University of Twente Student Theses

Login

Assessing road quality and driver behavior using movement sensors in wrist worn devices

Zwet, Dennis van der (2020) Assessing road quality and driver behavior using movement sensors in wrist worn devices.

[img] PDF
1MB
Abstract:In current society, people are involved in traffic every day. In order to travel to go to work, do the groceries, meet friends, study, visit a forest, in all these situations it is necessary to be involved in traffic. Because of this, it is very important that traffic is as safe as it can possibly be. Unsafe situations can be caused by many factors. An important factor is road quality: poorly maintained roads can cause damage to vehicles and even be a cause for accidents. Moreover, the way the road has been laid out can be very important for safety. Hard breaking is a potential cause for accidents. This means that roads that require hard breaking can be a cause for accidents to happen. The fact that it is difficult and labor intensive to check every road for damage and unsafe situations has become problematic. An automated way to detect these situations by people driving on the road would be a possible solution to increase awareness of which roads maintenance. Another important cause of accidents is driver behaviour. Distracted driving is one of the most important causes of accidents. However, it is easy to be distracted while driving. An automated system to notify a driver that they are being distracted could be beneficial for traffic safety. To improve these two causes of accidents, a solution must be created to automatically detect road safety, as well as driver behaviour. Therefore, this paper tries to answers two questions: • Can we detect unsafe situations on the road using smartwatches? • Can we detect the behavior of a driver using smartwatches? This paper proposes using sensors already available in wrist worn devices to measure both road safety and driver behaviour. Based on a literature review of existing solutions, this paper investigates three distinct machine learning techniques: Support Vector Machines, Long Short-Term Memory model and the InceptionTime model. Existing options are surveyed, after which the paper describes its own solution to the problems. The conclusion of the paper is that current methodology needs more improvements to create a viable solution.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science MSc (60300)
Link to this item:http://purl.utwente.nl/essays/84828
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page