Systems and methods for real-time crash detection using telematics data

A non-transitory computer-readable medium storing for improving crash prediction through correlating an observed response and a predicted response of a vehicle and including instructions that when executed by one or more processors cause the one or more processors to:

• generate the predicted response for the vehicle as a function of a response model and according to current vehicle inputs that are control inputs associated with steering, braking, and accelerating the vehicle, wherein the response model is a learning model that predicts behaviors of the vehicle;
• compute a residual indicating an extent of correlation between the predicted response and the observed response;
• and
• in response to determining the residual satisfies a crash threshold that indicates an anomaly between the predicted response and the observed response, provide an alert indicating the vehicle has likely crashed.

A method of improving crash prediction through correlating an observed response and a predicted response of a vehicle, comprising:

• generating the predicted response for the vehicle as a function of a response model and according to current vehicle inputs that are control inputs associated with steering, braking, and accelerating the vehicle, wherein the response model is a learning model that predicts behaviors of the vehicle;
• computing a residual indicating an extent of correlation between the predicted response and the observed response;
• and
• in response to determining the residual satisfies a crash threshold that indicates an anomaly between the predicted response and the observed response, providing an alert indicating the vehicle has likely crashed.