Patent Application
20170241778
1. Field of the Invention
The present invention relates to a method for predicting a condition of a road surface on which a vehicle is traveling, and, more particularly, to a method for predicting a road surface condition using road surface condition data estimated by vehicles traveling through a location within a predetermined range and vehicular information.
2. Description of the Related Art
In order to raise the travel safety of a vehicle, it is desired to accurately estimate the condition of the road surface on which the vehicle is traveling and to have the information fed back to vehicle control. If the road surface condition can be estimated, the safety of vehicular driving can be enhanced markedly since an advanced control of ABS braking, for instance, can be effected before the driver initiates a danger-avoiding control, such as braking or steering.
As methods having been proposed for estimating the condition of a road surface under a traveling vehicle, there are methods for estimating the condition of a road surface by detecting the vibration of the tire of the traveling vehicle and estimating the road surface condition from the time-series waveform of the detected tire vibration (see Patent Documents 1 to 3, for instance) and methods for estimating a road surface condition from a detected sound pressure level of tire noise by detecting tire noise arising from a tire (see Patent Document 4, for instance).
However, the known technologies as listed above allow prediction of the condition of a road surface on which a vehicle is traveling, but find it difficult to predict changes in road surface conditions for the location within a predetermined range from the estimation results of road surface conditions.
The present invention has been made to solve the foregoing problems, and an object of the invention is to provide a method for accurately predicting a road surface condition at a location within a predetermined range.
A method for predicting a road surface condition according to an embodiment of the present invention provides a method including the steps of obtaining vehicular information, which is information on the behavior of a vehicle during travel by an on-board sensor mounted on the vehicle and predicting a road surface condition at a location within a predetermined range, using road surface estimation decision values to be used in the estimation of road surface conditions, which are calculated using the vehicular information or estimated road surface conditions estimated using the road surface estimation decision values. In the step of predicting a road surface condition, predicted incidence rates SRp for respective road surface conditions, which are the incidence rates of road surface conditions at the location within the predetermined range, are calculated from time-dependent changes in the road surface estimation decision values calculated using the vehicle information obtained by the vehicles having passed the location within the predetermined range or time-dependent changes in the estimated road surface conditions, and then the road surface condition at the location within the predetermined range is predicted from the calculated predicted incidence rates SRp.
A server 20 includes a receiver 21, a data storage means 22, and a transmitter 23. A road surface condition predicting unit 30 predicts road surface conditions at the location within a predetermined range from the time-dependent changes in estimated road surface conditions estimated by a plurality of vehicles W at a plurality of times at the location within the predetermined range.
It is to be noted that the “number N of vehicles” is a “gross total of vehicles”. That is, when there is a plurality of data from a single vehicle within a predetermined space of time and at the location within a predetermined range, the plurality of these data are processed as separate data.
The server 20 and the road surface condition predicting unit 30 are installed in a road surface condition management center 2.
The road surface condition estimating means 13, the vehicular information collecting means 14, and the road surface condition predicting unit 30 are constituted, for instance, by computer software.
The acceleration sensor 11 is disposed approximately in the middle portion on the tire air chamber 42 side of the inner liner 41 of the tire 40 as shown in
The GPS unit 12, equipped with a not-shown GPS antenna and receiver, acquires position data of the vehicles Wi and calculates the travel speeds of the vehicles from the position data thereof.
The road surface condition estimating means 13 estimates the road surface under a traveling vehicle to be any one of DRY road surface, WET road surface, SNOW road surface, and ICE road surface, using the time-series waveform of tire vibration detected by the acceleration sensor 11. As a road surface condition estimating means 13 as mentioned above, there may be one which includes a vibration waveform extracting means 131 for extracting time-series waveforms of tire vibration from the acceleration sensor 11, a windowing means 132, a feature vector calculating means 133, a storage means 134 for storing four road surface models, a kernel function calculating means 135, and a road surface condition determining means 136 as shown in
Note that the road surface condition estimating means 13 may be installed inside the tire 40 or on the vehicle body. When the road surface condition estimating means 13 is installed on the vehicle body, the arrangement should preferably be such that the data of the acceleration waveform detected by the acceleration sensor is not sent from the tire 40 side to the vehicle body side, and a calculating unit is provided on the tire 40 side and band values to be used in the estimation of road surface conditions, such as the vibration levels in specific frequency bands detected from the acceleration waveform, or calculated values of the band values are sent to the road surface condition estimating means 13.
The vehicular information collecting means 14 collects the road surface conditions estimated by the road surface condition estimating means 13 (hereinafter referred to as “estimated road surface conditions”) and the position data of the vehicle obtained by the GPS unit 12 and sends them as vehicular information, together with the vehicle ID for identifying the vehicle, to the transmitter 15. The vehicular information includes the vehicle ID and the data on the time when the vehicular information is obtained (time data). It is to be noted that the time data to be used may be the time of extracting the time-series waveform of acceleration, the time of obtaining the position data, or the time of transmitting the data. These times, which are almost simultaneous, pose no problem as time data.
The transmitter 15 transmits the vehicular information, together with the estimated road surface condition data, position data of the vehicle, and vehicle ID for identifying the vehicle collected by the vehicular information collecting means 14, from a not-shown transmission antenna to the server 20 in the road surface condition management center 2.
The receiver 16 receives prediction data on the road surface condition at the location within the predetermined range predicted by the road surface condition predicting unit 30 in the road surface condition management center 2 and transmitted from the transmitter 23 of the server 20. It is to be noted that the received prediction data on a road surface condition may be displayed on a monitor provided in the vehicle so that the driver may be informed of the predicted result of the road surface condition for the location within the predetermined range.
Also, the predicted result of the road surface condition may be fed back to vehicle control, thereby improving the travel safety of the vehicle.
The server 20 receives the vehicular information, including the data on estimated road surface conditions, sent from the respective vehicles W1 (i=1 to N) by its receiver 21, classifies these data into data at the respective times for the location within the predetermined range, stores them in the data storage means 22, and transmits the prediction data on the road surface condition for the location within the predetermined range predicted by the road surface condition predicting unit 30 to the registered vehicles.
The vehicular information of the respective vehicles W1 is classified and stored in the data storage means 22.
More specifically, as shown in
It should be understood here that m1+m2+ . . . +mr+ . . . =N. Note also that the number nr of the times at which data is obtained may vary with locations Lr or be the same for all the locations Lr.
The data at the predetermined locations Lr are stored in order of time.
It is to be noted that the data at time tk refers to the data obtained within a predetermined time width Δtk including time tk (e.g., tk−Δtk/2≦t≦tk Δtk/2). The predetermined time width Δtk may not necessarily be fixed, but may vary with the predetermined locations.
The predetermined time width Δtk may be 1 to 5 minutes, for instance. Also, the “predetermined range” as used herein refers to a range that includes a pre-set location on a road map. Designation of a predetermined range may be made using a grid of appropriate size on a navigation road map as shown in
Also, the registered vehicles include not only the vehicles Wi carrying the road surface condition estimating means 13, but also any vehicles in communication with the server 20.
The road surface condition predicting unit 30 includes a statistical data generating means 31, an actual statistical data storage means 32, a predicted statistical data generating means 33, and a road surface condition predicting means 34. The road surface condition predicting unit 30 predicts road surface conditions at the location Lr within a predetermined range at future time tp (p>0), using the vehicular information obtained at t−n to time t−1.
The statistical data generating means 31 takes out the mr units of data on estimated road surface conditions obtained at the location Lr within the predetermined range at time t(r)k (k=−n1 to −1) before the present time t0 from the data stored in the data storage means 22 of the server 20, counts the numbers of the vehicles having estimated the DRY road surface, WET road surface, SNOW road surface, and ICE road surface, respectively, using these data, and generates maps tallying up the incidence rates of the estimated road surface conditions R at the location Lr within the predetermined range at time t(r)k for each of the estimated road surface conditions R (R: DRY road surface, WET road surface, SNOW road surface, and ICE road surface) (hereinafter referred to as actual statistical data M(r)k). It is to be noted that when a plurality of data are sent from a single vehicle within a predetermined space of time and at a location within a predetermined range, the data are processed as those from separate vehicles. In such a case, the number of vehicles is the gross total of the vehicles. Note also that whether the data is from the same vehicle or not can be determined by referring to the vehicle ID.
Prediction of a road surface condition is done for each of the locations Lr. However, the following description will be given of a case where the road surface condition at a location Lr is predicted. Note therefore that hereinbelow the suffix r indicating the location Lr is omitted such that the vehicle passing through the location Lr at time tk is denoted by Wj (j=1 to m), and the actual statistical data by Mk.
The actual statistical data Mk are the incidence rates SRk of estimated road surface conditions R (R: DRY, WET, SNOW, ICE) at time tk (k=−n1 to −1) calculated for each of the road surface conditions R. The predicted statistical data MpC are the predicted values SRp of incidence rates at a future time tp (p>0) (hereinafter referred to as predicted incidence rates) calculated for each of the road surface conditions R.
The incidence rates SRk are calculated by the formula: SRk=(count of an estimated road surface condition R)/(total count). Note that the actual statistical data Mk is generated for each of the times tk.
Generation of the actual statistical data Mk is normally done immediately after the storage of estimated road surface condition data at time t−1, which is the time immediately before the time tp at which a prediction is made.
The actual statistical data storage means 32 stores the actual statistical data Mk generated by the statistical data generating means 31 in order of time.
The predicted statistical data generating means 33 generates the predicted statistical data MpC for the location Lr within the predetermined range at a future time tp (p>0), using a plurality (10 here) of actual statistical data Mk arranged in a time series as shown in
More specifically, the time-dependent changes in the incidence rates SRk of estimated road surface conditions R at times t−10 to are respectively approximated by the n-th functions Gr(t) (n≧3), and the function values GR(tp) of the approximate functions (n-th functions) at t=tp are derived respectively. Then the predicted statistical data MpC are generated by calculating the predicted incidence rates SRp of road surface conditions R at time tp, using these four function values GR(tp), respectively.
The predicted incidence rates SRp are derived by dividing the function values GR(tp) of the respective road surfaces R by the sum (ΣRGR(tp)) of the four function values GR(tp).
That is, the predicted incidence rate of the DRY road surface is calculated by SDp=GD (tp)/ΣRGR(tp). And the predicted incidence rate of the WET road surface is calculated by SWp=GW(tp)/ΣRGR(tp). Likewise, the predicted incidence rate of the SNOW road surface is calculated by SSp=GS(tp)/ΣRGR(tp). And the predicted incidence rate of the ICE road surface is calculated by SIp=GI(tp)/ΣRGR(tp). The sum of the predicted incidence rates SDp, SWp, SSp, and SIp is 1.
The road surface condition predicting means 34 predicts the road surface condition R to be any one of the DRY road surface, WET road surface, SNOW road surface, and ICE road surface. More specifically, the road surface condition predicting means 34 selects an estimated road surface condition R indicating the highest predicted incidence rate SRp (DRY road surface in
By repeating the prediction of a road surface condition using prediction results of road surface conditions like this, the road surface condition R(tq) at a further future time tq (tq>tp) can be predicted.
Next, the operation of a road surface condition predicting system 1 is described with reference to the flowchart of
Firstly, estimated road surface conditions and vehicular information, such as vehicle positions, are obtained by the participating vehicles Wi (i=1 to N) (step S10). And these pieces of information, together with the vehicle IDs and data of acquisition times, are transmitted to the server 20 in the road surface condition management center 2 (step S11).
Next, at the server 20, the estimated road surface condition data and vehicular information are stored for each of the predetermined locations Lr in order of acquisition times (step S12).
Then, at the statistical data generating means 31, the actual statistical data Mk, which are the statistical data of the counts of vehicles having estimated a DRY road surface, vehicles having estimated a WET road surface, vehicles having estimated a SNOW road surface, and vehicles having estimated an ICE road surface at the predetermined location Lr, are generated for each of times tk (k=−n to −1), and they are arranged in a time series (step S13).
Next, the time-dependent changes in the incidence rates SRk of estimated road surface conditions being R at times t−n t−1 are respectively approximated by the n-th functions Gr(t) (step S14), and the function values GR(tp), which are the values of n-th functions GR(t)) at a future time tp, are derived, respectively. And the predicted incidence rates SRp of road surface conditions R at time tp, are derived using these four function values GR(tp), respectively (step S15).
And the estimated road surface condition R indicating the highest value of predicted incidence rates SRp is predicted to be the road surface condition R(tp) at time tp (step P16).
Finally, the information on the determined road surface condition is transmitted to the registered vehicles.
It is to be noted that when a road surface condition at a location Lr′, which is different from location Lr, is to be predicted, the procedure returns to step S13 and the prediction of the road surface condition is continued.
As described above, in the first embodiment, time-series waveforms of vibration of a running tire are detected by the acceleration sensor 11 provided on each of a plurality of vehicles Wi traveling through the location within a predetermined range. And using the information, the road surface conditions R at a plurality of times tk (k=−n to −1) are estimated, respectively. At the same time, the time-dependent changes in these estimated road surface conditions are respectively approximated by the n-th functions Gr(t), and from the function values GR(tp) of the n-th functions GR(t), the predicted incidence rates SRp of road surface conditions R at a future time tp at the location within the predetermined range are derived. And the future road surface condition R(tp) is predicted from the magnitudes of the predicted incidence rates SRp. Accordingly, the future road surface condition can be predicted with excellent accuracy.
It is to be noted that, in the first embodiment, the predicted incidence rates SRp of road surface conditions R at time tp are found by approximating the time-dependent changes in road surface conditions by the n-th functions Gr(t). However, the arrangement may be such that a time-series filter, such as a Kalman filter or a particle filter, which is used in predicting the ever-changing vehicle position from the output of an acceleration sensor or GPS data, for instance, may be used to obtain the predicted incidence rates SRp of road surface conditions R.
Also, in the first embodiment, the predicted incidence rates SRp of road surface conditions R(tp) at time tp are obtained from the time-dependent changes in the incidence rates SRk of estimated road surface conditions R at times t−10 to t−1, which are estimated by the road surface condition estimating means 13, provided on each of a plurality of vehicles Wi. However, a decision value estimating means for estimating road surface estimation decision values Kk used in estimating road surface conditions R, in the place of the road surface condition estimating means 13, may be provided in each of the vehicles Wi, and the predicted values Kp of the road surface estimation decision values at time tp may be found from the road surface estimation decision values Kk at times t−10 to t−1. And the road surface condition R(tp) at time tp may be predicted from the predicted values Kp.
Also, in the first embodiment, the statistical data of the estimated road surface conditions is classified into four conditions of DRY road surface, WET road surface, SNOW road surface, and ICE road surface. However, the SNOW road surface and the ICE road surface, which are both slippery, may be combined as a dangerous road surface, and the incidence probability of the dangerous road surface may be determined.
In the foregoing first embodiment of the invention, no consideration is given to changes in the weather at the location where the vehicles are traveling. However, as shown in
The weather correction means 35 includes a weather forecast data obtaining unit 35a, a weather model storage unit 35b, and a weather correction unit 35c.
The weather forecast data obtaining unit 35a obtains the forecast data of weather and temperature in the time slot including the time tp by connecting to the not-shown Internet.
The weather model storage unit 35b stores four weather models MR (R: DRY, WET, SNOW, ICE).
It is to be noted that the sum of the probabilities pTmR: Σp=pTmD+pTmW+pTmS+pTmI is not 1. For actual use, therefore, normalized probabilities are used. Hereinbelow, pTmR=pTmD/Σp is referred to as the incidence frequency.
For example, as shown in
The weather models MR are generated using past weather conditions and road surface condition data at the predetermined location Lr. It is to be noted that the weather models MR may be generated for different seasons or different months. In the present example, four different weathers and the temperature of −10° C. to 10° C. are used. However, the weathers may be more finely classified, and different ranges and graduation intervals of temperature may be used.
The weather correction unit 35c takes out the data of incidence frequencies pTmD, pTmW, pTmS, pTmI when the weather is m and the temperature is T from the weather models stored in the weather model storage unit 35b based on the forecast data of the weather m and temperature T in the time slot including the time tp obtained by the weather forecast data obtaining unit 35a and corrects the predicted incidence rates SRp of road surface conditions R calculated by the predicted statistical data generating means 33, using these incidence frequencies pTmD, pTmW, pTmS, pTmI.
More specifically, as shown in
After the correction, the predicted incidence rate of a DRY road surface is calculated by SDpz=PTmD·SDp, and the predicted incidence rate of a WET road surface is calculated by SWpz=PTmW·SWp. Also, the predicted incidence rate of a SNOW road surface is calculated by SSpz=PTmS·SSp, and the predicted incidence rate of an ICE road surface is calculated by SIpz=PTmI·SIp.
It is to be noted that
The road surface condition predicting means 34 predicts the road surface condition indicating the highest predicted value after correction (predicted incidence rate SRpz) to be the road surface condition Rz(tp) at the location Lk within the predetermined range at time tp.
By repeating the prediction of road surface conditions using the prediction results of road surface conditions like this, it is possible to predict the road surface condition Rz(tq) at a further future time tq (tq>tp) with weather taken into consideration.
In the foregoing second embodiment, the road surface condition Rz(tq) with weather taken into consideration is predicted by correcting the predicted incidence rates SRp of the road surface conditions R using the incidence frequencies PTmD, PTmW, PTmS, PTmI when the weather is m and the temperature T. However, as shown in
Note that the reference maps MR0 for the respective road surface conditions and the degrees of matching J will be described in detail below.
As shown in
Also, the reference map MW0 for WET road surface, created with the test vehicle traveling on WET road surfaces, shows a mapping of the rate PWD0 of estimating the road surface to be a DRY road surface, the rate PWW0 of estimating the road surface to be a WET road surface, the rate PWS0 of estimating the road surface to be a SNOW road surface, the rate PWI0 of estimating the road surface to be an ICE road surface. And reference map MS0 for SNOW road surface, created with the test vehicle traveling on SNOW road surfaces, shows a mapping of the rate PSD0 of estimating the road surface to be a DRY road surface, the rate PSW0 of estimating the road surface to be a WET road surface, the rate PSS0 of estimating the road surface to be a SNOW road surface, the rate PS10 of estimating the road surface to be an ICE road surface.
Also, the reference map MI0 for ICE road surface, created with the test vehicle traveling on ICE road surfaces, shows a mapping of the rate PID0 of estimating the road surface to be a DRY road surface, the rate PIW0 of estimating the road surface to be a WET road surface, the rate PIS0 of estimating the road surface to be a SNOW road surface, the rate PII0 of estimating the road surface to be an ICE road surface.
As a matter of course, PDD0 is the highest in the reference map MD0; PWW0 is the highest in the reference map MW0; PSS0 is the highest in the reference map MS0; and PII0 is the highest in the reference map MI0.
Hereinafter the rates PRR′ will be referred to as reference incidence rates. And R and R′ herein refer to any one of D, W, S, and I.
The degree of matching calculating means 37, as shown in
The DRY degree of matching JD is calculated by the following formula (1) using the predicted incidence rates SRp in the predicted statistical data MpC and the above-described reference incidence rates PDD0, PDW0, PDS0, PD10, in the reference maps MD0:
J
D
=esp{−(|SDp−PDD0|2+|SWp−PDW0|2+|SSp−PDS0|2+|SIp−PDI0|2)} (1)
In a similar manner, the WET degree of matching JW, the SNOW degree of matching JS, and the ICE degree of matching JI are respectively calculated by the following formulas (2) to (4):
J
W
=esp{−(|SDp−PWD0|2+|SWp−PWW0|2+|SSp−PWS0|2+|SIp−PWI0|2)} (2)
J
S
=esp{−(|SDp−PSD0|2+|SWp−PSW0|2+|SSp−PSS0|2+|SIp−PSI0|2)} (3)
J
I
=esp{−(|SDp−PID0|2+|SWp−PIW0|2+|SSp−PIS0|2+|SIp−PII0|2)} (4)
The weather probability adding unit 35d calculates the incidence rates SRpZ of road surface conditions when the weather is m and the temperature T, using the product of the incidence frequency PTmR when the weather is m and the temperature T and the degree of matching JR calculated by the degree of matching calculating means 37.
When the weather is m and the temperature T, the predicted incidence rate of DRY road surface is calculated by SDpZ=PTmD·JD. The predicted incidence rate of WET road surface is calculated by SWpZ=PTmW·JW. Also, the predicted incidence rate of SNOW road surface is calculated by SSpZ=PTmS·JS. And the predicted incidence rate of ICE road surface is calculated by SIpz=PTmI·JI.
It is to be noted that
The road surface condition predicting means 34 predicts the road surface condition indicating the highest predicted value (predicted incidence rate SRpZ) when the weather is m and the temperature T to be the road surface condition RZ(tp) at the location Lk within the predetermined range at time tp.
By repeating the prediction of road surface conditions using the prediction results of road surface conditions like this, it is possible to predict the road surface condition RZ(tq) at a further future time tq (tq>tp) with weather taken into consideration.
In the foregoing third embodiment, the predicted incidence rates SDpZ of road surface conditions R when the weather is m and the temperature T are predicted by adding the incidence frequencies PTmR of weather models to the degrees of matching JR obtained from the predicted incidence rates SRp in the predicted statistical data MpC and the reference maps MR0 for the respective road surface conditions. However, road surface weather models MRmT mapping the incidence rates PRmT of the respective road surface conditions R when the weather is m (m=1 to 4) and the temperature T may be generated, and the predicted incidence rates may be predicted from the predicted incidence rates SRp of road surface conditions R generated by the data generating means 33 and the road surface weather models MRmT.
Also, in the foregoing first to third embodiments, the road surface condition R(tp) at time tp is predicted using actual statistical data Mk of incidence rates SRk of road surface conditions estimated by a plurality of vehicles Wi calculated for the respective road surface conditions R. However, when there is only one vehicle Wx that has passed the predetermined location Lr at time tk, actual statistical data Mk at time tk cannot be generated.
In such a case, the reference maps MR0 for the respective road surface conditions as described in the foregoing third embodiment are used as actual statistical data Mk.
More specifically, when the estimated road surface condition estimated by the vehicle Wx is a DRY road surface, the reference map MD0 for DRY road surface is used as actual statistical data Mk at time tk.
The reference maps MD0, as already mentioned, are mappings of the rate PDD0 of the vehicle traveling on the DRY road surface estimating the road surface to be a DRY road surface, the rate PDW0 of the vehicle estimating the road surface to be a WET road surface, the rate PDS0 of the vehicle estimating the road surface to be a SNOW road surface, and the rate PDI0 of the vehicle estimating the road surface to be an ICE road surface. Therefore, use of the reference maps KD0 in substitution for the actual statistical data Mk may get the actual statistical data Mk closer to the actuality than when the incidence rate of DRY road surface is assumed to be 1.0. Hence, a sufficient accuracy can be ensured for the prediction of a road surface condition.
This can also be applied to cases where the number of vehicles having passed the predetermined location Lr is limited (e.g., less than 10 vehicles). For example, when there are only three vehicles having passed the location with the vehicles Wx and Wy estimating the road surface condition to be a DRY road surface and the vehicle Wz estimating it to be a WET road surface, the actual statistical data Mk may be generated from the reference map MD0 and the reference map MW0. In doing so, it is preferable that weighting is done according to the number of vehicles; for example, the reference map MD0 is weighted by 2 (weighting coefficient wD=2/3), and the reference map MW0 by 1 (weighting coefficient wW=1/3).
In this manner, by generating reference maps MR0 for the respective road surface conditions in advance, future road surface conditions can be predicted with accuracy even when the number of vehicles having passed the predetermined location Lr is limited.
Also, in the foregoing first to third embodiments, a plurality of vehicles Wi are used for the prediction of road surface conditions. However, by use of the reference maps MR0 for the respective road surface conditions, it is also possible to predict road surface conditions Rr(tp) at a plurality of predetermined locations Lr by operating a single vehicle Wp traveling the route including the plurality of predetermined locations a plurality of times.
That is, estimated road surface conditions at the predetermined location Lr are sent from the vehicle Wp to the server 20 at times ta, tb, tc, . . . , and estimated road surface conditions at the predetermined location Lr, at times ta+h, tb+h, tc+h, . . . . For example, the estimated road surface conditions R at a predetermined location Lr change as time passes like at times ta, tb, tc, . . . (a<b<c<0). Accordingly, the estimated road surface conditions R estimated by the vehicle Wp at time tm (m>0), substituted with the reference maps MR0 of road surfaces R, may be used as actual statistical data Mk. In this way, the road surface conditions at time tm can be predicted as in the foregoing first to third embodiments. It is to be noted that the reference maps MR0 may be changed for each of vehicles Wp.
In the foregoing first embodiment, the predicted incidence rates SRp of road surface conditions R(tp) at a future time tp (p>0) are predicted from the time-dependent changes in road surface conditions R at a plurality of times tk (k=−n to −1) estimated at the location within a predetermined range. However, as shown in
It is to be noted that, in the present example, the statistical data generating means 31 generates maps tallying up the incidence rates of estimated road surface conditions R estimated at the location Lr within the predetermined range at time t, which is time t (r)1 immediately before the time tp at which a prediction is made, for the respective estimated road surface conditions R, from the data stored in the data storage means 22 of the server 20.
The road surface condition transition predicting means 38 includes a transition model storage unit 38a and an incidence rate predicted value calculating unit 38b.
The transition model storage unit 38a, as shown in
The transition models Tr, which are generated for the respective road surface conditions R, represent the probabilities of the road surface conditions Rk at time tk becoming the road surface conditions R′ at time (hereinafter referred to as transition probabilities).
For example, with the DRY transition model TD in the upper middle of
The incidence rate predicted value calculating unit 38b calculates incidence rate predicted values SRp by correcting the incidence rate predicted value SR-1 for the estimated road surface conditions R at time t−1 using the above-mentioned transition probabilities qR,R′. The incidence rate predicted values SRpt are calculated for the respective road surface conditions R.
The incidence rate predicted values SDpt, SWpt, SSpt, SIpt for the respective road surface conditions R are respectively calculated by the following formulas (5) to (8):
S
Dpt
=S
D-1
·q
D,D
+S
W-1
·q
W,D
+S
S-1
·q
S,D
+S
I-1
·q
I,D (5)
S
Wpt
=S
D-1
·q
D,W
+S
W-1
·q
W,W
+S
S-1
·q
S,W
+S
I-1
·q
I,W (6)
S
Spt
=S
D-1
·q
D,S
+S
W-1
·q
W,S
+S
S-1
·q
S,S
+S
I-1
·q
I,S (7)
S
Ipt
=S
D-1
·q
D,I
+S
W-1
·q
W,I
+S
S-1
·q
S,I
+S
I-1
·q
I,I (8)
The road surface condition predicting means 34 predicts the road surface condition indicating the highest value of these incidence rate predicted values SDpt, SWpt, SSpt, SIpt to be the road surface condition R(tp) at the location Lk within the predetermined range at time tp.
In this manner, the predicted incidence rates SRpt of the road surface conditions R at a future time tp (p>0) are predicted from the transition probabilities qR,R′ of road surface conditions determined in advance and the incidence rates SR-1 of the estimated road surface conditions R at time t−1 as the past data. Thus, the road surface condition Rt(tp) at a future time tp can be predicted by a simple scheme.
In the foregoing fourth embodiment, the incidence rate predicted values SDpt, SWpt, SSpt, SIpt are obtained using the transition probabilities qR,R′ determined in advance from the past data. However, as the transition probabilities, the probability of the same road surface condition R continuing may beset as q′R,R=0.8, and the probability of a road surface condition R changing into a road surface condition R′ may be set as q′R,R′=(0.2)/3, for instance. It is to be noted that the denominator 3 is the number of road surface conditions R′ other than the road surface condition R.
In this case, the incidence rate predicted values S′Dpt, S′Wpt, S′Spt, S′Ipt for the respective road surface conditions R are respectively calculated by the following formulas (5′) to (8′):
S′
Dpt
=S
D-1·0.8+SW-1·0.2/3+SS-1·0.2/3+SI-1·0.2/3 (5′)
S′
Wpt
=S
D-1·0.2/3+SW-1·0.8+SS-1·0.2/3+SI-1·0.2/3 (6′)
S′
Spt
=S
D-1·0.2/3+SW-1·0.2/3+SS-1·0.8+SI-1·0.2/3 (7′)
S′
Ipt
=S
D-1·0.2/3+SW-1·0.2/3+SS-1·0.2/3+SI-1·0.8 (8′)
In this manner, the method for predicting a road surface condition at time tp using the pre-set transition probabilities q′R,R′ of estimated road surface conditions or the predetermined transition probabilities qR,R′ of estimated road surface conditions may be applied to the prediction in the initial phase when the number of past data is limited. This enables a prediction under the circumstances where there is no adequate amount of past data.
In the foregoing fourth embodiment, the road surface condition R(tp) at a future time tp is predicted using the incidence rate predicted values SRpt derived from the past data (t=t−1) and the predetermined transition probability of road surface conditions. However, the predicted incidence rates VRp of road surface conditions R at a future time tp may be obtained using the preceding predicted incidence rates VR-1, which are the predicted incidence rates at time t−1, the transition probability qR,R′ described in the fourth embodiment, and the predicted incidence rates SRp determined in the first embodiment. And using these incidence rate predicted values VDp, VWp, VSp, VIp, the road surface condition R(tp) at time tp may be predicted. In this manner, it is possible to predict a future road surface condition with even greater accuracy.
Now a description is given of the procedure for obtaining the predicted incidence rates VRp with reference to
Calculation of the predicted incidence rates VRp is done when preceding predicted incidence rates up to time t−1, namely, VR-n, . . . VR-2, VR-1 (calculated values) and actual statistical data M−n, . . . , M−2, M−1 are already known. Note that the initial preceding predicted incidence rate VR-n is the setting value.
Firstly, the products PR,R′p of the preceding predicted incidence rates VR-1 at time t−1 and the transition probabilities qR,R′ are obtained (procedure 1).
Next, the predicted incidence rates SRp at time tp are obtained from the actual statistical data M−n, . . . , M−3, M−2, M−1 at times preceding time tp, namely, t−n, . . . , t−3, t−2, t−1 (procedure 2). Note that the description of the method for predicting the predicted incidence rates SRk is omitted since it is the same as in the first embodiment.
Then the products VRp of the products PR,R′p obtained in the procedure 1 and the predicted incidence rates SRp obtained in the procedure 2 are calculated, and the products VRp are used as the predicted incidence rates of road surface conditions R at a future time tp (procedure 3).
Finally, these predicted incidence rates VDp, VWp, VSp, VIp are compared with each other, and the road surface condition showing the highest value is predicted to be the road surface condition R(tp) at the location Lk within the predetermined range at time tp (procedure 4). It is to be noted that procedure 1 and procedure 2 may be reversed.
It is to be appreciated that when the actual statistical data Mp at time tp are actually measured, it goes without saying that the time tp becomes “preceding time t−1”, the time becomes time t−2, and the initial time t−n becomes time t−n-1 (the past time increasing by 1).
In
By repeating this operation, it is possible to predict the road surface condition R(tk) at a further future time tq (tq>tp).
In the foregoing fifth embodiment, the products of the products PR,R′p and the predicted incidence rates SRp are used as the predicted incidence rates VRp of the road surface conditions R at a future time tp. However, the degrees of matching JRp between the reference incidence rates PR,R′ in the reference maps MR0, which are the models corresponding to the respective road surface conditions, and the predicted incidence rates SRp may be used in the place of the predicted incidence rates SRp.
The degrees of matching JRp are calculated by the following formula (9) in the same manner as in the third embodiment:
J
Rp=exp{−(|SDp−PRD0|2+|SWp−PRW0|2+|SSp−PRS0|2+|SIp−PRI0|2)} (9)
Also, in the foregoing fifth embodiment, the predicted incidence rates used are the predicted incidence rates SRp obtained in the first embodiment. However, the predicted incidence rates VRp of the road surface conditions R may be obtained using the predicted incidence rates SRpZ corrected by the weather model obtained in the second embodiment or the predicted incidence rates SRpz corrected using the maps MR0 of the respective road surface conditions and the weather model. In this manner, the prediction with weather changes taken into consideration can be made such that it is possible to predict future road surface conditions with even greater accuracy.
It is to be noted that in another method of prediction with weather changes taken into consideration, transition models TRT for the weather or the weather and temperature, which are the weather forecast data, may be generated, and the predicted incidence rates VRp of road surface conditions at a future time tp may be obtained using the transition models TRT, preceding predicted incidence rates VR-1, and predicted incidence rates SRp.
In the foregoing fourth embodiment, the predicted incidence rates SRpt of road surface conditions R(tp) at a future time tp are predicted from the transition models TR determined in advance and the incidence rates SR-1 at time t−1. However, by changing the transition models TR according to the transitional state of actual statistical data Mk, the accuracy of predicting the road surface condition R(tp) can be further improved.
Hereinbelow, a description is given of a method for changing the transition models TR.
Firstly, it is checked to find how the incidence rates SR-1 at time t−1, which are actual measurement data, have changed from the incidence rates SR-2 at time t−2, which is the time immediately before time t−1.
For example, as shown in
Here, if the change rate of incidence rate of the road surface condition R in the actual measurement data is vR-k=SR-k/SR-k-1, then vD-1=0.86, vW-1=2.0, and VS-1=VD-k=1.00.
Thus, the probability of a DRY road surface condition continuing in the transition model TD is changed from qD,D=0.7 to uD,D=VD,D·qD,D=0.6, and the probability of a DRY road surface changing into a WET road surface is changed from qD,W=0.2 to uD,W=vW-1·qD,W=0.4. Then uD,R is normalized such that the sum of the transition probabilities after the change is 1.
The transition models TR are changed by repeating the above operation for the other transition models TW, TS, and TI.
Hereinafter, the transition models after the change will be referred to as UR, and the probability of the road surface condition transiting from R to R′ as uR,R′.
Next, the incidence rate predicted values SRpT are calculated by correcting the incidence rates SR-1 of the estimated road surface conditions Rat time by use of the above-mentioned uR,R′.
The calculation formula for the incidence rate predicted value SRpT is as shown in the equation (10) below:
S
RpT
=S
D-1
·u
D,R
+S
W-1
·u
W,R
+S
S-1
·u
S,R
+S
I-1
·u
I,R (10)
The incidence rate predicted value SRpT is calculated for each of the road surface conditions R.
The road surface condition predicting means 34 predicts the road surface condition indicating the highest value of these incidence rate predicted values SDpT, SWpT, SSpT, SIpT to be the road surface condition RT(tp) at the location Lk within the predetermined range at time tp.
In this manner, the transition probabilities uR,R′ of road surface conditions are changed according to the transitional states of past data (actual measurement data Mk). Accordingly, the road surface condition RT(tp) at a future time tp can be predicted with even greater accuracy.
In the foregoing sixth embodiment, the transition models TR are changed using the change rates vR from the incidence rates SR-2 at time t−2 to the incidence rates SR-1 at time t−1. However, the transition models TR may be changed using the change rates VR-m, . . . , VR-2, vR-1 at a series of times t−m, . . . , t−3, t−2 preceding time t−1. As mentioned above, the change rates of incidence rates of road surface conditions R are calculated by vR-k=SR-k/SR-k-1.
More specifically, the change rates vR-m, . . . , vR-2, vR-1 of incidence rates at times t−m to t−1 are approximated by the n-th functions gR(t), and each of the function values gR(tp) of the approximation functions (n-th functions) at t=tp is derived. Then the change rates vRp of incidence rates at time tp are obtained from these four function values gR(tp). Following this, the transition models TR are changed by calculating the probabilities uR,R′ of the road surface conditions transiting from R to R′ by use of the change rates vRp of incidence rates.
Or the change rates vRp of incidence rates at time tp may be obtained by use of the computing equation of the change rates vR-k of incidence rates, for instance, using the mean value or linear combination of the change rates vR-m, . . . , vR-2, vR-1 of incidence rates.
Also, in the foregoing sixth embodiment, the incidence rate predicted values SRpT are calculated using the incidence rates SR-1 at time t−1 and the transition probabilities uR,R′. However, the road surface condition may be predicted using the preceding predicted rates VR-1, which are the predicted incidence rates at time t−1, and the transition probabilities uR,R′.
Also, the transition probabilities uR,R′ may be changed according to information on the geography and on the weather and temperature supplied as weather forecast.
Thus far, the invention has been described with reference to specific embodiments thereof. However, the technical scope of the invention is not limited to the described scope of the embodiments. And it should be evident to those skilled in the art that various modifications, changes, and improvements may be made thereto without departing from the spirit and scope of the invention.
For example, the foregoing first to sixth embodiments are not limited to the classification of the statistical data of estimated road surface conditions into the four conditions of DRY road surface, WET road surface, SNOW road surface, and ICE road surface. The classification may be done by the road surface friction coefficient μ or into “high-μ road (μ≧0.7)”, “intermediate-μ road (0.3<μ<0.7)”, “low-μ road (μ≦0.3)”, for instance.
Also, in the foregoing first to sixth embodiments, predicted values SRp of incidence rates of the road surface conditions R at time tp are obtained from the time-dependent changes in the incidence rates SRk of the estimated road surface conditions being R from time t−10 to t−1 estimated by the road surface condition estimating means 13 provided in each of a plurality of vehicles Wi. However, in the place of the road surface condition estimating means 13, a decision value estimating means for estimating road surface estimation decision values K used in estimating road surface conditions R may be provided in each of the vehicles Wi. And the predicted values Kp of the road surface estimation decision values at time tp may be obtained from the road surface estimation decision values Kk from time t−10 to time t−1. And a road surface condition Rp at time tp may be predicted from the predicted values Kp.
Also, in the foregoing first to sixth embodiments, a road surface condition estimating means 13 is provided in each of vehicles Wi. However, the arrangement may be such that a road surface condition estimating means 13 is provided at the road surface condition management center 2 and a plurality of band values used in estimating road surface conditions (vibration levels of specific frequency bands detected from acceleration waveform) or the calculated values of the band values are sent to the road surface condition management center 2.
By this arrangement, it is no longer necessary to install the road surface condition estimating means 13 within the tire 40, and thus the system within the tire can be made lighter in weight.
Also, in the foregoing first to sixth embodiments, a road surface condition determining unit is used which is configured to estimate the road surface condition to be one of the DRY road surface, WET road surface, SNOW road surface, and ICE road surface from the values of identification (discriminant) functions using kernel functions. However, other road surface condition determining means may be used. For example, such a road surface condition estimating unit may be configured to estimate the road surface friction coefficient μ by comparing the vibration level of vibration spectrum obtained by applying a frequency analysis to the time-series waveform of acceleration detected by the acceleration sensor 11 against a G-table showing the predetermined relationship between the road surface friction coefficient μ and the vibration level. Or such other road surface condition estimating unit may be configured to estimate the road surface condition from the time-series waveform supplied by an acceleration sensor attached to the tire or rim.
Or such a road surface estimating means may detect the tire noise arising from the running tire. And by comparing the mean value of sound pressure levels within a set frequency range of the detected tire noise against the reference sound pressure levels, it may estimate whether the road surface is an asphalt road amply wet with water, a slightly wet asphalt road, a dry asphalt road, or an ice-covered road.
A method for predicting a road surface condition according to an embodiment of the present invention includes the step of obtaining vehicular information, which is information on the behavior of a traveling vehicle by an on-board sensor mounted on the vehicle and the step of predicting a road surface condition of a location within a predetermined range, using road surface estimation decision values to be used in estimating road surface conditions, which are calculated using the vehicular information, or the estimated road surface conditions estimated using the road surface estimation decision values. In the step of predicting the road surface condition, predicted incidence rates SRp for respective road surface conditions, which are the incidence rates of road surface conditions at the location within the predetermined range, are calculated from the time-dependent changes in the road surface estimated values calculated using the vehicular information obtained by the vehicles having passed the location within the predetermined range or the time-dependent changes in the estimated road surface conditions, and then the road surface condition at the location within the predetermined range is predicted from the calculated predicted incidence rates SRp.
The suffix k of time tk refers to times (past) before the present time t0 when k<0 and times (future) after the present time t0 when k>0. It is to be noted that at the present time t=t0, no vehicular information is assumed to be available yet. Also, it is to be appreciated that the road surface condition at the location within the predetermined range to be predicted refers to the future road surface condition R(tp), which is the road surface condition at time tp (p>0), or the present time t.
Also, the road surface estimation decision values (or estimated road surface conditions) at time tk refer to the road surface estimation decision values (or estimated road surface conditions) obtained within the time width Δtk including time tk (for example, tk−Δtk/2≦t≦tk+Δtk/2).
In this manner, a future road surface condition at a location within a predetermined range is predicted using the road surface estimation decision values at a plurality of times obtained at the location within the predetermined range or the time-dependent changes in the estimated road surface conditions. Therefore, a future road surface condition can be predicted with excellent accuracy.
Also, in a method for predicting a road surface condition according to another embodiment of the present invention, the incidence rate predicted values VRp, which are the corrected values of predicted incidence rates SRp, are calculated by correcting the predicted incidence rates SRp of road surface conditions at the location within the predetermined range by use of the pre-set transition probabilities q′RR′ of estimated road surface conditions or the predetermined transition probabilities qRR′ of estimated road surface conditions and the preceding incidence rates VR-1, which are the predicted incidence rates at time t−1 before the time tp at which the road surface condition is predicted.
In this manner, the predicted incidence rates SRp at the location within the predetermined range are corrected using the pre-set transition probabilities q′RR′ of estimated road surface conditions or the predetermined transition probabilities qRR′ of estimated road surface conditions and the already calculated preceding incidence rates VR-1. As a result, it is possible to predict a future road surface condition with excellent accuracy.
A method for predicting a road surface condition according to still another embodiment of the present invention includes the step of obtaining vehicular information, which is information on the behavior of a traveling vehicle by an on-board sensor mounted on the vehicle and the step of predicting a road surface condition at a location within a predetermined range, using road surface estimation decision values to be used in the estimation of road surface conditions, which are calculated using the vehicular information, or the estimated road surface conditions estimated using the road surface estimation decision values. In the step of predicting a road surface condition, a road surface condition at the location within the predetermined range at a time after a predetermined time is predicted from the estimated road surface conditions estimated using vehicular information obtained by vehicles having passed the location within the predetermined range and the pre-set transition probabilities of estimated road surface conditions or the predetermined transition probabilities for the estimated road surface conditions.
In this manner, the road surface condition at time tp is predicted using the pre-set transition probabilities q′RR′ of estimated road surface conditions or the predetermined transition probabilities qRR of estimated road surface conditions. Thus, it is possible to predict the road surface condition at time tp easily.
Also, in a method for predicting a road surface condition according to still another embodiment of the present invention, the pre-set transition probabilities q′RR′ of estimated road surface conditions or the predetermined transition probabilities qRR′ of estimated road surface conditions are corrected using the incidence rates of road surface conditions estimated based on the vehicular information obtained at a time before the time at which the road surface condition is predicted.
In this manner, by correcting the transition probabilities q′RR′ or the transition probabilities qRR′ by the actually determined incidence rates of road surface conditions, the accuracy in predicting a future road surface condition can be further improved.
Also, in another method, the predicted road surface condition R(tp) is corrected based on weather forecast information, such as weather, temperature, rainfall, wind speed, and sunshine hours.
Here, “correction” means determining whether or not the predicted road surface condition changes with the information on weather or temperature provided as the weather forecast, and when there is any change, predicting which of the road surface conditions will show up.
In this manner, the predicted road surface condition is corrected based on the weather forecast information. Therefore, the accuracy in predicting a road surface condition can be improved further.
Also, the information to be used in correction may include the estimation results of road surface conditions at other locations within a predetermined range. More specifically, among the past conditions (weather, traffic volume, estimation results of road surface conditions), the estimation results of road surface conditions at the location B within a predetermined range having a correlation to the location A within a predetermined range may be used to correct the estimation result at the location A within a predetermined range.
Also, in a method for predicting a road surface condition according to yet another embodiment of the present invention, a road surface condition R (tp) at a time tq (tq>tp) even after the time tp at which a prediction is made is predicted using the predicted road surface condition R(tp).
In this manner, by repeating the prediction of a road surface condition using the predicted results, a future road surface condition can be further predicted.
It is to be understood that the foregoing summary of the invention does not necessarily recite all the features essential to the invention, and subcombinations of all these features are intended to be included in the invention.
As described herein, the present invention provides methods for determining a road surface condition within a predetermined space of time at a location within a predetermined range with excellent accuracy. Therefore, the travel safety of vehicles may be improved if the determination results are communicated to the vehicles traveling along the location within the predetermined range.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2014-210322 | Oct 2014 | JP | national |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/JP2015/079029 | 10/14/2015 | WO | 00 |
