JPH08124085A - Aspect design device - Google Patents

Abstract

PURPOSE: To provide a new aspect design device which is capable of easily designing an aspect for a beginner. CONSTITUTION: When the traffic amount of the right-turn, the direct advance and the left-turn for every advancing direction of each road according to the shapes of the roads of plural intersections operated in the past is inputted, a neural network 2 performs learning. When the traffic amount of the right-turn, the direct advance and the left-turn for every advancing direction of each road according to the shape of the road of the intersections to be a present instruction design object is inputted, the flow line number of the intersection of the present instruction object is outputted by defining the learning result as an educator signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a display design apparatus for designing a display that is a traffic right used for traffic signal control.

[0002]

2. Description of the Related Art In general, traffic flows in different directions are sequentially given traffic lights. That is, only vehicles and pedestrians in a specific direction who are given the right to pass can see the green light, and the green light (right to pass) given to this group of traffic flows is "presented". Say. For example, in the case of a general crossroad intersection as shown in FIG. 7, when the up-down direction display and the left-right direction display of the drawing are alternately switched, this is referred to as two display, which is the most common display. . If there are many right-turn vehicles, a special sign for the right turn is inserted between the signs, which is called 3 signs.

Designing a display used for traffic signal control in this way means designing a display map having the displays ~ as shown in Fig. 8 from the shape of each intersection and the traffic conditions. Conventionally, when designing such a presentation,
Experts were based on their experience and skills.

[0004]

However, the conventional method has a problem that it cannot be designed easily because the expert designs the presentation based on his experience and technique.

The present invention has been made in view of the above problems of the prior art, and an object of the present invention is to provide a novel marking designing apparatus which enables a beginner to easily design a marking.

[0006]

In order to achieve the above object, the invention as set forth in claim 1 has been operated in the past in a display design apparatus for designing a display which is a right of use used for traffic signal control. The first input means for inputting the traffic volume of right turn, straight ahead, left turn for each traveling direction of each road according to the shape of the road at the plurality of intersections, and A second input means for inputting the traffic volume of right turn, straight turn, and left turn for each traveling direction of the road, and past operation data input via the first input means are learned by a neural network, and the second operation means is used. Learning means for outputting the actual number of the intersection of the actual design target input through the input means. The learning means according to a second aspect of the present invention is characterized in that learning is carried out at a saturation ratio of the maximum traffic volume that can pass through and the actual traffic volume for each of the traveling directions of each road, such as a right turn, a straight road, and a left turn. A learning unit according to a third aspect of the present invention has an input layer including an arithmetic unit corresponding to the number of pixels of the saturation ratio bar graph. The learning means according to claim 4 outputs the streamline diagram of the intersection of the actual design object together with the actual number.

[0007]

According to the first aspect of the present invention, the traffic volume of a right turn, a straight turn, and a left turn is learned by a neural network for each traveling direction of each road according to the shapes of roads at a plurality of intersections that have been operated in the past. The streamline number of the designated design intersection is output. Therefore, according to the shape of the road at the intersection for which the display is designed, the display can be easily designed only by inputting the traffic volume for the right turn, the straight travel, and the left turn for each traveling direction of each road. According to the second aspect of the present invention, learning is performed by the saturation ratio of the maximum passing traffic volume and the actual traffic volume. According to the third aspect of the present invention, learning is performed by the input layer configured by the arithmetic units corresponding to the number of pixels in the saturation ratio bar graph.
Further, in the invention according to claim 4, the streamline diagram of the intersection of the actual design object is output together with the actual number.

[0008]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a visual design apparatus according to the present invention, and FIG. 2 is a flow chart for explaining a learning operation. In FIG. 1 and FIG. 2, first, data of hundreds of intersections which are currently in operation and whose indications are known are prepared and input through the data input unit 1 (step S
1). In this process, an intersection such as a three-way intersection, a four-way intersection, a five-way intersection, etc. is selected, a learning mode screen as shown in FIG. 3 is displayed, and past operation data can be input. On this screen, for example, you can enter the average hourly traffic volume for right, straight, and left turns in each direction of travel in the north, south, east, and west directions with the saturation ratio (%). The "End of setting" button is displayed.

Here, the saturation ratio refers to a ratio between the maximum traffic volume that a vehicle can pass in one hour and the actual traffic volume, and the traffic volume of left turn, straight turn, and right turn for each traveling direction. When the saturation ratio ~ is input, the bar graph is displayed as shown in FIG. Also, when the saturation ratio of the pedestrian in that direction is input, it is also displayed as a bar graph.

The neural network 2 is composed of, for example, three layers of an input layer j, an intermediate layer k and an output layer 1 as shown in FIG. 5, and the input layer j is the number of pixels of the bar graphs shown in FIGS. 3 and 4. The arithmetic unit of is used. The learning signal is transmitted from the input layer j to the output layer 1 through the intermediate layer k.
The middle layer 32 is changed by the learning time, and the output layer 3
3 is composed of one unit and outputs a display number.

The weight w is applied to the connection from the input layer j to the intermediate layer k.
The weight wkl is defined for the connection of jk from the intermediate layer k to the output layer l, and the error function is represented by E = (1 / 2j) (tj-vj) @ 2. However, tj: teaching data of unit j of output layer 1 (desired output) vj: actual output of unit j of output layer l vj = fj (Σwjk · vk−θj)

Learning of the neural network 2 is performed by using the back propagation method, and is performed according to the flow shown in FIG. The backpropagation method is an error function E
Weighting factor w between the output layer 1 and the intermediate layer k so that
kl, and the weighting coefficient wjk of the intermediate layer k and the input layer j is changed. Therefore, the weighting coefficient correction amounts Δwjk, Δ
Calculate wkl and repeat the correction until it becomes less than a certain value.

[0013]

[Equation 1]

[0014]

[Equation 2]

Θj and θk are thresholds, and the output is always "-
The unit of "1" is considered, and the value of the weight of the link is expressed from that unit. This makes it possible to associate a plurality of input patterns with output patterns.

In FIG. 2, first, data of several hundreds of intersections which are currently in operation and whose indications are known are prepared and input (step S1). Here, the present number used as a teacher signal is shown. When the "learning execution button" is pressed after inputting data of hundreds of known intersections on the learning mode screen shown in FIG. 3, learning is started (step S2), and the overall error is the reference value. The parameters are adjusted until (step S3 → S4). Here, the reference value of the error is set by the learning time or the like.

Upon completion of such learning, the shape (3 difference road, 4 difference road, 5 difference road, etc.) of the intersection to be designed is selected and, for example, a 4 difference road input screen as shown in FIG. 6 is displayed. Is displayed. On the four-lane input screen shown in FIG. 5, the average 1 hour traffic volume of right turn, straight road, and left turn and the traffic volume of pedestrians can be entered in the saturation ratio (%) for each traveling direction of north, south, east, and west.
The "execution" mode and the "learning" mode can be designated. Further, the saturation ratio of traffic volume for left turn, straight road, and right turn for each traveling direction and the saturation ratio of pedestrians in that direction are displayed in a bar graph.

When the "execution" button is pressed by inputting the saturation ratios ~ and the pedestrian saturation ratios of the traffic volume for left turn, straight road, and right turn for all traveling directions, the neural network 2 causes the neural network 2 as shown in FIG. Is displayed. Therefore, the representation is designed by the neural network 2.

[0019]

As described above, according to the first to fourth aspects of the present invention, right turn, straight turn and left turn traffics are made for each traveling direction of each road according to the shape of the roads at a plurality of intersections that have been operated in the past. The quantity is learned by the neural network, and the streamline number of the intersection of the actual design target is output. Therefore, according to the shape of the road at the intersection for which the display is designed, the display can be easily designed only by inputting the traffic volume for the right turn, the straight travel, and the left turn for each traveling direction of each road.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a physical design apparatus according to the present invention.

FIG. 2 is a flowchart for explaining a learning operation.

FIG. 3 is an explanatory diagram showing a learning mode screen.

FIG. 4 is an explanatory diagram showing a saturation ratio in one traveling direction.

5 is a configuration showing the neural network of FIG. 1. FIG.

FIG. 6 is an explanatory diagram showing an input screen of a current design target.

FIG. 7 is an explanatory diagram showing a general representation of a crossroad intersection.

FIG. 8 is an explanatory diagram showing a current drawing.

[Explanation of symbols]

 1 Data input section 2 Neural network

Claims (4)

[Claims] 1. A display designing device for designing a display that is a traffic right used for traffic signal control, in which a right turn is made for each traveling direction of each road according to the shapes of roads at a plurality of intersections that have been operated in the past. The first input means for inputting the traffic volume for straight or left turn, and the second input means for entering the traffic volume for right turn, straight line, left turn for each traveling direction of each road according to the shape of the road at the intersection of the design design target Input means and the past operation data input via the first input means are learned by a neural network, and the display number of the intersection of the display design target input via the second input means is output. And a learning means for performing the presentation designing device. 2. The learning means learns at a saturation ratio of a maximum traffic volume that can be passed rightward, straight ahead, and leftward and an actual traffic volume for each traveling direction of each road. The design design device described. 3. The display designing apparatus according to claim 2, wherein the learning unit has an input layer formed of arithmetic units corresponding to the number of pixels of the bar graph of the saturation ratio. 4. The physical design apparatus according to claim 1, wherein the learning unit outputs a streamline diagram of the intersection of the physical design target together with the physical number.