Planning Set-Based Trajectory Envelops in Dynamic Driving Environments for Guaranteed-Safe Guardian Autonomy

Dealing with uncertainty in motion planning is a key challenge facing researchers in automated driving. This project seeks to further investigate this area by drawing on various domains from probabilistic to reachability-based methods. This project also specifically looks at planning in safe sets which would result in a Guardian system allowing the driver to act within the trajectory envelope, but would provide a correcting action were the driver attempting to exit the envelope. 

 

During the course of this project, we will develop algorithms to characterize the safe behavior of a vehicle in the presence of other drivers, bicyclists, and pedestrians in terms of safe trajectory envelopes. We will consider two approaches to constructing such trajectory envelopes: (1) reachability analysis and (2) probabilistic analysis. We will study methodologies to trade off safety and efficiency, possibly through the combination of these approaches. This project will also stress mathematically guaranteed performance, on-line computational speed, and experimental validation in simulation and on hardware systems. This will be demonstrated on a shared i-Road development platform with TRI.