Driver motion prediction using behaviour classifications of vehicles

A significant portion of decision making, path planning and navigation algorithms for Autonomous Vehicles (AV) rely heavily on accurate estimation of the current location as well as future trajectories of the surrounding road users. There are different kinds of drivers in urban environments, and an expert human driver will identify dangerous drivers and avoid them accordingly. However, existing autonomous driving systems often treat all neighboring vehicles the same and do not take actions to avoid the dangerous drivers.
For active safety and reduced reaction times, Gatik’s AVs need to accurately predict the behaviours of surrounding agents to be able to make safe & reliable complex decisions such as merging, unprotected left turns, lane change,
etc
The goal of this research project is to develop new techniques for enabling accurate & reliable driver behaviour
prediction to ensure safer reactions in avoiding dangerous neighboring drivers, pedestrians and cyclists, and
efficient navigation around careful drivers.

Faculty Supervisor:

Krzysztof Czarnecki

Student:

Prarthana Bhattacharyya

Partner:

Gatik Inc

Discipline:

Engineering - computer / electrical

Sector:

Professional, scientific and technical services

University:

University of Waterloo