sufficiently short time and repeats the predictive calculation until a predetermined future time is reached.

The present invention is also a method of predicting a traffic situation in a wide area road traffic space to be controlled, which is present in the first step of capturing traffic data in the traffic space and in the traffic space. For all vehicles, the position after a sufficiently small time is determined from the traffic signal instruction information and straight ahead ratio, right turn ratio, left turn ratio and straight road vacancy, left turn road vacancy, right turn road vacancy information Provided is a method of predicting a traffic situation in a wide area road traffic space, which comprises the second step and a third step of controlling the second step to be repeated for a desired time.

[0014]

According to the method for predicting the traffic condition in the wide area traffic space of the present invention, the traffic condition in the traffic space to be controlled can be accurately predicted. As a result, in a wide-area transportation network, optimal control can be realized so as to eliminate traffic congestion that occurs when the traffic volume approaches the allowable traffic capacity. Optimum here means not only to improve the serviceability of vehicles and drivers existing in a specific area or system, but also to obtain control effects that improve serviceability, safety, and economic efficiency over the entire area. It means that.

The signal control method in the traffic system according to the present invention uses the operation pattern of the traffic signal installed in the target area at that time (often referred to as a signal ladder pattern) as a starting point, and possible operations are possible. Within a change range of the pattern, the situation is predicted after a certain time when the pattern is changed and the corresponding operation pattern is executed. Since the target area is wide, the number of target traffic lights and the scale of the network are large, so the process of changing the control plan and performing prediction after a certain period of time is Since the method is such that they can operate in parallel, and the probability of selection as the original data of the iterative processing is increased by the value of the evaluation value for a plurality of control plans, a good control plan can be determined in an extremely short time.

[0016]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration example of a signal control system according to the present invention. This signal control system controls a wide area traffic space, that is, a wide area road traffic space intentionally divided. The present system includes a signal control device 1 that determines an optimum signal control pattern, a display device 2 that displays an evaluation of a control pattern or a control result determined by the signal control device 1, a constraint condition and an objective function for control. Input device 3 for inputting the signal to the signal control device 1, a signal input device 9 for fetching traffic data from the space measurement sensor, and traffic signal devices installed at the traffic site for the signal pattern determined by the signal control device 1. The signal output device 10 outputs the signal to the control unit 11A. A large number of traffic signals 11 are installed as part of the traffic system in the traffic space to be controlled, and at least one traffic signal 11 can change its signal parameter, that is, a signal pattern. .

The signal control device 1 determines an optimum signal control pattern based on the input traffic data and input information. The display device 2 can be realized by hardware such as a CRT and a liquid crystal display device, and the input device 3 can be realized by a keyboard. The signal input device 9 takes in traffic data from an ultrasonic sensor 13 installed at an intersection or a road 12 or a space measurement sensor 14 that processes an image captured by a television camera. The signal control device 1 can be realized by a conventional process control computer or microcomputer of the sequential execution type, but when the number of control target signal devices exceeds 100, it has a structure in which a plurality of processing devices are connected in parallel,
A parallel computer capable of performing at least two types of independent processing on the same time axis is required.

Since the signal control method of the present invention has a logical structure in which the efficiency is not deteriorated even on a parallel computer, the target traffic space becomes vast and the number of control target signals is increased. Even if the number becomes large, it is possible to easily cope with it by preparing a sufficient number of processing devices. That is, unlike the conventional upper limit of the performance of the sequential execution type computer and the conventional parallel logic in which the effect remarkably decreases when the degree of parallelism increases, the method has no limit performance by the hardware of the computer.

FIG. 2 shows an example of the hardware configuration of the signal control device 1 when performing wide area control by the signal control system of the present invention. In this example, an input device control unit 8 which is connected to the input device 3 and takes in input information, an output device control unit 7 which is connected to the output device 2 and outputs output information, and a shared memory 6 which stores constants or variables. , A cell processor massively parallel control unit 4 for controlling the entire device, and a plurality of processing units (MPU) 5 connected to the shared memory 6 and the massively parallel control unit 4 and capable of executing respective processes at the same time. To be done.
In this example, the number of MPUs 5 is 825, but the number can be increased or decreased according to the required processability. As described above, the computer may be a sequential execution type computer having one processor.

Next, FIG. 3 shows a configuration example of a wide area transportation network in this system. In urban areas where buildings and houses are densely located, the main line that is often seen in the suburbs is the center, and the other lines are not subtly relatively subordinate. In addition, the road 12 forms a complicated two-dimensional network. In this example, the intersection marked with a circle indicates that the traffic signal S is installed, and the number of traffic signals is 140 (units). In this system, the traffic space is grasped as a network as shown in Fig. 3, and the information such as the position of the traffic light, the connection with the route, the traffic capacity of the route, the probability value obtained from the actual value of the straight, right turn and left turn at the intersection, etc. Is stored in advance in the traffic information storage means.

FIG. 4 shows a logical configuration of the signal control device 1. In this signal control device, at least
A process signal capturing means 102 for capturing a process signal,
Process signal output means 1 for outputting a control result to a process
04, an input device control means 106 for receiving input information from the input device 3, a display device control means 108 for controlling the display device 2, and traffic information for storing information on the traffic network and characteristics of traffic signals necessary for control. The storage unit 110, a traffic flow prediction unit 112 that predicts a traffic situation after a certain time when a signal control pattern is given, and an optimization unit 114 that determines an optimum signal pattern. The optimizing unit 114 determines and outputs a signal control command that takes in the captured traffic information, the constraint condition, and the objective function and maximizes or minimizes the given objective function value.
For example, it is constituted by a genetic optimization means.

FIG. 5 is an explanatory diagram of a one-dimensional vector having, as elements, a signal pattern for controlling the target area and a signal pattern number for each signal in the present invention. If the control execution cycle time is set to 160 (seconds), G (green), Y (yellow), R
Allocating time to (red) is equivalent to allocating passing traffic to the opposite route. Defining the control cycle here as 160 (seconds) is essentially different from the conventional cycle value, and is limited to the meaning of simply giving the control timing. In the example of FIG. 5, the number of possible patterns is 100. For example, the traffic light Si can have 100 signal bridges (patterns) and performs specified operations. This signal pattern 1 indicates that from that time point, G is lit for 20 seconds, Y and R are lit for 100 seconds, G is lit for 30 seconds again. G, Y, R for realistic traffic lights
Besides, there are right turn instructions under R, bus, pedestrian priority, etc., but these are only increasing the types of lighting,
If the combination of G, Y, and R is possible, the processing is the same.

In this way, the number of signal patterns is set as many as can be considered within the range in which the control conditions are held. Next, the signal pattern Pi of each traffic signal Si (i = 1 to 140 in this example)
A one-dimensional vector Xi whose element is is defined as X = (P ₁ , P ₂ , ... ......, P ₁₄₀ ). This X is hereinafter referred to as a chromosome. The signal state of the target space is uniquely determined by X. Proper definition of the elements of X is the essential purpose of the present invention.

The space of X, that is, the number n of possible cases
Since (X) has 100 patterns in each example and 140 signals in this example, n (X) = 100 ¹⁴⁰ = 10 ^{280, which} is a huge number. For each combination, the traffic situation after a certain period of time is predicted, and, for example, the combination with the least traffic congestion is obviously optimal. However, as mentioned above, making predictions for all of a large number of cases is
It is extremely difficult within the allowable time, and even with parallel computers, it is impossible with today's computers. Another important object of the present invention is to provide a method capable of doing this in a sufficiently short time.

FIG. 6 is a processing flow chart of the signal control device 1. This process consists of 12 steps and is started every 160 seconds. The operation will be described below. The hardware used in this example is a parallel computer in which 10 processing devices are connected in parallel.

(Step 1) Acquisition of Traffic Data Traffic data such as the number of vehicles existing in the traffic network, the moving speed, the traffic jam condition, the signal condition, etc. is acquired from the route installed on the traffic network and the sensor installed on the intersection.

(Step 2) Estimation of traffic conditions Since it is extremely rare that accurate sensors are installed on all routes and intersections on the traffic network, no sensors are installed by the traffic flow prediction means. Route,
The intersection is estimated based on the information obtained in (step 1), and the conditions of all the target transportation networks are grasped.

(Step 3) Generation of desired control plan As an initial value of the control plan, a chromosome X having the signal pattern given to each traffic light at that point as an element is generated.

(Step 4) Creation of control plan selection table Since the control plans X given to the 10 CPUs are the same in the initial state, each control plan can be an optimal solution candidate with the same probability. A control plan selection table as shown in 18 is created.

The following (step 5) to (step 9)
It is executed concurrently by different processing units (MPUs).

(Step 5) Two control plan selections at random The control plan selection table shown in FIG. 18 is 0.0 to 100.0.
(%) Has a range of 0.0-10
The control plan M (i, j) is determined for a value of 0.0. In this step, two kinds of random numbers r1 and r2 uniformly distributed in the range of 0.0 to 100.0 are generated, and two kinds of control plans are determined. Here, it is assumed that X ₁ and X ₂ are selected.

(Step 6) Mutation Operation The signal patterns of a part of the _two control plans X ₁ and X ₂ selected by the control plan selection in the above (step 5) are mutated.
This is called a mutation operation. FIG. 7 shows a detailed procedure of this mutation operation processing, and FIG. 8 shows a mutation example using an actual example. First, in (Step F10), from 1 to the number of traffic signals 1
Random number random1 uniformly distributed in the range of 40 and 0 or 1
A random number RAND1 that uniformly takes the value of is generated. Next, in step (F20), rand2 and RAND2 are similarly generated. FIG. 8 shows the process of (step F30). Here, if rand1 = 5 RAND1 = 0, the signal pattern of the traffic light 5 of X ₁ is 88, so this is rewritten to 87. That is, the signal number ra
The signal pattern of nd1 is updated in descending order when RAND1 is 0 and in ascending order when RAND1 is 1. As a result, X ₁ is partially mutated and a new control plan X ₁ ′ is generated. It goes without saying that this method is an example of the location of mutation, and other methods may be used. (Step F 40) the same manner, the X ₂ is a second control plan, the mutation operation was added, a new control plan X ₂ '
To generate.

(Step 7) Synthesis of control plan (mating) Combine the two control plans generated in (Step 6) above
(Mating) to generate a new control plan. Fig. 9 shows the main mating site
The detailed procedure of the processing is shown in FIG.
First, in step G10 in FIG.
Consists of a string where each element is uniformly 0 or 1.
Generate a template TP. In the example of FIG. TP = (0,1,0,0,0,1,1,0, ...
…) Has become. Next (step G30)
Generate a new control plan M (i, j) repeatedly for the number of units
It Here, j is the number of the MPU being processed and i is the number of generations.
It M (i, j) is X based on TP₁’X₂′ Property
Is synthesized to take over. That is, the value of TP is 0
In this case, the signal pattern of the traffic signal of Tp is X ₁′ Thing
Is set, and if the value of TP is 1, X₂'Set
To do so. In the example of FIG. 10, the traffic signals S1, S3, S
X for 4, S5 and S8₁'Pattern is S2
X in S6 and S7₂′ Pattern is adopted, new control
It can be seen that the plan M (i, j) is generated.

(Step 8) Prediction of Traffic Situation After Certain Time Predicted traffic situation after a certain time when the control plan M (i, j) generated in (Step 7) above is executed by the traffic flow prediction means. To do.

The processing procedure for predicting the traffic condition after a fixed time is shown in FIG. 11 and two typical cases are shown in FIG. 12, and the operation will be described. Although various methods have been proposed for making accurate predictions, very few methods are able to consider even complex mutual influences of intersections and routes on a wide area network. The prediction in the present invention is basically to drive one vehicle in a short time transition unit, and the calculation amount is large, but it includes a lot of uncertain information such as trend prediction by statistical information. There is no.

In the example of FIG. 11, the traffic condition prediction after 160 (seconds) is performed every 4 (seconds) by running a vehicle existing on the network and repeating this 40 times. . In (Step H20), (Step H30) to (Step H30) for all routes on the network.
The control of 60) is repeated. (Step H3
In 0), calculate the position of the vehicle 4 seconds after on the route,
Determine the number of vehicles that flow out from the relevant route to others. This is obtained by the product of the green time TG of the traffic light and the average vehicle speed VS, TG * VS. FIG. 12 shows an example of the current position of the vehicle on the route and the calculated position after 4 seconds. Case 1 shown as (a)
At, the vehicle can move straight ahead, turn right, and turn left. Currently 100 (units) on the route A as shown in the upper part of the figure.
However, if the oncoming traffic signal continues G for 4 seconds (TG = 4), considering the distance on the corresponding route with the average vehicle speed as VS, 4 seconds later, as shown in the lower part of the figure, 40
Cars can leak.

Next, in (Step H40), the outflow direction of the corresponding outflowing vehicles, that is, the straight, right turn, and left turn routes are sorted. This is calculated based on the past long-term statistical data including road conditions such as road width and number of lanes, regional customs, etc., using the straight-line rate, right turn rate, and left turn rate at each intersection that are accurately determined.

In case 1 of FIG. 12, the straight traveling rate is 50.
(%), The right turn rate is 25 (%), and the left turn rate is 25 (%), so 40 cars that leaked are 20 (vehicles) and 10 (vehicles) in the directions of straight, right, and left turns, respectively. 10 (vehicles) are moving smoothly.

(Step H50) calculates the number of vehicles that can go straight, turn right, and turn left depending on the vehicle conditions on the adjacent route. In Case 1, a sufficient margin was met, but in Case 2 shown in FIG. 12B, although there is a margin in the left turn direction, the vehicle is buried in the straight direction and the right turn direction is 8
Only (unit) can proceed. For this reason, 10 (vehicles) were able to proceed in the left turn direction, but 8 (vehicles) were not able to make a right turn and none were straight. In this example, it is assumed that the number of lanes is sufficient, but if this is not the case, the vehicle in the originally vacant direction will also be stopped by stopping the vehicle in the other direction.
It may become impossible to proceed. The number of vehicles on each route is updated based on this result. The above is the processing of (step H60).

As described above, since the vehicle is microscopically run and the prediction is made after a certain period of time, accurate prediction accuracy is realized. It is certain that the space measurement type using a TV camera will become the mainstream as a traffic condition sensor in the future, so that the number of vehicles at intersections and routes,
The speed of movement and the degree of traffic congestion can be obtained extremely accurately. This method improves the accuracy and diversification of such information,
It is possible to exert further effects.

(Step 9) Evaluation of Control Result This is a step of evaluating the traffic condition after a certain period of time, which is obtained in the above (Step 8). In this example, the congestion margin is designated as the control objective function, and this value is maximized. Here, the congestion degree TJ is defined as "the number of vehicles having a speed of 0 other than waiting for an R signal". Of course, a vehicle that cannot pass through the intersection at the next G timing even during the R lighting is included in the congestion degree TJ. On the other hand, the congestion margin CTG (vehicle) is defined by CTJ (vehicle) ← (allowable traffic (vehicle))-(congestion degree (vehicle)). Since the allowable traffic volume is uniquely determined if the target transportation network is determined, TJ and CTJ are qualitatively the same objective function. That is, the minimization of the congestion degree TJ is nothing but the maximization of the congestion margin CTJ. FIG. 13 shows a processing procedure of (step 9) control result evaluation, and FIG. 14 shows an example of a realistic traffic jam condition. (Step I20) is performed for all the routes, and TJ (vehicle) is calculated. Figure 1
This will be explained in Section 4.

In FIG. 14, signals are installed at four intersections A, B, C and D which are close to each other, and the lighting state at that time is as shown in the figure. At this time, the number of vehicles is indicated by a rectangle, and is indicated by a solid arrow when the travelable speed is not zero, and by a broken line when the speed is zero. When the signal is an R-light, there is no traffic even if the speed is zero, so the rectangle is displayed as white. On the other hand, vehicles that are not free to move due to traffic congestion are vehicles subject to traffic congestion,
The rectangle is hatched. As a result, the congestion degree TJ is obtained as described below by calculating the total sum of vehicles in the hatched portions on all routes. In this example, the allowable traffic on the network is 1,000 (units)
Then, the traffic congestion margin is CTJ (vehicle) = 1,000−TJ = 235 (vehicle).

In this example, traffic energy is specified as the control objective function, and this value is maximized. Traffic energy TE (i) is: TE (i) = TC (i) × AU, where TC (i) is the traffic volume on a certain route and AU (i) (km / h) is the average speed. Defined in (i), the traffic energy in the traffic network diagram is TEA = TE (i).

The aim of maximizing the traffic energy is whether the traffic volume is increased or the average speed is increased. Increasing traffic volume and increasing average speed is nothing but realizing smooth traffic conditions.

In this example, the purpose of control is to maximize the congestion margin, but the minimum number of waits, the minimum wait time, the minimum travel time, the minimum fuel consumption, the minimum time benefit cost, etc. may be set. Needless to say. These may be dynamically changed according to the traffic of the existing vehicle and the corresponding network at the control timing.

(Step 10) In the preparation of the selection table (Step 5) to (Step 9), the proposal M (i, j) for each generation j is generated and evaluated by 10 processing devices i in this example. Was done. In (Step 10), a selection table for preferentially selecting a control plan with a high degree of matching for a given purpose in the process of the next generation is created as follows.

In FIG. 16, the vertical axis represents the number i of the processing unit MPU and the horizontal axis represents the number of generations j, which is the number of repetitions controlled by (step 11), and the evaluation value of the control plan M (i, j) is shown. It is a thing. As described above, in the initial state (generation = 0), the control pattern for continuing the signal pattern at that time point is set, and this is assigned to all processing numerical values, so that the evaluation values are almost the same. However, in the generations 1 and later, the evaluation value varies due to the mutation operation in (Step 6) and the mating operation in (Step 7). In (step J10) in FIG. 15, all the control plans M in the generation i
The sum of the evaluation values CTJ (j) of (i, j) (j = 1 to 10) is calculated. For example, in generation 1, ΣCTJ (1) = CTJ (1,1) + CTJ (2,1) + CTJ (3,1) + CTJ (4,1) + ... + CTJ (10,1) = 1,770 . Next, in (Step J20), the total ΣCTJ
The evaluation value CTJ (i,
The ratio of j) is calculated.

FIG. 17 shows this situation.
In generation 1, M (6,1) has the highest evaluation, and the overall 2
It can be seen that it has an evaluation value of 8.2 (%). On the other hand, M (4,1), M (5,1), M (9,1) is 5.6.
(%), Indicating that the degree of agreement with the purpose is low. In (Step J30), a priority selection table of the control plan is generated based on the value calculated in (Step J20).

FIG. 18 is a selection table for generations 1 and 100 created based on the values shown in FIG. The pie chart has a scale of 0.0 to 100.0, and each value is assigned a control plan number. When a control plan is selected, variables that are uniformly distributed in the range of 0.0 to 100.0 are generated, and by defining the control plan by this, the control plan is determined in the order of priority of evaluation values. Become. That is, the control plan M (6,1) to which 28.2 (%) is allocated has a probability of 5 times or more compared to M ⁵ ₁ to which only 5.6 (%) is allocated. Will be selected as. This leaves many good control plans,
Those that are not are shown to be weeded out.

It can be seen that in the initial stage of generation = 1, the evaluation values greatly vary, but when generation = 100, the evaluation values of all control plans are high and become uniform. In other words, by updating the generations, evolution has taken place and the standard has risen.

(Step 11) Repeating a predetermined number of generations (Step 5) to (Step 10) is controlled to repeat a predetermined number of generations. The number of generations is determined by the scale of the target problem and the parallel processing frequency, that is, the parallel number.

(Step 12) Output of the best control plan Among the finally existing control plans, the one with the best evaluation value is selected as the optimal plan and output. In the example of FIG. 16, MPU10
Since the control scheme in 1) is the best, this signal pattern is output.

FIG. 19 is a comparison diagram of the effects of the signal control system according to the present invention and the conventional control system. For the signal control system according to the present invention, the maximum congestion margin is given as the objective function and the strategy. The conventional control system is
This is a method in which the signal pattern is fixed. In the two graphs, the horizontal axis represents the time transition from 7:00 to 10:30, and the vertical axis defines the congestion degree in the upper figure and the congestion margin degree in the lower figure.
As described above, the congestion degree and the congestion allowance have the same contents, and therefore, the congestion degree graph will be described here. In the above figure, the solid line shows the control result of the signal control system according to the present invention, and the broken line shows the result of the conventional control system. From this graph, it can be seen that a significant improvement is achieved by the present invention particularly in the time zone of 7:50 to 9:30 when the traffic volume is large. In other words, when the traffic volume is low and the cross-spreading of intersections and routes is small, the conventional method will produce good results, but when the traffic volume approaches the traffic capacity of the target traffic space, the conventional control method will no longer be effective. It turns out that can't be expected.

FIG. 20 shows the target traffic space three-dimensionally. The first and second axes show the map of the target area.
It is the figure which defined the congestion intensity as the 3rd axis | shaft, The upper figure shows the conventional fixed signal pattern control, and the lower figure shows the congestion situation at time 8:30 of the signal pattern control which maximizes the traffic energy by this invention. It is a thing. In the above figure, there are large peaks of congestion everywhere. On the other hand, in the figure below, the control is performed after observing the entire target area comprehensively and widely, so the peaks are well suppressed and the traffic volume is evenly distributed, resulting in congestion. It can be seen that is greatly improved and a smooth traffic flow is realized. That is, as in the conventional method, the local optimization rather reduces the traffic processing capacity.

In FIG. 21, not only the degree of congestion but also the evaluation items
It is expanded to traffic volume, number of stops, waiting time, travel time, number of vehicles in dilemma zone, total cost, accident probability, and these evaluation values are plotted for each time transition. Compared to the result of the conventional fixed signal pattern control shown in the above figure,
It can be seen that most of the items are greatly improved by the signal pattern control according to the present invention shown in the figure below.

By properly controlling the traffic signal in the traffic space in this way, the following effects can be obtained. (1). Congestion is eliminated or reduced over the entire area. (2). Local congestion peaks are suppressed. (3). The number of stops is reduced. (4). The waiting time is reduced. (5). Travel time is reduced. (6). Useless fuel consumption is reduced. (7). Transportation benefits are improved.

[0057]

As described above, according to the present invention, it is possible to accurately predict the traffic condition for all traffic routes included in the traffic space to be controlled, so that the entire traffic space is covered. Thus, it becomes possible to configure a control plan that alleviates or reduces traffic congestion.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an embodiment of a signal control system of the present invention.

FIG. 2 is a diagram illustrating a hardware configuration example of a signal control device.

FIG. 3 is a diagram in which a traffic space to be controlled is represented by a network.

FIG. 4 is a diagram showing a functional configuration of a signal control system.

FIG. 5 is a diagram showing an example of a chromosomal expression of a traffic spacetime having a traffic light pattern as an element.

FIG. 6 is a flowchart of signal control processing.

FIG. 7 is a flow chart of mutation operation processing of a traffic light state.

FIG. 8 is an example of a mutation operation.

FIG. 9 is a flowchart of a mating operation process.

FIG. 10 is an example of a mating operation.

FIG. 11 is a traffic situation prediction processing flowchart.

FIG. 12 is an example of traffic condition prediction.

FIG. 13 is a control result evaluation processing flowchart.

FIG. 14 is an example of traffic congestion evaluation.

FIG. 15 is a flowchart of a control plan selection table creation process based on evaluation values.

FIG. 16 shows a transition of the control plan evaluation value for each generation.

FIG. 17 shows the evaluation value of each control plan with respect to the previous evaluation value.

FIG. 18 is an example of a control selection table.

FIG. 19 is a diagram showing an effect when the objective function is set to the maximum congestion margin.

FIG. 20 is a diagram showing a three-dimensional representation of a traffic space and showing a comparison of congestion degree distributions by the conventional control and the control of the present invention.

FIG. 21 is an explanatory diagram of a simultaneous improvement situation of multinomial evaluation values.

[Explanation of symbols]

1 ... Signal control device, 2 ... Display device, 3 ... Input device, 4 ...
Massively parallel control unit, 5 ... Micro processing unit, 6 ... Shared memory, 7 ... Display device control unit, 8 ... Input device control unit, 9 ... Signal input device, 10 ... Signal output device.

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Shozo Shiozawa             Hitachi 2-3-1, Saiwaicho, Hitachi-shi, Ibaraki             Engineering Co., Ltd. (72) Inventor Kenichi Nakamura             Hitachi 2-3-1, Saiwaicho, Hitachi-shi, Ibaraki             Engineering Co., Ltd. (72) Inventor Masahiro Yahiro             5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture             Ceremony company Hitachi Ltd. Omika factory (72) Inventor Yoshizo Ito             5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture             Ceremony company Hitachi Ltd. Omika factory (72) Inventor Yoshiyuki Sato             5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture             Ceremony company Hitachi Ltd. Omika factory (72) Inventor Yutaka Sano             5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture             Ceremony company Hitachi Ltd. Omika factory (72) Inventor Takayoshi Yokota             7-1-1, Omika-cho, Hitachi-shi, Ibaraki Prefecture             Inside the Hitachi Research Laboratory, Hitachi Ltd. F-term (reference) 5H180 AA01 CC04 CC11 DD02 DD04                       EE02 EE03 JJ02

Claims (2)

[Claims] 1. A method of predicting a traffic condition in a wide area road traffic space to be controlled, wherein the traffic condition at that time is input from a traffic sensor of the traffic space, and the traffic space is sufficiently short. A method of predicting a traffic situation in a wide area road traffic space, comprising predicting the behavior of all vehicles existing in the vehicle, and repeating the predictive computation until a predetermined future time is reached. 2. A method of predicting a traffic condition in a wide area road traffic space to be controlled, comprising: a first step of capturing traffic data in the traffic space; and all vehicles existing in the traffic space. On the other hand, the position after a sufficiently small time is straight ahead ratio, right turn ratio, which is determined proactively with the traffic signal instruction information.
It is characterized by comprising a second step for determining left turn ratio, vacant straight road, vacant left road and vacant right road information, and a third step for controlling the second step to repeat for a desired time. Method for predicting traffic conditions in a wide area road traffic space.