Machine Learning-based Methods for Driver Identification and Behavior Assessment: Applications for CAN and Floating Car Data

Doctoral thesis (2020)

The exponential growth of car generated data, the increased connectivity, and the advances in artificial intelligence (AI), enable novel mobility applications. This dissertation focuses on two use-cases ... [more ▼]

The exponential growth of car generated data, the increased connectivity, and the advances in artificial intelligence (AI), enable novel mobility applications. This dissertation focuses on two use-cases of driving data, namely distraction detection and driver identification (ID). Low and medium-income countries account for 93% of traffic deaths; moreover, a major contributing factor to road crashes is distracted driving. Motivated by this, the first part of this thesis explores the possibility of an easy-to-deploy solution to distracted driving detection. Most of the related work uses sophisticated sensors or cameras, which raises privacy concerns and increases the cost. Therefore a machine learning (ML) approach is proposed that only uses signals from the CAN-bus and the inertial measurement unit (IMU). It is then evaluated against a hand-annotated dataset of 13 drivers and delivers reasonable accuracy. This approach is limited in detecting short-term distractions but demonstrates that a viable solution is possible. In the second part, the focus is on the effective identification of drivers using their driving behavior. The aim is to address the shortcomings of the state-of-the-art methods. First, a driver ID mechanism based on discriminative classifiers is used to find a set of suitable signals and features. It uses five signals from the CAN-bus, with hand-engineered features, which is an improvement from current state-of-the-art that mainly focused on external sensors. The second approach is based on Gaussian mixture models (GMMs), although it uses two signals and fewer features, it shows improved accuracy. In this system, the enrollment of a new driver does not require retraining of the models, which was a limitation in the previous approach. In order to reduce the amount of training data a Triplet network is used to train a deep neural network (DNN) that learns to discriminate drivers. The training of the DNN does not require any driving data from the target set of drivers. The DNN encodes pieces of driving data to an embedding space so that in this space examples of the same driver will appear closer to each other and far from examples of other drivers. This technique reduces the amount of data needed for accurate prediction to under a minute of driving data. These three solutions are validated against a real-world dataset of 57 drivers. Lastly, the possibility of a driver ID system is explored that only uses floating car data (FCD), in particular, GPS data from smartphones. A DNN architecture is then designed that encodes the routes, origin, and destination coordinates as well as various other features computed based on contextual information. The proposed model is then evaluated against a dataset of 678 drivers and shows high accuracy. In a nutshell, this work demonstrates that proper driver ID is achievable. The constraints imposed by the use-case and data availability negatively affect the performance; in such cases, the efficient use of the available data is crucial. [less ▲]

Non-intrusive Distracted Driving Detection Based on Driving Sensing Data

in 4th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2018) (2018, March)

Nowadays Internet-enabled phones have become ubiquitous, and we all witness the flood of information that often arrives with a notification. Most of us immediately divert our attention to our phones even ... [more ▼]

Nowadays Internet-enabled phones have become ubiquitous, and we all witness the flood of information that often arrives with a notification. Most of us immediately divert our attention to our phones even when we are behind the wheel. Statistics show that drivers use their phone on 88% of their trips and on 2015 in the UnitedKingdom 25% of the fatal accidents were caused by distraction or impairment. Therefore there is need to tackle this issue. However, most of the distraction detection methods either use expensive dedicated hardware and/or they make use of intrusive or uncomfortable sensors. We propose distracted driving detection mechanism using non-intrusive vehicle sensor data. In the proposed method 9 driving signals are used. The data is collected, then two sets of statistical and cepstral features are extracted using a sliding window process, further a classifier makes a prediction for each window frame, lastly, a decision function takes the last l predictions and makes the final prediction. We evaluate the subject independent performance of the proposed mechanism using a driving dataset consisting of 13 drivers. We show that performance increases as the decision window become larger.We achieve the best results using a Gradient Boosting classifier with a decision window of total duration 285seconds which yield ROC AUC of 98.7%. [less ▲]

An Open Dataset for Human Activity Analysis using Smart Devices

Report (2017)

The study of human mobility and activities has opened up to an incredible number of studies in the past, most of which included the use of sensors distributed on the body of the subject. More recently ... [more ▼]

The study of human mobility and activities has opened up to an incredible number of studies in the past, most of which included the use of sensors distributed on the body of the subject. More recently, the use of smart devices has been particularly relevant because they are already everywhere and they come with accurate miniaturized sensors. Whether it is smartphones, smartwatches or smartglasses, each device can be used to describe complementary information such as emotions, precise movements, or environmental conditions. In this short paper, we release the applications we have developed and an example of a collected dataset. We propose that opening multi-sensors data from daily activities may enable new approaches to studying human behavior. [less ▲]

A Car Hacking Experiment: When Connectivity meets Vulnerability

in Globecom Workshops (GC Wkshps), 2015 IEEE (2015, December)

Interconnected vehicles are a growing commodity providing remote access to on-board systems for monitoring and controlling the state of the vehicle. Such features are built to facilitate and strengthen ... [more ▼]

Interconnected vehicles are a growing commodity providing remote access to on-board systems for monitoring and controlling the state of the vehicle. Such features are built to facilitate and strengthen the owner’s knowledge about its car but at the same time they impact its safety and security. Vehicles are not ready to be fully connected as various attacks are currently possible against their control systems. In this paper, we analyse possible attack scenarios on a recently released all-electric car and investigate their impact on real life driving scenarios. We leverage our findings to change the behaviour of safety critical components of the vehicle in order to achieve autonomous driving using an Open Vehicle Monitoring System. Furthermore, to demonstrate the potential of our setup, we developed a novel mobile application able to control such vehicle systems remotely through the Internet. We challenge the current state-of-the-art technology in today’s vehicles and provide a vulnerability analysis on modern embedded systems. [less ▲]