Title ActNormal
Summary Driving Behavior Model Identification
Supervisor Tony Larsson, Stefan Byttner, Cristofer Englund
Level Master
Status Open

Vehicle driving can be assisted to make the driving more cooperative, safer and energy efficient. A hypothesis is that if we “act normal” we can drive more safely and energy efficient and thus if a driver deviates too much from this norm the driver should be informed. To distinguish abnormal driver behaviors from the more normal a model of driving and traffic behavior with acceptable deviations is needed.

Research Questions: 1) How combine and process context awareness messages in a server sent from vehicles via the 3/4G radio communication network? 2) How create a space-time mapping of the driving behavior? 3) What message information and periodicity is needed to sufficiently map the “normal” behavior and deviation intervals along a road?

Expected Results: 1) A system architecture level description. 2) A smartphone client “app” that periodically sends CAM messages to a server via the 3/4G network. 3) A server “app” that creates a norm model for each road segment at different time intervals, i.e. a model logged in a space-time map. 4) A smartphone client “app” that compares a driver’s behavior to the norm for a specific road-time segment acquired from the server and gives a warning message if a dangerous deviation is detected. 5) The system consisting of the above “apps” tested in a limited scenario like a few blocks in a city or a few km of a rural 2-lane road with a few crossings. 6) An analysis of the relations between sampling accuracy, periodicity and detection sensitivity.