Research
Behavior Modeling for Autonomous Agents in Virtual
Environments
Our
bicycling studies serve as a rich source of motivation for computational
research on methods to control the behaviors of autonomous vehicles and
pedestrians, and to create robust, replicable scenarios for experiments.
This includes methods to represent simulated roadways,
methods to control the behaviors of semi-autonomous vehicles that
populate the virtual environment, and techniques to create controlled
traffic patterns in order to study bicycling behavior.
Our
research highlights the relationship between behavior programming and
roadway modeling. We represent
roadway surfaces as three-dimensional ribbons that make the local
orientation of the road explicit and allow relative distances on the road to
be simply computed. Roads and intersections are connected to form an
interconnected network of ribbons.
Vehicle behaviors are simplified by representations that facilitate
both short-term steering and long-term wayfinding.
We’ve
developed multi-component behaviors that plan routes and safely navigate
through traffic filled road networks – tracking lanes, shifting lanes to
avoid congestion and prepare for upcoming turns, negotiating intersections,
and respecting the rules of the road.
P. Willemsen,
Joe Kearney,
and H. Wang