Learning aggressive driving behavior - Android Application

Question

I am beginner therefore please please please help me out.
I am developing an android app which will learn the aggressive behavior of the driver and when learned it will rate a driver accordingly.
I have fixed the acceleration and deceleration values, whenever exceeded, reported as a wrong action. But turns are the tricky part.
Up-til i have implemented, i get the location where driver starts taking turn and where the turn is completed, the degree of turn taken too. Degree of turn will help me categorize type of turn (U-Turn or Simple) . Plus i can get speed of entering and exiting the turn. It will help me evaluate the turn taken.

• But the problem are the long curved roads, how can i handle them?
• Or how can i handle sharp turn during the smooth curve?
• Or what should i do to handle these both situations?

.
There could be pattern matching technique but i think its very much complex and i have no idea about it.
Thanks :)

Answer 1

I guess you could use the phone built-in accelerometer to measure acceleration force during turns. This likely should not result higher than 0.8/1g (a F1 car reaches 5/6 g in turns to have an idea). The problem lies mainly in defining a threshold of acceptable g-force values, which should probably be defined by physically testing the sensor, and in the accuracy of the sensor itself. Indeed I firmly believe that the sensor must be mounted in some fixed position so to obtain consistent values.

Still, even if it seems a hard path to measure the "comfort" of a turn with this kind of device I think it would be the correct way since the g-force is the one giving us the feedback on how delicate was the turns.

This would solve also the problem of sharp turns during the smooth curves, where the g-force would be added to the one already in effect. All of this is under the assumption that the device can continuously keep track of the g-force without resetting it after a period in which the variation detection occurred.