TRI Liaison: Avinash Balachandran
This project seeks to understand feedback interactions between groups of autonomous decision makers (both humans and machines), and to use this understanding to design decision and control policies for autonomous driving that are guaranteed to be safe in the presence of these feedback interactions. We will consider three specific kinds of interactions: (1) human-machine shared autonomy in one vehicle, (2) pairwise interactions between two vehicles, and (3) large-scale interactions among large numbers of vehicles. We will validate our theoretical and algorithmic results at every stage with experiments on test car platforms available through the PIs’ existing labs. This project will advance the science of cooperative autonomy, while also directly augmenting Toyota’s autonomous driving capabilities.
- Human-car interactions: Design safe assistive driving policies that explicitly consider coupled dynamic interactions between the car and the human driver.
- Pairwise Interactions: To develop decision policies for common maneuvers that are proven, analytically and experimentally, to be safe in pair?wise feedback interactions between two autonomous vehicles, or between one autonomous and one human?driven vehicle.
- Large-Scale Interactions: To produce theoretically proven and experimentally validated composition rules for large groups of both human and autonomous drivers in realistic traffic conditions.