Method for predicting a modification of the linkage conditions of a terminal to a cellular network
A method for predicting, for a vehicle that is connected to a current cellular access point and is traveling on a road network, at least one characteristic associated with a modification of the linkage conditions of the vehicle to the cellular network. The method includes: training a first prediction model associated with the current access point from crowdsourced data collected from at least one training vehicle, the data comprising at least a first location, a speed and a direction of the training vehicle as well as a second location of the training vehicle collected when a disconnection from the current access point is detected; and predicting a third location at which a second vehicle will be disconnected from the current access point while traveling on the road network by applying the prediction model to a location, a speed, and a direction of the second vehicle.
1. A method for predicting, for a vehicle that is connected to an access point to a current cellular network and is traveling on a road network, at least one characteristic associated with a modification of the linkage conditions of the vehicle to said cellular network, the method being characterized in that it comprises the following steps:
training a first prediction model associated with the current access point from crowdsourced data collected from at least one training vehicle, comprising at least, for locating the at least one training vehicle connected to a current cellular access point:
a first location, a speed, and a direction of the at least one training vehicle, and
a second location of the at least one training vehicle, collected when a disconnection from the current access point is detected,
predicting a third location at which the vehicle will be disconnected from the current access point while traveling on said road network by applying the prediction model to:
a location, a speed, and a direction of the vehicle;
training a second prediction model from crowdsourced data collected from the at least one training vehicle, the data comprising at least, for locating the at least one training vehicle connected to a current cellular access point:
a location, a speed, and a direction of the at least one training vehicle, and
an identifier of a second cellular access point obtained after detecting disconnection of the at least one training vehicle from the current cellular access point,
predicting an identifier of a next cellular access point by applying the second prediction model to:
a location, a speed, and a direction of the vehicle, wherein the vehicle locations are locations, relative to the current cellular access point, that are each specified in polar coordinates in the form of a respective angle and a respective distance in relation to the current cellular access point.
2. The method as claimed in claim 1 , wherein the first prediction model is a regression multilayer perceptron neural network.
3. The method as claimed in claim 1 , wherein the second prediction model is a classification multilayer perceptron neural network.
4. A device for predicting, for a vehicle that is connected to a current cellular network and is traveling on a road network, at least one characteristic associated with a next modification of the linkage conditions to said cellular network, the device comprising a memory and a processor which is configured by instructions stored in the memory, said instructions being configured to implement:
a module for training a first prediction model associated with the current access point from crowdsourced data collected from at least one training vehicle, comprising at least, for locating the at least one training vehicle connected to a current cellular access point:
a first location, a speed, and a direction of the at least one training vehicle, and
a second location of the at least one training vehicle, collected when a disconnection from the current access point is detected,
a module for predicting a third location at which the vehicle will be disconnected from the current access point while traveling on said road network by applying the prediction model to:
a location, a speed, and a direction of the vehicle;
a module for training a second prediction model from crowdsourced data collected from the at least one training vehicle, the data comprising at least, for locating the at least one training vehicle connected to a current cellular access point:
a location, a speed, and a direction of the at least one training vehicle, and
an identifier of a second cellular access point obtained after detecting disconnection of the at least one training vehicle from the current cellular access point,
a module for predicting an identifier of a next cellular access point by applying the second prediction model to:
a location, a speed, and a direction of the vehicle, wherein the vehicle locations are locations, relative to the current cellular access point, that are each specified in polar coordinates in the form of a respective angle and a respective distance in relation to the current cellular access point.
5. The device as claimed in claim 4 , wherein the first prediction model is a regression multilayer perceptron neural network.
6. The device as claimed in claim 4 , wherein the second prediction model is a classification multilayer perceptron neural network.