Vehicle-based predictive data transmission system and method for traffic state analysis
A vehicle ( 101 ) includes: a plurality of sensors (S); an acquisition section ( 11 ) that acquires detection data from the plurality of sensors (S); a prediction section ( 12 ) that uses a trained model ( 31 ) to predict a future state of the vehicle ( 101 ) on the basis of detection data (D 1 ) and that outputs a prediction result for the future state as prediction data (D 2 ); and a communication section ( 2 ) that transmits, to a server 102 , usage information which is based on either the detection data (D 1 ) or the prediction data (D 2 ), after transmitting the usage information, the communication section ( 2 ) transmitting subsequent usage information on the basis of a determination result indicating whether it is possible to transmit the subsequent usage information.
1 . A system comprising a vehicle and server, the vehicle comprising:
a plurality of types of sensors;
an acquisition section that acquires a plurality of types of detection data from the respective plurality of types of sensors every time a certain time period has elapsed;
a vehicle side prediction section that, every time the plurality of types of detection data are acquired, uses a vehicle side trained model constructed by machine learning to predict a state of the vehicle after a second time period from a time when the detection data is acquired and that outputs a prediction result for the transition of the state of the vehicle as prediction data, wherein the second time period is longer than the certain time period;
a comparison section that, every time the plurality of types of the detection data are acquired, compares a measured value indicating the state of the vehicle which state is based on the detection data with a predicted value of the state of the vehicle, at the time when the detection data is acquired; and
a communication section that transmits the detection data to a server which uses the prediction data to predict a future state of the vehicle,
wherein in a case where an error which is not less than a threshold set in advance has occurred between the predicted value and the measured value after transmission of the detection data, the communication section transmits, to the server, the detection data corresponding to the measured value,
the server comprising a server side prediction section,
wherein the server side prediction section is configured to repeat an operation of predicting a state of the vehicle and outputting a prediction result for the state of the vehicle as prediction data, the operation being repeated in a cycle identical to that of the vehicle side prediction section,
at a timing at which detection data is received from the vehicle, the server side prediction section carries out prediction by inputting, into a server side trained model constructed by machine learning, a part of the detection data received from the vehicle, and
at a timing at which the detection data is not received from the vehicle, the server side prediction section carries out prediction by inputting, into the server side trained model, as the detection data, prediction data which the server side prediction section outputted last time.
2 . The system as set forth in claim 1 , wherein the vehicle further comprises a determination section that determines whether it is possible to transmit the detection data.
3 . The system as set forth in claim 2 , wherein, while the determination section is determining that an error which has occurred between the predicted value and the measured value is less than the threshold, the communication section transmits no detection data to the server even in a case where the vehicle side prediction section predicts the transition of the state of the vehicle.
4 . The system as set forth in claim 1 ,
wherein the server side prediction section carries out prediction of the future state of the vehicle based on the detection data and traffic data of a traffic participant present around the vehicle.
5 . The vehicle as set forth in claim 1 , wherein the reception section receives the result of the prediction of the future state of the vehicle from the server, and wherein the vehicle further comprises a display section that displays the received result of the prediction of the future state of the vehicle.
6 . The system as set forth in claim 1 , wherein
at a timing at which the detection data is transmitted to the server, the vehicle side prediction section carries out prediction by inputting, into the vehicle side trained model, a part of the detection data transmitted to the server, and
at a timing at which the detection data is not transmitted to the server, the vehicle side prediction section carries out prediction by inputting, into the vehicle side trained model, as the detection data, prediction data which the vehicle side prediction section outputted last time.
7 . The system as set forth in claim 1 , wherein the detection data include update information which is updated periodically and map information.