Master thesis presentation by Rasmus Åberg and Gabriel Matuszczyk, MPSYS: Smartphone based automatic incident detection algorithm and crash notification system for all terrain vehicle drivers

On Wednesday, 8 June Rasmus Åberg and Gabriel Matuszczyk, MPSYS, presents their master thesis with the title “Smartphone based automatic incident detection algorithm and crash notification system for all terrain vehicle drivers”.

When? 9 am. on 8 June, 2016
Where? Lunnerummet (room 3311), Hörsalsvägen 11, 3rd floor
Examiner: Stefan Candefjord


All-Terrain Vehicle (ATV) drivers face a different kind of danger than posed by most other means of travel. An ATV is mainly designed to travel forests and unpaved areas. It is a versatile vehicle often only used by one person at a time, meaning if an accident occurs far out in the wilderness help is hard to come by, especially if the driver is incapacitated. As a result of the prevalence of GPS-enabled smartphones, an application for them implementing an accurate Incident Detection Algorithm (IDA) could save even an unconscious driver, via an automatic message to an In Case of Emergency contact (ICE). This thesis investigates the possibility of designing such an application, as well as the feasibility of running it in real time on a smartphone.
Around 55 hours of normal ATV driving motion data is used, with the help of machine learning methods (specifically one class support vector machines (OC-SVM), to create an IDA that can satisfactorily identify several different types of accidents. Motion data was collected containing abnormalities in the form of a test person falling and rolling in several directions, in order to simulate a number of crash scenarios. Together with a cancellation algorithm to reduce false positives, and a decision tree within the IDA, few false alarms are raised while alarms do occur for all incidents simulated in the crash dataset. A design for a smartphone application to enable this automatic alarm is proposed in the form of a flow chart. Investigations of required functionality support the claim that a smartphone is capable of running such an IDA in real time and with low battery consumption. With the limited normal driving dataset, and only simulated crash data, further investigations need to be performed to ensure no overfitting has taken place. The next step in the development would be a test group consisting of regular ATV drivers to evaluate the performance of the IDA in real life situations. It is the authors’ opinion that with additional trials and tweaking of parameters, a well-functioning smartphone application could be released to the public and potentially serve as a life saver, perhaps even for other vehicles, in cases where the driver is otherwise helpless.