Driver Behaviour Learning (DBL)

This is a collaborative project with industry partner Magna International to collect real world data about human driving behaviour in order to enable training of accurate predictive models of human driving for modern ADAS.

DOMAINS | autonomous-driving | driver-behaviour-learning

FIELDS | machine-learning | deep-learning | human-study | time-series

METHODS | deep-neural-networks | lstm

WEBPAGE: | https://uwaterloo.ca/scholar/mcrowley/dblstudy | /news/2021-10-14-AutolineInterview/

The Real-Time Embedded Software Group and UWECEML labs at the Department of Electrical and Computer Engineering at the University of Waterloo are working to conduct a study on driving behaviour. This project is part of a research collaboration between the University of Waterloo and Magna International Inc. to develop a driving model for self-driving vehicles based on learning from human driver behaviour on roads using a variety of automotive sensors, data fusion and artificial intelligence algorithms. The goal is to create a comfortable in-vehicle experience when a self-driving mode is engaged. The project requires monitoring and recording the driving behaviour of 100 drivers using a specially equipped SUV (LiDAR, Radar, Cameras and more) in order to have sufficient data to create a self-learning model using machine learning algorithms.

Get Involved: See the project webpage to find out more or to sign up to be a driving participant in our study.

Learning Problems

Problem 1: Given an observed driving execution with a known driving plan, efficiently and effectively extract a driver behavior model.

Problem 2: Given a driving plan and a driver model, predict the vehicle state at points on the driving plan.