Data Analytics of ADAS Interventions

Background of thesis project
Advanced driver assistance systems (ADAS) play an important role in making the roads safer for everyone. By monitoring the truck behavior and surrounding environment, active safety features can predict traffic critical situations and intervene to prevent accidents from happening. However, accidents do still occur and when they do, data from truck sensors is logged and stored.

The hypothesis is that this data can be utilized to predict hazardous situations. By providing input to the HMI (Human Machine Interface) system display settings and warning strategies could be adapted with the aim of reducing the probability of incidents and creating a better driving experience.

This thesis will be running in parallel with a thesis that is centered around the adaptivity of the HMI output. Cooperation between the two is possible and encouraged.

Description of thesis work
The thesis aims at using historical data from logged traffic incidents and interventions to predict hazardous situations where the HMI could affect the driver and help make safer choices. The primary dataset includes short sequences of logged CAN data around the time of a function intervention and has been collected over several years and markets from customer trucks.

The dataset from historical events will likely mainly consist of difficult and/or hazardous situations. How to deal with an unknown and skewed dataset will be a key challenge. Solutions could include combining the current dataset with other internal data sources, collecting more data and augmenting the existing data.


The thesis could include, but not be limited to:
Data exploration of historic data from traffic incidents, function activations and other public and internal data sources to identify and correlate key features describing the events.
Build and tune/train a model using statistical learning, machine learning or other toolboxes for predicating hazardous traffic situations based on a historical dataset.
Apply and evaluate the model on logged timeseries data from trucks, emulating a real-world situation
The thesis is recommended for one or two students with an interest in active safety and data science.

The following skills would be highly valuable:
Python programming
SQL
Statistics
Data analytics
Machine learning



If you find this proposal interesting send your application with CV and grades to:
hans.deragarden@volvo.com


Thesis Level: Master

Language: English


Starting date: January 2023


Number of students: 1 or 2


Tutor
Contact persons:
Therese Gardell – Volvo GTT
tel: +46 73 80 80 332
mail: therese.gardell@volvo.com


Hans Deragården – Volvo GTT
tel: +46 31 3228054
mail: hans.deragarden@volvo.com
Kindly note that due to GDPR, we will not accept applications via mail. Please use our career site.