FR3130431A1 - Method for predicting a surface condition of a road segment - Google Patents

Abstract

The invention relates to a method for predicting at least one weather condition on the surface of a road segment comprising steps of obtaining (200) weather data for a geographical area comprising at least one particular road segment, of predicting ( 201) of at least one surface weather condition of the road segment by applying a predictive model to the weather data obtained, of determining (203) a frequency of updating the prediction for the segment, the frequency being proportional a level of risk associated (202) with the predicted surface condition for the segment, and updating the prediction for the segment from new weather forecasts according to the determined frequency. Figure 2.

Description

Method for predicting a surface condition of a road segment

The invention belongs to the automotive field and particularly relates to techniques for predicting a weather condition on the surface of a road segment.

Prior art

Advanced driver assistance systems and autonomous driving are currently experiencing strong development. To travel with maximum safety, it is particularly important that vehicles equipped with such systems know the surface conditions of the road on which they are traveling in order to adapt their speed, their safety distances, and to estimate the distances braking.

To this end, models have been developed to predict the state of the roadway from forecasts and/or atmospheric weather observations. Such models are for example trained from weather observations and ground observations, carried out for example from road weather stations (RWIS, for Road Weather Information System in English) distributed over the road network, so as to establish correlations between atmospheric conditions and the state of the roadway. It is thus possible, for example, to determine the temperature or the drying time of a road surface from an atmospheric temperature, a wind speed, cloud cover and a type of traffic lane.

Weather observations and forecasts are updated regularly by weather data providers, for example every 15 minutes. These forecasts have a granularity of about 0.01° x 0.01°. and are generally available continentally. Thus, by applying predictive models as soon as new forecasts or weather observations are available, we provide vehicles with updated surface condition predictions.

However, frequent updating of predictions for a large geographical area requires a computational infrastructure that is expensive to set up and maintain.

There is thus a need for a technique for predicting surface conditions at any point of a road network that makes it possible to limit the costs of installing and operating the computing infrastructure.

To this end, a method is proposed for predicting at least one weather condition on the surface of a road segment comprising the following steps:

• Obtaining atmospheric weather data for a geographical area including at least one particular road segment,
• Prediction of at least one surface weather condition of the road segment by applying a predictive model to the weather data obtained,
• Determining a prediction update frequency for the segment, the frequency being proportional to a level of risk associated with the predicted surface condition for the segment, and
• Update of the prediction for the segment from new weather data according to the determined frequency.

The method thus makes it possible to optimize the use of calculation resources by adapting the frequency at which a predictive model is inferred according to surface conditions predicted for this segment. In other words, the predictions made for a particular segment are associated with a period of validity that is all the shorter as the predicted surface condition constitutes a high risk for the vehicles in circulation. Thus, the frequency of updating the predictions is higher for a road segment for which a risky surface condition has been predicted, for example an event likely to affect the friction between the tires of a vehicle and the road. On the contrary, the predictions are recalculated less frequently for the segments for which a surface weather condition has been predicted involving a low risk for motorists, for example a dry road surface for the next few hours. In this way, the use of computing resources is reduced without degrading the quality of the predictions transmitted to the vehicles.

In a particular embodiment, the method is such that the prediction update frequency for a first segment is equal to the prediction update frequency for a second segment, when the level of risk associated with the condition of surface of the second segment is greater than the level of risk associated with the surface condition predicted for the first segment and that the distance separating the first segment from the second segment is less than a threshold.

This increases the frequency of updating the predictions for the road segments which are close to zones for which a risky weather event is predicted. The computational resources are thus concentrated on the segments likely to see their surface conditions modified in the near future.

According to a particular embodiment, the method is such that the prediction update frequency for a segment is also proportional to a traffic density value associated with said segment.

Thus, the method makes it possible to adapt the frequency of update of the predictions of surface conditions of a road segment according to the frequentation of the segment, that is to say according to an observed or forecast traffic density for the segment. A modification of the surface conditions having a greater impact on a very frequented segment, the reactivity of the prediction system is increased for this type of segment. Since a risk is assessed on the basis of a hazard and an issue, the fact that a large number of motorists are confronted with dangerous surface conditions increases the level of risk.

According to a particular embodiment, the method is such that the frequency of updating the prediction for a segment is proportional to a number of requests received by an online service from vehicles in circulation in a geographical area comprising the segment.

In this way, the frequency of updating the predictions of the surface conditions for a given region is adapted according to a degree of solicitation of an online service, preferably according to a degree of solicitation of a service providing surface conditions . Such an arrangement makes it possible to improve the reliability of the predictions on the busiest segments without it being necessary to consult a third-party supplier specialized in the supply of road traffic data. This improves the prediction reliability for a given area at a lower cost.

According to a particular embodiment, the prediction update frequency for a segment is such that an interval between two successive predictions for a segment is between 15 minutes and 7 hours.

In this way, when a risky surface weather condition is predicted for a segment, the predictive model is inferred as soon as new weather data is available, for example every 15 minutes. Conversely, when the predictions made for a segment indicate stable and safe conditions for the next few hours, the predictive model is inferred less often.

According to a particular embodiment, a high level of risk is associated with a surface condition when the predicted condition includes at least one of the following conditions:

• Rain,
• Black ice,
• Snow,
• Temperature below 4 degrees.

The presence of rain, ice or snow modifies the grip of a vehicle on the road and are therefore risky situations for motorists. Thus, it is proposed to assign to prioritize the processing of road segments for which such situations are predicted. Other parameters or combination of parameters can be taken into account to determine a processing priority for a given segment, such as the height of a film of water or a wind speed. For example, a segment predicted to have a combination of low temperature, wind, and rain should be watched closely, as such conditions can quickly lead to icy patches.

The invention also relates to a prediction of at least one weather condition on the surface of a road segment comprising a processor and a memory in which are stored instructions configured to implement the following steps, when they are executed by the processor, the following steps:

• Obtaining atmospheric weather data for a geographical area including at least one particular road segment,
• Prediction of at least one surface weather condition of the road segment by applying a predictive model to the weather data obtained,
• Determining a prediction update frequency for the segment, the frequency being proportional to a level of risk associated with the predicted surface condition for the segment, and
• Update of the prediction for the segment from new weather data according to the determined frequency.

According to another aspect, the invention relates to a server comprising a prediction device as described previously.

According to yet another aspect, the invention relates to an information medium comprising computer program instructions configured to implement the steps of a prediction method as described previously, when the instructions are executed by a processor.

The information medium can be a non-transitory information medium such as a hard disk, a flash memory, or an optical disk for example.

The information carrier can be any entity or device capable of storing instructions. For example, the medium may comprise a storage means, such as a ROM, RAM, PROM, EPROM, a CD ROM or even a magnetic recording means, for example a hard disk.

On the other hand, the information medium can be a transmissible medium such as an electrical or optical signal, which can be conveyed via an electrical or optical cable, by radio or by other means.

Alternatively, the information carrier may be an integrated circuit in which the program is incorporated, the circuit being adapted to execute or to be used in the execution of the method in question.

The different aforementioned embodiments or characteristics can be added independently or in combination with each other, to the steps of the prediction method. Servers, devices and information carriers have at least advantages similar to those conferred by the process to which they relate.

Brief description of figures

Other characteristics and advantages of the invention will become apparent on reading the description which follows. This is purely illustrative and must be read in conjunction with the appended drawings on which:

There represents an environment suitable for implementing the invention according to a particular embodiment,

There is a flowchart on which are represented the main steps of a prediction method according to a particular embodiment, and

There represents the architecture of a device adapted to implement the prediction method according to a particular embodiment.

Description of a particular embodiment

There represents an environment suitable for implementing the prediction method according to a particular embodiment.

There shows segments S1, S2 and S3 of a road network on which a vehicle 104 is traveling. The vehicle 104 is equipped with a GNSS (Global Navigation Satellite System) receiver enabling it to obtain its geographic coordinates and with a telematics unit allowing it to establish connections with a communication network 100 via a cellular access network 105, and in particular to exchange messages with a server 101 of the communication network 100.

The server 101 is a processing server configured to calculate surface condition predictions for a set of segments of a road network from weather data transmitted regularly by a weather forecast provider 102. The weather data obtained from the provider 102 may consist of atmospheric weather observations or forecasts. For this, the server 101 comprises a predictive model trained in a conventional manner from a history of atmospheric weather observations associated with observations made on the ground. Such training allows the predictive model to learn to make correlations between particular atmospheric weather conditions and pavement surface conditions resulting from these atmospheric conditions. In order to improve the predictions, the training data may additionally include a type of taxiway, a type of surface or any other characteristic likely to have an influence on the surface conditions. Thus, by applying the predictive model to weather forecasts provided by the supplier 102 for a particular geographical location, the server 101 can determine the probability of occurrence of different conditions on the surface of a road segment of the geographical area in question.

THE also shows a geospatial database 103 in which is stored a digital representation of a road network. For example, the database 103 includes the characteristics of each segment of the road network supervised by the server 101, such as the geographical coordinates of the starting point and of the ending point of the segment, identifiers of segments to which the points of departure and/or arrival of the segment, direction of traffic, type of traffic lane, etc.

In particular, the database 103 includes at least the information allowing the server 101 to identify the segments included in a particular geographical area for which weather data is available from the server 102. In other words, when the server 101 obtains forecasts and/or weather observations for a particular geographical area, the database 103 enables it to find the segments included in this area to associate them with the predictions made and transmit them to the vehicles traveling on these segments, for example to the vehicle 104.

A particular embodiment of a prediction method will now be described with reference to the according to which the method is implemented by the server 101 of the .

During a first step 200, the server 100 obtains weather data from a supplier 102 of forecasts and/or atmospheric weather observations. The forecast provider 102 (WTF for Wether Forcast Provider) provides the server 101 with atmospheric weather forecasts updated every 15 minutes for geographical areas of approximately 100 km ² . Of course, other update frequencies and other geographic resolutions can be considered depending on the weather data provider without it being necessary to modify the invention.

In step 201, the server 101 applies the weather data obtained to the predictive model so as to obtain predictions of surface conditions corresponding to the various weather forecasts. More specifically, upon receipt of weather forecasts from the supplier 102, the server 101 interrogates the database 103 to find out the road segments concerned by these forecasts, that is to say the segments located in each geographical area for which a weather forecast was obtained. For each segment of this set, the server 101 optionally obtains particular characteristics from the database 103, such as a type of road or surface, and constructs a characteristic vector comprising at least values representative of the weather data and of the characteristics. particulars obtained for the segment. By applying the feature vector to the predictive model, server 101 obtains a probability associated with a predictable surface condition. Thus, for each segment for which the server has obtained weather forecasts, the predictive model provides probabilities associated with conditions such as "rain", black ice", "snow", "dry", or even water levels or surface temperatures.

The predicted conditions are for example stored in the database 103 in association with the information relating to the segment concerned. In this way, upon receipt of a surface condition request from a vehicle, the server determines the segment on which the vehicle is located, for example from a geographical location of the vehicle contained in the request, and queries database 103 to obtain the latest predicted surface conditions for that segment and transmits them in a response to the vehicle. The predictions are thus immediately available and can be transmitted to the vehicles concerned without it being necessary to calculate a new prediction for each request transmitted by a vehicle.

During a step 202, the server 101 determines, for each segment considered, a level of risk according to the predicted surface conditions.

The level of risk is proportional to the influence of a predicted surface condition on the coefficient of friction of a vehicle. In other words, the more the conditions predicted for a particular segment are likely to deteriorate the grip conditions of a vehicle traveling on the segment, the higher the level of risk associated with the segment for the segment. A high level of risk corresponds, for example, to surface conditions such as rain, ice or snow.

At step 203, the server 101 determines, for each segment, a prediction update frequency, that is to say a delay after which steps 200 to 204 of the prediction method will be implemented again. . The prediction update frequency for a particular segment is determined such that it is inversely proportional to a level of risk associated with the segment. In other words, the higher the risk level associated with a segment, the higher the prediction update frequency for this segment, and the shorter the interval between two successive updates. Conversely, the update time is longer for segments with a low level of risk. In other words, the duration of validity of a prediction for a particular segment is inversely proportional to the level of risk associated with the prediction.

According to a particular embodiment, the frequency of updating the prediction for a first segment is also determined according to a level of risk associated with a second segment when the distance separating the first segment from the second segment is less than a threshold. More specifically, in the event that a low risk is determined for a first segment, for example because "dry pavement" type conditions are predicted for the segment, but a higher level of risk is associated for a second segment close to the first segment, for example for a second segment located in a weather cell contiguous to the weather cell in which the first segment is located, then the update frequency of the prediction assigned to the first segment is equal to the update frequency update of the prediction determined for the second segment. According to a particular embodiment, the first segment is assigned an intermediate update frequency between the update frequency initially determined for the first segment and the prediction update frequency determined for the second segment.

According to a particular embodiment, the frequency of updating the prediction for a segment is proportional to a traffic density measured or forecast on said segment. For this, the server 101 can interrogate a traffic data supplier in order to obtain data relating to a traffic density for a particular segment and to determine a level of risk for a segment which is proportional to the traffic density. According to a particular embodiment, the traffic density is estimated from time slots and/or days of the week. For example, the update frequency of a prediction for a segment located in an urban area can be increased during office hours, in the early morning and late afternoon during weekdays, and reduced on weekends. -end and/or during off-peak hours of the day or night.

According to a particular embodiment, at least three update frequencies are defined corresponding to three durations of validity of the predictions associated with a particular segment. For example, we define an intensive update frequency according to which the predictions are deemed valid for a maximum of 15 minutes, a normal update frequency according to which the predictions are deemed valid for a maximum of 60 minutes and an update frequency reduced day that predictions are deemed valid for a maximum of 120 minutes.

According to a particular embodiment, the update frequency determined according to a number of requests for access to an online service. For this, the server 101 can count the number of requests received from vehicles in circulation in a particular geographical area and determine an update frequency that is all the higher as the number of requests received increases. For example, a reduced update rate is associated with segments in a particular geographic area when no requests are received from that area for the last 60 minutes, while an intensive update rate is associated with segments of a particular geographical area when at least 100 requests have been received from this area during the last 60 minutes. As a variant, the frequency of updating the predictions for a particular geographical area is proportional to the number of requests for downloading cartographic data for this particular area. Such an arrangement makes it possible to adapt the frequency of updating of the predictions of surface conditions according to routes followed by vehicles.

In a step 204, the server starts a timer when a prediction is calculated for a particular segment, the timer being configured with the time interval corresponding to the update frequency determined for the segment. When the timer expires, steps 200 through 204 are repeated with new weather forecasts obtained from the weather data provider.

There represents the architecture of a device adapted to implement the prediction method according to a particular embodiment.

The device 300 comprises a storage space 302, for example a memory MEM, a processing unit 301 equipped for example with a PROC processor. The processing unit can be controlled by a program 303, for example a computer program PGR, implementing the prediction method described with reference to the , and in particular the steps of obtaining atmospheric weather data for a geographical area comprising at least one particular road segment, of predicting at least one surface weather condition of the road segment by applying a predictive model to the weather data obtained, determining a prediction update frequency for the segment, the frequency being proportional to a level of risk associated with the predicted surface condition for the segment, and updating the prediction for the segment from new weather data according to the determined frequency.

On initialization, the instructions of the computer program 303 are for example loaded into a RAM memory (Random Access Memory in English) before being executed by the processor of the processing unit 301. The processor of the unit processing 301 implements the steps of the method according to the instructions of the computer program 303.

For this, in addition to the memory and the processor, the device 300 comprises communication means 304, for example a COM network interface suitable for exchanging messages with other devices, and in particular for receiving weather data from a forecast provider. The means of communication correspond for example to an Ethernet network card driven by computer program instructions configured to transmit a request to a weather forecast server, for example an http request, accompanied by parameters such as a geographical zone identifier , and to receive meteorological data for said geographical area in response, for example in the form of a message conforming to JSON, XML, or any other suitable format.

The device 300 also includes a prediction module 305, configured to predict a surface weather condition of a road segment from atmospheric weather forecasts obtained by the module 304 for a geographical area in which the segment is located. The module 305 is for example implemented by computer program instructions adapted to implement an automatic learning model, for example a regression or classification model, or a more complex model such as a model based on decision trees (random forest, or gradient boosting), or even a neural network. The learning model is previously trained in a conventional way from a history of atmospheric weather forecasts associated with surface conditions observed on the ground.

The device 300 also comprises a module 306 for determining a level of risk for each road segment for which a prediction is made by the prediction module 305. The module 306 is implemented by program instructions configured to assign a level a risk that is all the higher when a predicted surface condition is likely to deteriorate a coefficient of friction between the tires of a vehicle and the roadway. For this, the module 306 can use tables and rules established from observations and measurements. Such a table is for example recorded in the memory 302 of the device and contains correspondences between particular surface conditions, or particular combinations of surface conditions and atmospheric forecasts, with a level of risk on a determined scale. For example, the table may contain an association between a combination ("wind", "temperature < 4°C", "wet road") with a high level of risk, for example a risk of level 3 on a scale of 0 to 3.

According to a particular embodiment, the level of risk associated with a road segment is determined according to an observed or forecast traffic density on the road segment, such that the level of risk increases with the density of vehicles in circulation. For this, the module 306 comprises program instructions configured to count the number of requests received by the communication interface 304 from vehicles located in a particular geographical area, and to compare the values obtained with predefined thresholds with which are associated validity periods of the predictions. Alternatively, the module 306 determines the traffic density by controlling the communication means 304 to interrogate a server of a traffic data provider.

The device 300 finally comprises a planning module 307 suitable for determining a date or an update interval for a surface condition prediction for a segment or a set of particular road segments, and programming the execution of the update. day on the determined date, that is to say program a new execution of steps 200 to 204 of the prediction method. For this, the module 307 comprises computer program instructions configured to consult a table stored in the memory 302 in which risk levels and prediction update intervals have been associated. The highest risk levels are associated in the table with short update intervals, for example 15 minute intervals. Conversely, a low level of risk is associated with a longer update interval, for example 7 hours.

According to a particular embodiment, the device is integrated into a processing server, for example a server included in a “cloud” type architecture.

Claims (9)

Method for predicting at least one weather condition on the surface of a road segment comprising the following steps:

• Obtaining (200) weather data for a geographical area comprising at least one particular road segment, and
• Prediction (201) of at least one surface weather condition of the road segment by applying a predictive model to the weather data obtained,
• Determining (203) a prediction update frequency for the segment, the frequency being proportional to a level of risk associated with the predicted surface condition for the segment,
• Updating the prediction (200, 201) for the segment from new weather data according to the determined frequency.

The method of claim 1 wherein the prediction update rate for a first segment is equal to the prediction update rate for a second segment, when the level of risk associated with the surface condition of the second segment is greater than the risk level associated with the surface condition predicted for the first segment and the distance separating the first segment from the second segment is less than a threshold. A method according to any preceding claim wherein the prediction update rate for a segment is further proportional to a traffic density value associated with said segment. Method according to any one of the preceding claims, in which the frequency of updating the prediction for a segment is proportional to a number of requests received by an online service from vehicles circulating in a geographical area comprising the segment. Method according to any one of the preceding claims, in which the frequency of updating the prediction for a segment is such that an interval between two successive predictions for a segment is between 15 minutes and 7 hours. A method according to any preceding claim wherein a high level of risk is associated with a surface condition when the predicted condition includes at least one of the following conditions:

• Rain,
• Black ice,
• Snow,
• Temperature below 4 degrees

Device for predicting at least one weather condition on the surface of a road segment comprising a processor (301) and a memory (203) in which are stored instructions (303) configured to implement the following steps, when they are executed by the processor (301), the following steps:

• Obtaining, from communications means (304), weather data for a geographical area comprising at least one particular road segment, and
• Prediction of at least one surface weather condition of the road segment by applying a predictive model (305) to the weather data obtained,
• Determining a prediction update frequency for the segment, the frequency being proportional to a level of risk associated with the predicted surface condition for the segment,
• Update of the prediction for the segment from new weather data according to the determined frequency.

Server comprising a prediction device according to claim 7. Information carrier comprising computer program instructions configured to implement the steps of a prediction method according to any one of claims 1 to 6, when the instructions are executed by a processor.