JPH1079096A - Device for providing path selection support information - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prepare a predicted congestion level by learning a traffic situation by a neural network. SOLUTION: A section required time is calculated by a section required time calculating means 3 by using a spatial mean speed measured by a vehicle sensor 1, and a required time result value is measured by AVI automatic vehicle identibication system 2. A hierarchical neural network 6 for prediction inputs the latest section required time data, and calculates a required time predicted value by using a weighting factor updated by a weighting factor updating means 7. A hierarchical neural network learning means 9 operates learning by comparing the required time result value as a teacher signal with the output of a hierarchical neural network 8 for learning which inputs the section required time data, and calculates weighting factor correction amounts. Then, when a required time predicted error in the hierarchical neural network learning means 9 is within a set range, a prediction arithmetic operation is turned ON by a perdition starting timing judging means 10, and a predicted congestion level is calculated by a predicted congestion level calculating means.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a traffic information providing system and, more particularly, to an information providing system for a traffic control system on an expressway, the degree of congestion in a route section, and a route selection support.

[0002]

2. Description of the Related Art Expressways, railroads, and the like are major transportation modes in urban areas. At present, there is an increasing need for information that assists the user in making a decision such as a route selection. ing.

In route selection, information is provided through a display of a navigation system or the like in order to determine an optimum route based on the traffic conditions of various selectable routes, the length of the route distance, the toll when using a toll road, and the like. ing. Recently, there has been proposed a method of obtaining an optimal route in a short time by applying a neural network from data of only the route based on map information.

[0004]

The above-mentioned conventional methods only provide data at the present time,
It does not take into account how traffic conditions will change thereafter. Therefore, in order to provide more useful information in an urban area where the traffic situation changes rapidly, it is important to predict the change in the traffic situation in addition to the current status report.

(Object of the Invention) The present invention predicts a change in traffic conditions based on current data by using a method such as a neural network, and creates support information for route selection and the like, thereby providing more useful support. We propose a method to create information.

[0006]

According to a first aspect of the present invention, there is provided a route selection support information providing apparatus comprising: (a) a plurality of vehicle sensors installed in a route section; AVI systems installed at both ends of the section (c) Using the spatial average speed obtained from the vehicle detector,
Section required time calculating means for calculating required time in each section when the target route section is divided (d) Section required time storing means for storing section required time obtained by section required time calculating means (e) Obtained from AVI system Required time actual value storage means for storing the required required time actual value (f) A hierarchical neural network for prediction which receives a section required time obtained from the section required time calculating means as an input and outputs a required time predicted value based on a weight coefficient ( g) Hierarchical neural network for learning, which inputs section required time data obtained from section required time storage means and outputs a predicted value of required time based on a weighting factor. (h) Hierarchical neural network for prediction and hierarchical neural for learning. Used on both sides of the network Weight coefficient updating means for updating the weight coefficient (i) Correction of the weight coefficient using the required time predicted value output from the learning hierarchical neural network for learning and the required time actual value obtained from the required time actual value storage means Hierarchical neural network learning means for calculating the quantity (j) Predictive start timing determining means for determining the timing to start prediction based on the required time prediction error obtained from the hierarchical neural network learning means (k) Predictive start timing determining means A predicted traffic congestion degree calculating means for calculating a predicted traffic congestion degree based on the required time predicted value obtained as described above is provided.

The route selection support information providing device according to the second aspect of the present invention includes: (a) a plurality of vehicle sensors installed in a route section; and (b ') a section required time stored in the section required time storage means. A required time actual value calculation means for calculating a required time actual value calculated value in consideration of a lapse of time due to traveling; (c) using a spatial average speed obtained from a vehicle sensor,
Section required time calculating means for calculating required time in each section when the target route section is divided (d) Section required time storing means for storing section required time obtained from section required time calculating means (e ') Required time actual Required time actual value storage means for storing the required time actual value obtained from the value calculating means (f) Prediction hierarchical neural network for inputting the section required time obtained from the section required time calculating means and outputting the required time predicted value (G) Hierarchical neural network for learning which receives section required time data obtained from the section required time storage means and outputs a predicted value of required time (h) Both the neural network for prediction and the neural network for learning Weight coefficient that updates the weight coefficient used in Updating means (i) Hierarchical neural network learning for calculating the correction amount of the weighting factor using the required time predicted value output from the learning hierarchical neural network for learning and the required time actual value obtained from the required time actual value storage means Means (j) Predicted start timing judging means for judging timing to start prediction based on required time prediction error obtained from hierarchical neural network learning means (k) Based on required time predicted value obtained from predicted start timing judging means And a predicted congestion degree calculating means for calculating the predicted congestion degree.

According to a third aspect of the present invention, there is provided a route selection support information providing apparatus which divides each target line into several sections, and installs the route selection support information providing apparatus according to the first aspect in each section. (L) Predicted congestion degree table creation means for creating a predicted congestion degree table for the target route based on the predicted congestion degree obtained from the predicted congestion degree creation device installed in each section of each route (m) Predicted congestion degree A route selection support information providing unit for generating route selection support information is provided based on a predicted congestion degree table for each route obtained from the table creation unit.

According to a fourth aspect of the present invention, there is provided a route selection support information providing apparatus which divides each target route into several sections, and installs the route selection support information providing apparatus according to the third aspect in each section.
Further, (l) a predicted congestion degree table creating means for creating a predicted congestion degree table for the target route based on the predicted congestion degree obtained from the predicted congestion degree creation device installed in each section of each route. A route selection support information providing unit for generating route selection support information is provided based on a predicted congestion degree table for each route obtained by the degree table creation unit.

According to a fifth aspect of the present invention, there is provided a route selection support information providing apparatus, comprising: (a) a plurality of vehicle sensors installed in a route section; and (b ") a section based on the spatial average speed obtained from the spatial average speed storage means. A required time actual value calculation means for calculating a more strict required time actual value in consideration of the overshoot between (c) using the spatial average speed obtained from the vehicle sensor,
Section required time calculating means for calculating the required time in each section when the target route section is divided (d) Section required time storing means for storing the section required time obtained from the section required time calculating means (e ") Actual required time Required time actual value storage means for storing the required time actual value obtained from the value calculating means (f) Prediction hierarchical neural network for inputting the section required time obtained from the section required time calculating means and outputting the required time predicted value (G) Hierarchical neural network for learning which receives section required time data obtained from the section required time storage means and outputs a predicted value of required time (h) Both the neural network for prediction and the neural network for learning Weight coefficient that updates the weight coefficient used in Updating means (i) Hierarchical neural network learning for calculating the correction amount of the weighting factor using the required time predicted value output from the learning hierarchical neural network for learning and the required time actual value obtained from the required time actual value storage means Means (j) Predicted start timing judging means for judging timing to start prediction based on required time prediction error obtained from hierarchical neural network learning means (k) Based on required time predicted value obtained from predicted start timing judging means Predicted congestion degree calculating means for calculating predicted congestion degree (n) Spatial average speed storage means for storing the spatial average speed obtained from the vehicle sensor is provided.

According to a sixth aspect of the present invention, there is provided a route selection support information providing apparatus which divides each target line into several sections, and installs the route selection support information providing apparatus according to the fifth aspect in each section. (L) Predicted congestion degree table creation means for creating a predicted congestion degree table for the target route based on the predicted congestion degree obtained from the predicted congestion degree creation device installed in each section of each route (m) Predicted congestion degree A route selection support information providing unit for generating route selection support information is provided based on a predicted congestion degree table for each route obtained from the table creation unit.

Further, the apparatus for preparing a predicted congestion degree table according to claim 7 includes: (a) a vehicle sensor installed in a plurality of sections; (n) a spatial average speed storage for storing a spatial average speed obtained from the vehicle sensor. Means (o) Occupancy calculating means for calculating occupancy closely related to traffic density using data obtained from the vehicle sensor (p) Traffic density for calculating traffic density using occupancy obtained from occupancy calculating means Arithmetic Means (q) A neural network that inputs a traffic density (or occupancy) of a section and traffic densities (or occupancy) of a plurality of downstream sections of the section and calculates a predicted spatial average speed of the section (r) The actual traffic density value (or actual occupancy value) of the section, the actual traffic density value (or A neural network learning means for learning the above-mentioned neural network from the actual value of the spatial average speed of the section and the actual value of the spatial average speed of the section. (S) The spatial average speed is calculated based on the predicted spatial average speed obtained from the neural network. Traffic flow prediction means for predicting the transition of (vehicle speed), traffic density and traffic volume (t) Vehicle speed transition prediction storage means (l) for storing the transition of the spatial average speed obtained from the traffic flow prediction means ') A predicted congestion degree table creating means for creating a predicted congestion degree table based on the transition of the spatial average speed obtained from the vehicle speed transition prediction storage means is provided.

In the route selection support information providing device according to the present invention, the predicted congestion degree table creation device according to the seventh aspect is installed on each target route, and (m ') the predicted congestion degree table creation device. A route selection support information providing unit that creates route selection support information is provided based on the predicted traffic congestion degree table of each route obtained as a basis.

Next, the operation of each of the above means will be described.

[0015] In the route selection support information providing device according to the first aspect, (a) the vehicle sensor measures the traffic volume, the spatial average speed, and the occupancy of the traveling vehicle.

(B) The AVI system measures the actual value of the required time by recognizing the number of the traveling vehicle by image.

(C) is means for (a) receiving the spatial average speed data from the vehicle sensor and calculating the required section time using the set section length.

(D) is a means for receiving the section required time from the (c) section required time calculating means and storing a portion corresponding to the set time.

(E) means (b) means for receiving the actual required time value from the AVI system and storing a value corresponding to the set time.

(F) receives the section required time data from the (d) section required time storage means;
Is used to calculate the required time prediction value from the hierarchical neural network using the weighting factor updated by the above.

(G) is means for receiving the required section time in the past from (d) and calculating the predicted value of the required time by the hierarchical neural network using the weighting factor updated in (h). This means has the same structure as the above (f). The point is that since the required time is predicted using the past section required time data, the corresponding required time actual value is already stored in (e).

Here, the hierarchical neural network used in the above means (f) and (g) will be described.

As shown in FIG. 9, the hierarchical neural network employs a configuration in which a signal flows only in one direction such as an input layer → intermediate layer → output layer. The signal flow in the hierarchical neural network of FIG. 9 will be described.

The following symbols are as follows:

M: number of elements in the input layer s: number of elements in the intermediate layer n: number of elements in the output layer Uk (t): input to the k-th element in the input layer (k = 1 to m) Ik (t): input Output from the k-th layer element (k = 1 to m) NetHj (t): Input to the j-th element in the middle layer (j = 1 to s) Hj (t): Output from the j-th element in the middle layer (j = 1 to s) Net0i (t): Input to the i-th element in the output layer (i = 1 to n) Xi (t): Output from the i-th element in the output layer (i = 1 to n) ajk: k to the input layer Coupling weight coefficient bij from the element output end to the j-th element input end of the intermediate layer bij: coupling weight coefficient from the j-th element output end of the intermediate layer to the i-th element input end of the output layer fH (•): input / output conversion function in the intermediate layer f0 (•): input / output conversion function in the output layer In the present invention, the input Ui (t) is the section required time data of each section from the section required time storage means,
The output Xi (t) corresponds to the required time predicted value. -Input unit Receives an external input and determines the activation value k (t) according to the following equation.

K (t) = Uk (t) (k = 1 to m) (X.1) Intermediate unit The sum of signals from the input unit is received as input NetHj (t), and a new activation value Hj (t) is received. ).

[0027]

(Equation 1) Here, fH (·) in the equation (X.3) represents an input / output function in the intermediate unit. As the input / output function, a sigmoid-shaped (S-shaped) function is usually used. Output unit The sum of the outputs from the intermediate units is received as input Net0i (t), and a new activation value 0i (t) is determined, which is taken as the output Xi (t) of the entire neural network.

[0028]

(Equation 2) Here, f0 (·) in the equation (X.5) represents an input / output function in the output unit.

As described above, by separating the units into layers and determining the activation value synchronously for each layer, the input / output relationship of the entire neural network can be expressed by the following equation.

[0030]

(Equation 3) (H) is means for receiving the correction amounts Δapq and Δbpq of the weighting factors apq and bpq of the neural network from (i) and updating the weighting factors by the following equation.

Apq (t + 1) = apq (t) + Δapq (t) (X.7) bpq (t + 1) = bpq (t) + Δbpq (t) (X.8) (i) is obtained from (e). A means for calculating the correction value of the weighting factor by receiving the actual required time value from the past and the predicted required time value corresponding to the actual required time value from (e), using the actual required time value as a teacher signal, and learning the network of (g). is there.

FIG. 10 shows a processing flow in this means. Here, the processing of FIG. 10 will be described. The following symbols are newly defined.

E: Error evaluation function Vi (t): Teacher signal for output from the i-th element in the output layer Δapq: Correction amount of weight coefficient apq Δbpq: Correction amount of weight coefficient cpq ε: Weight coefficient learning parameter α: Weight coefficient Learning Parameter In this means, Vi (t) corresponds to the actual required time value from the AVI system.

In this means, the correction amount Δ
Apq (t) and Δbpq (t) are calculated as follows.

Δapq (t) = Δsapq (t) + α · Δapq (t−1) (X.9) Δbpq (t) = Δsapq (t) + α · Δbpq (t−1) (X.10. ) Where Δs apq in equations (X.9) and (X.10).
(K) and Δs bpq (k) are calculated as follows using the k-th sample data.

[0036]

(Equation 4) Here, the error evaluation function E is the network output Xi (t)
And the teacher signal Vi (t),

[0037]

(Equation 5)

[0038]

(Equation 6) Becomes

(J) is a means for receiving the prediction error of the required time from (i) and outputting the predicted value of the required time received from (f) when the prediction error becomes smaller than the set value. is there.

(K) is a means for receiving a predicted value of the required time from (j) and calculating a predicted congestion degree indicating a future congestion state based on the predicted value of the required time. For example, assuming that the required time of a target line during non-traffic congestion is T [seconds] and the required time predicted value is Tp [seconds], the predicted congestion degree C is calculated by the following formula.

C = Tp / T Non-congestion: C = 1 Congestion (somewhat congestion): 1 <C <2 Congestion: 2 <C
(A), (b), (c), (d), (e), (f),
For (g), (h), (i), (j), and (k),
It is the same as the predicted congestion degree calculation device of claim 1, and by installing it on a plurality of target routes, the predicted congestion degree of each line is calculated, and the route selection support information is calculated by (l) and (m). Is to create. Specifically, (l),
(M) has the following effects.

(L) is a means for creating a predicted congestion degree table as shown in FIG. 11 based on the predicted congestion degree obtained from the predicted congestion degree calculation device for each route.

(M) is means for creating route selection support information based on the predicted congestion degree table obtained from (l). As the route selection support information, for example, a route that is least congested (or predicted as such), a route toward which the traffic congestion is resolved (or predicted as such), or congestion becomes severe Routes (or expected to do so).

Further, the apparatus for calculating a predicted congestion degree according to the third aspect of the present invention is characterized in that (a), (c), (d), (f), (g),
Regarding (h), (i), (j), and (k), the operation is the same as that of the predictive congestion degree calculation device of claim 1.
(B ') and (e') have the following operations.

(B ') is means for calculating the calculated value of the required time actual value from the section required time obtained by the section required time storage means, taking into account the passage of time as shown in FIG.

The operation flow of FIG. 12 will be described next. The following symbols are as follows.

T: Required time actual value I: Section number t: Time ST (I, t): Section required time of section I at time t R: Total number of sections Δt: Sample period [STEP1] Initialization of each variable Do.

T = 0, I = 0, t = 0 [STEP 2] The section required time of the section I (time t) is added to the required time T.

T = T + ST (I, t) [STEP 3] It is checked whether or not the required section time of all sections has been added.

When the addition is completed, the required time actual value calculation processing ends. If not, the process proceeds to [STEP 4]. [STEP 4] The value of the actual time required value T currently being calculated is
A check is made to see if it is larger than a value obtained by adding Δt to the time t in the calculation process. If not, select [S
TEP5].

If it is larger, the time of the operation process is advanced by the following operation.

T = t + int {(T-t) / 5} × 5 (Note: int is a function for converting an integer value by rounding down decimal places.) [STEP 5] Advance section I by one.

I = I + 1 [STEP 6] Go to [STEP 2].

As described above, [STEP 1] to [STEP 6] are the flow of the (b ') required time actual value calculation means shown in FIG.

Further, according to the fourth aspect of the present invention, there is provided a route selection support information providing apparatus comprising: (a), (b '), (c), (d),
(E '), (f), (g), (h), (i), (j),
(K) is the same as the predicted congestion degree calculation device of claim 3, and is installed on a plurality of target routes to calculate the predicted congestion degree of each line, and (l), (m) Thus, route selection support information is created.
(L) and (m) are the same as those of the route selection support information providing device according to the second aspect.

Further, the apparatus for calculating a predicted congestion degree according to the fifth aspect of the present invention is characterized in that (a), (c), (d), (f), (g),
Regarding (h), (i), (j), and (k), the operation is the same as that of the predictive congestion degree calculation device of claim 1.
(B ″), (e ″), and (n) have the following operations.

(B ″) is means for strictly calculating the calculated value of the actual required time value by the algorithm of the flowchart of FIG. 13 from the spatial average speed of the route section obtained from (l).

The operation flow of FIG. 13 will be described below. The following symbols are as follows.

I: section number J: counter relating to time K: number of repetitions L (I): section length of section I (during operation,
The value of L (I) changes. ) RL (I): distance obtained by subtracting the distance traveled through section I from section length of section I to Δt V (I, J × Δt): spatial average velocity of section I at (initial time + J × Δt) Δt: Sample period ST (I): Required time of section I of section I T: Actual required time [STEP 1] Initial values of each variable are set as follows.

J = 0 I = 1 ST "(I) = 0 L (I) = section length of section I [STEP 2] Initialization of the number of repetitions K is performed.

K = 0 [STEP 3] The remaining distance RL (I) of the section I after proceeding during Δt is calculated.

RL (I) = L (I) −Δt × V (I, J × Δt) [STEP 4] It is checked whether the remaining distance RL (I) of section I is greater than zero.

If not, it is considered that section I has finished traveling, and the process proceeds to [STEP 5].

If it is larger, it is considered that the vehicle has not traveled in section I, and the following calculation is performed.
3].

L (I) = RL (I) K = K + 1 J = J + 1 [STEP 5] The state immediately before the end of section I running is calculated, and the required time T of section I is calculated in consideration of the result.
(I) is calculated.

ST ′ (I) = L (I) / V (I, J × Δt) (T ′ (I) <Δt) ST (I) = K × Δt + ST ′ (I) + ST ″ (I) , ST '(I), ST "(I) are ST' (I): time immediately before the end of section I running ST" (I): time in consideration of the amount of overshoot from the previous section. (I = the number of sections) is checked as to whether or not the calculation of the required time for all sections has been completed.

If all sections have not been completed, [S
TEP7].

If all sections have been completed, the required time of all sections is totaled by the following calculation, the required time of the target route section is calculated, and the processing is terminated.

[0069]

(Equation 7) [STEP 7] Section I in section I + 1
Overtime ST ″ (I + 1) taking into account excessive overshoot
Is calculated.

ST ″ (I + 1) = Δt−ST ′ (I) Also, the remaining distance RL (I +
1) is calculated.

RL (I + 1) = L (I + 1) -ST "
(I + 1) × V (I + 1, J × Δt) [STEP 8] When considering the overshoot of the section I, it is checked whether or not the section I + 1 has been passed.

If RL (I + 1)> 0 is satisfied,
Assuming that it has not passed through section I + 1, [STE
P12].

If RL (I + 1)> 0 is not satisfied, it is assumed that the signal passes through section I + 1, and [STEP]
9] Perform the following work. [STEP 9] The running distance L1 in the remaining time ST "(I + 1) is calculated.

L1 = ST ″ (I + 1) × V (I + 1, J × Δt) [STEP 10] It is checked whether or not the next section is exceeded.

If L1> L (I + 1) is satisfied, it is determined that the next section has not been exceeded, the following calculation is performed, and the process proceeds to [STEP 12].

RL (I + 1) = L (I + 1) -ST "
(I + 1) × V (I + 1, J × Δt) If L1> L (I + 1) is not satisfied, it is determined that the next section is exceeded, and the process proceeds to [STEP 11]. [STEP 11] An operation is performed when the value exceeds the next section I + 1.

First, I is updated.

I = I + 1 Next, the running time of the next section (because I has been updated, the next section is hereinafter referred to as I) is determined.

ST ′ (I) = L (I) / V (I, J × Δt) Next, the remaining time is obtained by the following arithmetic expression, and [STEP 9]
Proceed to.

ST ″ (I + 1) = ST ″ (I) −ST ′ (I) [STEP 12] Each variable is updated as follows.

L (I + 1) = RL (I + 1) J = J + 1 I = I + 1 [STEP 13] Go to [STEP 2].

As described above, [STEP 1] to [STEP 13]
The above is the flow of (b ″) required time actual value calculation means shown in FIG.

FIG. 14 shows an example of calculating the required time of section I when there is no overshoot of the next section (I + 1).

(E ″) is means for storing the calculated value of the actual required time value obtained from (b ″).

(N) is a means for storing the spatial average speed obtained from (a) the vehicle sensor.

The route selection support information providing device according to the sixth aspect of the present invention is characterized in that (a), (b ″), (c), (d),
(E ″), (f), (g), (h), (i), (j),
(K) and (n) are the same as those of the predicted congestion degree calculating device according to claim 5, and by installing them on a plurality of target routes, the predicted congestion degree of each route is calculated, and (l) ,
By (m), route selection support information is created. (L) and (m) are the same as those of the route selection support information providing device according to claim 2.

Further, the apparatus for creating a predicted congestion degree table according to claim 7 is similar to the apparatus for producing a predicted congestion degree according to claims 1 and 5 with respect to (a) and (n).
(O) The occupancy calculating means is means for calculating (a) an occupancy which is closely related to the traffic density from the measured value from the vehicle detector.

(P) is a means for calculating the traffic density based on the occupancy obtained by the occupancy calculation means and the average vehicle length.

Specifically, the calculation is performed by the following equation, for example.

Traffic density [vehicles / m] = occupancy [%]
/ (100 · average vehicle length [m / vehicle]) (q) is input to the neural network with the traffic density of a certain section and the traffic densities of a plurality of sections on the downstream side, and output the neural network as the spatial average of the section. This is a means for obtaining a speed prediction value. An example of the operation of the neural network is the case of the route selection support information providing device according to the first aspect. This process is performed in the same manner for each section.

(R) learns the weight count in the (q) neural network from the actually measured data. Specifically, the measured vehicle density of a certain section and the measured values of traffic density of a plurality of sections downstream thereof are input to a neural network, and the vehicle speed predicted value and the measured vehicle speed are compared. Is modified so that is minimized. The correction (learning operation) of the weight count is performed on various measurement data to improve the prediction accuracy of the neural network.

In the hierarchical neural network, which is a typical neural network, the detailed operation method of the present means is the same as the learning method of the hierarchical neural network described in claim 1.

(S) is based on the spatial average velocity predicted value obtained from the above (q) neural network.
This is a means for performing a prediction calculation of a change in spatial average speed (vehicle speed), a change in traffic density, and a change in traffic volume.

FIG. 15 shows an example of a flow representing the above processing flow. Further, the calculation contents of FIG. 15 will be described below. [STEP 1] (s) The time t managed in the traffic flow prediction means is set to the current time t0. [STEP 2] (p) By multiplying the traffic density of each section obtained from the traffic density calculation means by the section length, the number of vehicles E in each section j (j = 1, 2, 3,...) At the current time t0 (J,
t0) is calculated. ([STEP3]-[STEP1
0] is repeated for the required time. [STEP 3] Traffic density K of each section at time t
(J, t) is obtained by dividing the number of existing vehicles E (j, t) in each section by the section length L (j) of each section as in the following equation.

K (j, t) = E (j, t) / L (j) [STEP 4] Set the section number j to 0.

J = 0 (The operations of [STEP5] to [STEP6] are repeated by the required number of sections.) [STEP5] The section number j is advanced by one.

J = j + 1 [STEP 6] The traffic density at the time t of the adjacent sections j, j + 1, j + 2,... Is input to the (q) neural network as the traffic density on the section j and the downstream side thereof. Vehicle speed V (j,
t) is obtained. [Step 7] The section number j is set to 0. ([ST
The operations of [EP8] to [STEP10] are repeated by the required number of sections. [STEP 8] The section number j is advanced by one.

J = j + 1 [STEP 9] The traffic QIN (j, t to t + Δt) flowing into the section j between the time t and t + Δt is calculated by the following equation.

QIN (j, t to t + Δt) = V (j−
(1, t) · K (j−1, t) Also, the traffic volume QOUT (j, t to t) flowing out of the section j
+ Δt) is calculated by the following equation.

QOUT (j, t to t + Δt) = V (j,
t) · K (j, t) (However, a method of giving the traffic volume flowing from the end of the route using an average value of past traffic volume data is known.) [STEP 10] Time t + Δt Is calculated by the following equation using the number of existing vehicles E (j, t + Δt) in the section j.

E (j, t + Δt) = E (j, t) + QI
N (j, t to t + Δt) −QOUT (j, t to t + Δ
t) [STEP 11] Advance the time t by Δt. ([STE
After repeating the operations of [P3] to [STEP 11] for the required time, the process ends. (T) means for obtaining and storing a future transition of the vehicle speed in each section from the spatial average speed obtained from the (s) traffic flow prediction means.

(L ') is a means for creating a predicted congestion degree table as shown in FIG. 16 based on the transition of the spatial average speed obtained from the (t) vehicle speed transition prediction storage means.

The route selection support information providing device according to claim 8 is characterized in that (a), (n), (o), (p), (q),
(R), (s), (t), and (l ') are the same as those in the predicted traffic congestion degree table creating apparatus according to claim 7, and by installing them on a plurality of target routes, A predicted congestion degree table is created, and route selection support information is created by (m). (M) is the same as the route selection support information providing device according to claim 2.

FIG. 17 is a table showing a calculation example in which the actual required time value is calculated by the method of the present invention from the required section passage time when the target section is divided into sections of several hundred meters.

In this table, the vertical axis represents time, and the horizontal axis represents each section. At a certain time, for example, 0:00, the time required for passing through each section is 2,
4, 6, 3, 2, 2, 1 minute. The method of summing the respective times is a conventional required time calculation method. However, this is completely different from the time required when the specific vehicle sequentially passes from section 1 to the subsequent sections.

In the present invention, a viewpoint centered on the running of a specific vehicle is taken and the following calculation is made. It is now assumed that the time is 0:00, and the vehicle in section 1 passes to section 7. It takes 2 minutes for this vehicle to pass through section 1 and 4 minutes to pass through section 2
Takes minutes. Therefore, when the vehicle reaches the section 3, six minutes have passed since the vehicle started running in the section 1, and the vehicle has reached 0:06.

Therefore, in order to know the passage time of section 3, it is necessary to look at section 3 at 0:05 in the chart of FIG. This is 5 minutes. Therefore, when the vehicle enters the next section 4, the ratio is 0:11. So, looking at section 4 at 0:10, it's 4 minutes. In this way, the section passing time of the specific vehicle is obtained, and when these are integrated, the time required for passing through all sections is obtained. In the example of the table of FIG. 17, 2 + 4 + 5 + 4 + 4 + 6 + 6 = 31 minutes.

That is, the frame surrounded by the thick line in FIG. 17 is the traffic flow when the specific vehicle runs. Therefore, within the range shown in this diagram, the specific vehicle is in a relatively bad flow in sections 1 to 7.
That is why I ran up to.

On the other hand, for example, 0:15 or 0: 2
A vehicle that starts running in section 1 around 0 can pass through all sections in a shorter time.

[0110]

BEST MODE FOR CARRYING OUT THE INVENTION

(Configuration of First Embodiment) FIG. 1 is a block diagram showing an embodiment of the present invention. In FIG. 1, a vehicle sensor 1 (installed in each section on a road), AV
I system 2 (installed at both ends of target section), section required time calculation means 3, section required time storage means 4, required time actual value storage means 5, hierarchical neural network for prediction 6, weighting coefficient updating means 7 Learning hierarchical neural network 8, hierarchical neural network learning means 9, prediction start timing determining means 1
0, the estimated congestion degree calculating means 11.

(Operation of the First Embodiment) First, the section required time is calculated by the section required time calculating means 3 using the spatial average speed measured by the vehicle sensor 1, and the required section time is stored. 4 is stored. Similarly, AV
The required time actual value is measured by the I system 2 and stored in the required time actual value storage means 5. The prediction hierarchical neural network 6 receives the latest section required time data obtained by the section required time calculating means 3 as an input,
Using the weighting factor updated by the weighting factor updating means 7, a required time prediction value is calculated.

The weight coefficient used at this time is updated using the weight coefficient correction amount calculated by the hierarchical neural network learning means 9. The hierarchical neural network learning means 9 uses the actual required time value stored in the required actual time value storage means 5 as a teacher signal and the section required time data stored in the required section time storage means 4 as an input. Learning is performed by comparing the output with the output of the neural network 7, and the weight coefficient correction amount is calculated. This correction amount is passed to the weighting factor updating means 7, and the weighting factor is updated.

The prediction start timing determining means 10 determines whether it is appropriate to use the output of the hierarchical neural network 7 for prediction as a predicted value. Specifically, at the beginning of online, the prediction calculation is turned off,
The weight coefficient is updated by a learning operation. Then, the required time prediction error in the hierarchical neural network learning means 9 (the output of the learning neural network 8 for learning,
When the required time actual value storage means 5 (difference from the actual value) falls within the set range, the prediction start timing judgment means 10 turns on the prediction calculation.

Finally, the predicted congestion degree calculating means calculates the predicted congestion degree based on the required time predicted value obtained from the prediction start timing judging means 10.

The weight update in the operation of the present embodiment is performed for each sample (for example, every 5 minutes). The section required time data and the required time actual value data used for learning are n samples before (for example, one sample). For 5 minutes,
If it is 12 samples before, learning is performed using the data 60 minutes before. ) Data will be used. (Effects of First Embodiment) According to the first embodiment, (1)
By using the hierarchical neural network, the traffic situation is learned by the hierarchical neural network, so that a predicted congestion degree indicating a future change in the congestion state can be created.

(2) Since the learning of the neural network using the actual measurement data is performed sequentially, it is possible to continuously improve the model accuracy in response to the secular change such as the route and the economic condition. It has the effect of. (Configuration of Second Embodiment) FIG. 2 is a block diagram of a second embodiment of the present invention. In FIG. 2, roads 1 to m
Are provided with the predicted congestion degree calculation devices 12A-1 to 12-n shown as the first embodiment, the predicted congestion degree table creation means 13-1 to m, and the route selection support information providing means 14 for the roads 1 to m. ing. (Operation of the Second Embodiment) First, the predicted congestion degree of each route is calculated by the predicted congestion degree calculation devices 12A-1 to 12n installed in each section of each route. Next, on the basis of the predicted congestion degree of each section calculated by the predicted congestion degree calculation devices 12A-1 to 12A-n installed in each section, the predicted congestion degree table creating means 13-1 to 13-1 installed on each route. Based on m, a predicted congestion degree table for each route is created.

Finally, based on the result of comparing the predicted congestion degree tables of the respective routes, the route selection support information providing means 14 determines whether the route is likely to be reduced in the future or the route is likely to be reduced in the future. Create route selection support information such as "". (Effects of Second Embodiment) As described above, in the second embodiment, in addition to the effects of the first embodiment described above,
(3) Based on the predicted congestion degrees of a plurality of routes, it is possible to provide support information in consideration of prediction of future traffic conditions (congestion conditions) as information for selecting a route.

This has the effect of: (Configuration of Third Embodiment) FIG. 3 is a block diagram showing a third embodiment of the present invention. Vehicle detector 1 (installed in each section on the road), required time actual value calculation means 11, section required time calculation means 3, section required time storage means 4, required time actual value storage means 5a, for prediction It is composed of a hierarchical neural network 6, a weight coefficient updating means 7, a learning hierarchical neural network 8, a hierarchical neural network learning means 9, a prediction start timing determining means 10, and a predicted congestion degree calculating means 11. (Operation of Third Embodiment) Next, the operation of the third embodiment of the present invention will be described. First, using the spatial average speed measured by the vehicle sensor 1, the section required time calculating means 3
The section required time is calculated and stored in the section required time storage means 4. Based on the section required time obtained from the section required time storage means, the required time actual value calculation means 11 calculates the required time actual value and stores it in the required time actual value storage means 5a.

FIG. 6 is a calculation example of the required time actual value calculation means 11. The prediction hierarchical neural network 6 receives the latest section required time data obtained by the section required time calculating means 3 as input, and calculates a required time predicted value using the weight coefficient updated by the weight coefficient updating means 7. .

The weight coefficient used at this time is updated using the weight coefficient correction amount calculated by the hierarchical neural network learning means 9. In the hierarchical neural network learning means 9, the required time actual value storage means 5a
The learning is performed by comparing the required time actual value stored in the section required time data as a teacher signal with the output of the learning hierarchical neural network 7 that receives the required time data of the section stored in the required time storage section 4 as input. Calculate the coefficient correction amount.

This correction amount is passed to weight coefficient updating means 7, and the weight coefficient is updated. Further, the prediction start timing determining means 10 determines whether it is appropriate to use the output of the hierarchical neural network 7 for prediction as a predicted value. Specifically, the forecast calculation is initially OF
F, and the weight coefficient is updated by a learning operation.

The required time prediction error in the hierarchical neural network learning means 9 (the deviation between the output of the learning hierarchical neural network 8 for learning and the actual value in the required time actual value storage means 5a) is set within a set range. When it is settled, the prediction calculation is turned ON by the prediction start timing determination means 10.

Finally, the predicted congestion degree calculating means calculates the predicted congestion degree based on the required time predicted value obtained from the prediction start timing judging means 10.

The weight update in the operation of the present embodiment is performed for each sample (for example, every 5 minutes). The section required time data and the required time actual value data used for learning are n samples before (for example, one sample). 5 minutes, 1
If it is two samples before, learning will be performed using the data 60 minutes before. ) Data will be used. (Effects of Third Embodiment) As described above, in the third embodiment, in addition to the effects of the first embodiment described above, (3) means for calculating the actual required time value is provided. Therefore, it is possible to adapt to routes without an AVI system. It has the effect of. (Structure of Fourth Embodiment) FIG. 4 shows a fourth embodiment of the present invention. The basic configuration is the same as that of the second embodiment. Instead of the predicted congestion degree calculating devices 12A-1 to 12-n in each section of the roads 1 to m, the roads 1 to m Predicted congestion degree calculating devices 12B-1 to 12B according to claim 3 for each section.
It consists of. (Operation of the Fourth Embodiment) Next, the operation of the fourth embodiment of the present invention will be described. First, the predicted congestion degree of each route is obtained by the predicted congestion degree calculation devices 12B-1 to 12B installed in each section of each route. Next, based on the predicted congestion degree of each section calculated by the predicted congestion degree calculation devices 12B-1 to 12B-n installed in each section, the predicted congestion degree table creating means 13-1 to 13-1 installed on each route. Based on m, a predicted congestion degree table for each route is created. Finally, based on the result of comparing the predicted congestion degree tables of the respective routes, the route selection support information providing unit 14 determines that the route is a route that will be less congested in the future.
For example, route selection support information such as "routes in which traffic congestion tends to worsen" is created. (Effects of Fourth Embodiment) As described above, in the fourth embodiment, in addition to the effects of the third embodiment, (4) information for selecting a route based on the predicted congestion degrees of a plurality of routes. As a result, it is possible to provide support information in consideration of prediction of future traffic conditions (congestion conditions). It has the effect of. (Structure of Fifth Embodiment) FIG. 5 is a block diagram showing a fifth embodiment of the present invention. Vehicle sensor 1 (installed in each section on the road), spatial average speed storing means 12, required time actual value calculating means 11a, section required time calculating means 3, section required time storing means 4, required time actual It is composed of a value storage means 5b, a hierarchical neural network for prediction 6, a weight coefficient updating means 7, a hierarchical neural network for learning 8, a hierarchical neural network learning means 9, a prediction start timing determining means 10, and a predicted congestion degree calculating means 11. Is done. (Operation of Fifth Embodiment) Next, the operation of the fifth embodiment of the present invention will be described. First, the section required time is calculated by the section required time calculating means 3 using the spatial average velocity measured by the vehicle sensor 1 and stored in the section required time storing means 4. At this time, the spatial average speed measured by the vehicle sensor 1 is simultaneously stored in the spatial average speed storage unit 12.

The actual time required value calculating means 1 is calculated by using the spatial average speed obtained from the spatial average speed storing means 12 storing the spatial average speed measured by the vehicle sensor 1.
In 1a, the required time actual value in consideration of the overshoot between sections is calculated and stored in the required time actual value storing means 5b.
The prediction hierarchical neural network 6 receives the latest section required time data obtained by the section required time calculating means 3 as input, and calculates a required time predicted value using the weight coefficient updated by the weight coefficient updating means 7. .

The weight coefficient used at this time is updated using the weight coefficient correction amount calculated by the hierarchical neural network learning means 9. In the hierarchical neural network learning means 9, the required time actual value storage means 5b
The learning is performed by comparing the required time actual value stored in the section required time data as a teacher signal with the output of the learning hierarchical neural network 7 that receives the required time data of the section stored in the required time storage section 4 as input. Calculate the coefficient correction amount.

This correction amount is passed to the weight coefficient updating means 7, and the weight coefficient is updated. Further, the prediction start timing determining means 10 determines whether it is appropriate to use the output of the hierarchical neural network 7 for prediction as a predicted value. Specifically, the forecast calculation is initially OF
F, and the weight coefficient is updated by a learning operation.

Then, the required time prediction error in the hierarchical neural network learning means 9 (the deviation between the output of the learning hierarchical neural network 8 for learning and the actual value in the required time actual value storing means 5b) falls within the set range. Then, the prediction calculation is turned ON by the prediction start timing determination means 10.

Finally, the predicted congestion degree calculating means calculates the predicted congestion degree based on the required time predicted value obtained from the prediction start timing judging means 10.

The weight update in the operation of this embodiment is performed for each sample (for example, every 5 minutes), and the section required time data and the required time actual value data used for learning are n samples before (for example, one sample). 5 minutes, 1
If it is two samples earlier, learning will be performed using the data 60 minutes ago). (Effects of Fifth Embodiment) As described above, in the fifth embodiment, in addition to the effects of the third embodiment, (4) the actual required time is calculated based on the spatial average speed in consideration of overshoot between sections. Since the value is calculated, a more exact required time can be calculated. Since the predicted congestion degree is created based on this, a more reliable predicted congestion degree can be created. It has the effect of. (Structure of Sixth Embodiment) A sixth embodiment of the present invention will be described. The basic configuration is the same as that of the second embodiment. Instead of the predicted congestion degree calculating devices 12A-1 to 12-n in each section of the roads 1 to m, the roads 1 to m Each section is provided with the predicted congestion degree calculation devices 12C-1 to 12C-n described in claim 5. (Operation of Sixth Embodiment) Next, the operation of the sixth embodiment of the present invention will be described. First, the predicted congestion degree of each route is calculated by the predicted congestion degree calculation devices 12C-1 to 12C installed in each section of each route. Next, on the basis of the predicted congestion degree of each section calculated by the predicted congestion degree calculation devices 12C-1 to 12C-n installed in each section, the predicted congestion degree table creating means 13-1 to 13-1 installed on each route. Based on m, a predicted congestion degree table for each route is created. Finally, based on the result of comparing the predicted congestion degree tables of the respective routes, the route selection support information providing means 14 uses the route selection support information providing means 14 such as "Routes in which congestion tends to be reduced in the future" or "Routes in which congestion tends to be worse in the future" Create route selection support information. (Effects of Sixth Embodiment) As described above, in the sixth embodiment, in addition to the effects of the fifth embodiment, (5) information for selecting a route based on predicted congestion degrees of a plurality of routes. As a result, there is an effect that it is possible to provide support information in consideration of prediction of a future traffic situation (congestion situation). (Configuration of Seventh Embodiment) FIG. 7 is a block diagram showing a seventh embodiment of the present invention. Vehicle detector 1, installed in each section on the road, spatial average speed storage means 16a, occupancy calculation means 17, traffic density calculation means 18, traffic flow prediction means 19, neural network 2
0, a neural network learning means 21, a vehicle speed transition prediction storage means 22, and a predicted congestion degree table creation means 23. (Operation of Seventh Embodiment) Next, the operation of the seventh embodiment of the present invention will be described. First, regarding the vehicle sensor 1 and the spatial average speed storage means 16a, claim 1 and claim 5
The occupancy calculating means 17 calculates the occupancy which is closely related to the traffic density from the measurement value from the vehicle sensor 1 in the same manner as the predicted congestion degree creating device described.

Next, the traffic density creating means 18 is means for calculating the traffic density based on the occupancy obtained by the occupancy calculating means 17 and the average vehicle length. On the basis of this traffic density, the neural network 20 inputs the traffic density of a certain section and the traffic densities of a plurality of sections downstream thereof to the neural network 20, and obtains a predicted value of the spatial average speed of the section as the output of the neural network. . This process is similarly performed for each section to obtain a predicted value of the spatial average velocity of each section.

Also, the neural network learning means 2
In step 1, the weight count in the neural network 20 is learned from the actually measured data. More specifically, a vehicle speed predicted value obtained when a measured value of traffic density in a certain section and a measured value of traffic density in a plurality of sections downstream thereof are input to the neural network 20 are compared with actual measured values of vehicle speed. Modify the weight count so that the error is minimized. Such correction of the weight count (learning calculation) is performed on various measurement data, and the prediction accuracy of the neural network 20 is improved.

[0133] The traffic flow prediction means 19 calculates the prediction of the transition of the spatial average speed (vehicle speed), the transition of the traffic density and the transition of the traffic volume, based on the predicted value of the spatial average speed obtained from the neural network 20 described above. Do.

The vehicle speed transition prediction storage means 22 obtains and stores the future transition of the vehicle speed in each section from the spatial average speed obtained from the traffic flow prediction means 19.

Finally, the predicted congestion degree table creating means 23
, A predicted congestion degree table is created based on the transition of the spatial average speed obtained from the vehicle speed transition prediction storage means 22. (Effect of Seventh Embodiment) As described above, in the seventh embodiment, (1) the neural network learns the traffic situation by using the neural network, so that the change of the traffic congestion situation in the future can be monitored. (2) Neural network learning using actual measurement data is performed sequentially, so that it can continuously cope with the secular change of the route, economic conditions, etc. (3) One route is divided into several sections, and the traffic flow can be predicted up to a certain time ahead for all sections. (Configuration of Eighth Embodiment) An eighth embodiment of the present invention will be described.
For the roads 1 to m, the predicted congestion degree table creation devices 24-1 to 24-1
m (this is a predicted traffic congestion degree table creation device corresponding to (claim 7) of the present invention, and is installed on each target route), and is constituted by route selection support information providing means 14a. (Operation of the Eighth Embodiment) First, the predicted congestion degree tables of the respective routes are prepared by the predicted congestion degree table preparing devices 24-1 to m installed on the respective routes. Based on the result of comparing the predicted congestion degree tables of the respective routes based on this result, the route selection support information providing means 14a indicates that the routes on which the traffic congestion will tend to be reduced and the routes on which the congestion will tend to worsen in the future. And route selection support information such as "time until a certain route is cleared of congestion" and "time when non-congestion continues in a certain route". (Effects of Eighth Embodiment) As described above, in the eighth embodiment, in addition to the effects of the seventh embodiment described above,
(4) Based on the predicted congestion degrees of a plurality of routes, it is possible to provide support information in consideration of prediction of future traffic conditions (congestion conditions) as information for selecting a route.

(5) Since the predicted traffic congestion degree table up to a certain period of time can be used, it is possible to create route selection support information in consideration of further changes in traffic conditions. It has the effect of. (Other Embodiments) In the above embodiment, the actual required time value is measured by the AVI system, or is calculated from the section required time and the spatial average speed, and is used as a teacher signal for the neural network. In the case of a section, the actual required time value may be obtained as follows and used for calculating the predicted congestion degree.

By actually running the test vehicle at regular time intervals, the required travel time between each point is measured.

[0138]

As described above, the present invention has the following effects.

(1) By using the neural network, the traffic condition is learned by the neural network, so that a predicted congestion degree indicating a future change in the congestion condition can be created.

(2) Since the learning of the neural network using the actual measurement data is performed sequentially, it is possible to continuously improve the model accuracy in response to the secular change of the route, the economic condition, and the like.

(3) Based on the predicted congestion degrees of a plurality of routes, information on traffic selection (traffic congestion state) as information for selecting a route.
Support information in consideration of the prediction of

(4) Since the means for calculating the actual required time value is provided, it is possible to adapt to routes without an AVI system.

(5) Since the actual required time value is calculated from the spatial average speed in consideration of overshoot between sections,
More strict required time can be calculated. Since the predicted congestion degree is created based on this, a more reliable predicted congestion degree can be created.

(6) One route is divided into several sections, and it is possible to predict the traffic flow up to a certain time ahead for all sections.

(7) Since the predicted traffic congestion degree table up to a certain period of time can be used, it is possible to create route selection support information in consideration of further changes in traffic conditions.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a block diagram showing a second embodiment of the present invention.

FIG. 3 is a block diagram showing a third embodiment of the present invention.

FIG. 4 is a block diagram showing a fourth embodiment of the present invention.

FIG. 5 is a block diagram showing a fifth embodiment of the present invention.

FIG. 6 is a block diagram showing a sixth embodiment of the present invention.

FIG. 7 is a block diagram showing a seventh embodiment of the present invention.

FIG. 8 is a block diagram showing an eighth embodiment of the present invention.

FIG. 9 is a general configuration diagram of a hierarchical neural network.

FIG. 10 is a flowchart of a learning method of a hierarchical neural network.

FIG. 11 is an example of creating a predicted congestion degree table created in the second, fourth, and sixth embodiments of the present invention.

FIG. 12 is a flowchart for calculating a required time actual value from a section required time.

FIG. 13 is a flowchart for calculating a required time actual value from a section average speed.

FIG. 14 is a calculation example of the required time of section I when there is no passage of the next section.

FIG. 15 is a flowchart showing the calculation contents of a traffic flow prediction means which is a part of FIG. 7;

FIG. 16 is an example of creating a predicted congestion degree table created in the seventh embodiment of the present invention.

FIG. 17 is an explanatory diagram for calculating a required actual time value from a section required time.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Vehicle sensor 2 AVI system 3 Section required time calculation means 4 Section required time storage means 5, 5a, 5b Required time actual value storage means 6 Prediction hierarchical neural network 7 Weight coefficient updating means 8 Learning hierarchical neural network 9 Hierarchical neural network learning means 10 Prediction start timing judgment means 11, 12A, 12B, 12C Predicted congestion degree calculating means 13, 23, 24 Predicted congestion degree table creation means 14, 14a Route selection support information providing means 15, 15a Required time results Value calculation means 16 Spatial average speed storage means 17 Occupancy calculation means 18 Traffic density calculation means 19 Traffic flow prediction means 20, 28 Neural network 21, 29 Neural network learning means 22 Vehicle speed transition prediction storage means 25 Riding rate storage means 6 occupancy table creation means 27,27a ride the train selection support information providing unit 30 prediction occupancy table creation means

Claims (8)

[Claims] A plurality of vehicle detectors installed in a predetermined closed section of an automobile road to detect a spatial average speed of a vehicle traveling in the closed section; An AVI system for determining a time required for the vehicle to travel in a closed section, and a time required in each of the sections obtained by dividing the target closed section using a spatial average speed from the vehicle detector. Section required time calculating means for calculating; section required time storing means for holding section required time from the section required time calculating means; required time actual value storing means for storing required time actual value from the AVI system; The section required time from the section required time calculating means is input and a required time predicted value calculated based on the weight coefficient is output. A layered neural network for measurement, a layered neural network for learning that receives section required time data from the section required time storage means and outputs a predicted value of required time calculated based on a weighting factor, and the hierarchical layer for prediction. Weighting factor updating means for updating the weighting coefficients used in both the neural network and the learning hierarchical neural network; and a required time predicted value output from the learning hierarchical neural network, and the required time actual value storing means. A neural network learning means for calculating the correction amount of the weighting coefficient using the required time actual value, and a prediction for determining a timing to start prediction based on a required time prediction error from the hierarchical neural network learning means. Start timing determining means; Serial prediction starting based on the required time predicted value from the timing determining means, route selection assistance information providing apparatus equipped with the prediction congestion degree calculating means for calculating a predicted congestion degree. 2. The route selection support information providing device according to claim 1, wherein, instead of the AVI system and the required actual time value storage means, a section required time stored in the section required time storage means is used. A required time actual value calculating means for calculating a required time actual value calculated value in consideration of a lapse of time due to traveling; a required time actual value storing means for storing a required time actual value obtained from the required time actual value calculating means; A route selection support information providing device, comprising: 3. A route selection support information providing device that divides a plurality of motorway lines into several sections and calculates a predicted congestion degree for each of the divided sections, and a route selection support information providing device installed in each section. A predicted congestion degree table creating means for creating a predicted congestion degree table based on the predicted congestion degree of each section obtained from the section, and a predicted congestion degree table for each route obtained from the predicted congestion degree table creating means installed on each route. A route selection support information providing device that creates route selection support information based on the route selection support information providing device. 4. The route selection support information providing device according to claim 3, further comprising a predicted congestion degree calculating device for dividing the plurality of routes into several sections and calculating a predicted congestion degree for each of the divided sections. Route selection support information providing device. 5. The predicted congestion degree calculating device according to claim 1, wherein a space average speed storage unit for storing a space average speed obtained from the vehicle sensor, instead of the AVI system and the required time actual value storage unit. From the spatial average speed obtained by the spatial average speed storage means, taking into account the passage of time due to traveling, a required time actual value calculation means for calculating a more exact required time actual value, and the required time actual value calculation means An estimated traffic congestion degree calculating device, comprising a required time actual value storage means for storing an obtained required time actual value. 6. The route selection support information providing device according to claim 2, wherein a plurality of routes are divided into several sections, respectively, instead of the predicted congestion degree calculating apparatus according to claim 1, and the predicted congestion of each divided section is divided. A route selection support information providing device provided with the predicted congestion degree calculation device according to claim 5 for calculating a degree. 7. An occupancy which is closely related to traffic density is calculated by using a plurality of vehicle sensors installed in a closed section set on an automobile road and data obtained from the vehicle sensors. An occupancy calculating means; a traffic density calculating means for calculating a traffic density using the occupancy obtained from the occupancy calculating means; a spatial average speed storing means for storing a spatial average speed obtained from the vehicle sensor; A neural network that receives the traffic density (or occupancy) and the traffic density (or occupancy) of a plurality of sections downstream of the section and calculates a predicted spatial average speed of the section, and a traffic density actual value (or occupancy) of the section Actual values), actual traffic density values (or actual occupancy values) for multiple sections downstream of the section, and And a neural network learning means for learning the above-mentioned neural network from the actual value of the spatial average speed of the section, and a transition of the spatial average speed (vehicle speed) and traffic based on the spatial average speed predicted value obtained from the neural network. A traffic flow prediction unit that performs a prediction calculation of a change in density and a change in traffic volume; a vehicle speed change prediction storage unit that stores a change in the spatial average speed obtained from the traffic flow prediction unit; and a vehicle speed change prediction storage unit. A route selection support information providing device, comprising: a predicted congestion degree table creating unit that creates a predicted congestion degree table based on the obtained transition of the spatial average speed. 8. A predicted traffic congestion degree table creating device for calculating a predicted traffic congestion degree table for each route of a plurality of motorways, and a predicted traffic congestion degree table for each route obtained from the predicted traffic congestion degree table producing device. A route selection support information providing device comprising: a route selection support information providing unit for generating route selection support information;