This competition is designed to test the capabilities of learning-based robot decision-making algorithms to safely cope with events or uncertainties that are not known at design time. The task we consider is based on a nano quadrotor platform (Bitcraze’s Crazyflie). The quadrotor is required to navigate, as fast as possible, through an environment with a set of obstacles. Uncertainties the quadrotor may encounter include model uncertainties (e.g., parametric uncertainty in the robot’s mass and inertia), stochastic initial conditions, and external disturbances applied to states, inputs (e.g., noisy or biased sensing and actuation), and to the dynamics (e.g., wind). The participating teams are required to show that their algorithms can safely navigate the quadrotor through increasingly cluttered environments and can learn to cope with uncertainties.
The task execution is considered unsafe if it violates any of the safety constraints (e.g., violating a minimum safety distance from the obstacles, flying too close to the ground, leaving the boundaries of the flight arena, etc.). The evaluation criteria will be based on (1) the robot’s task performance (e.g., measured by its task completion time), (2) the satisfaction of the safety constraints, and (3) the maximal level of clutteredness in the environment that the algorithm can withstand.
The competition includes two simulation phases (virtual) and one experimental phase (through remote access). The fully virtual simulation components of the competition will be based on an open-source software benchmark suite. In the final experimental phase of the competition, we will provide remote access to our Flight Arena at the University of Toronto Institute for Aerospace Studies in Toronto, Canada.
- Angela Schoellig (University of Toronto, Vector Institute)
- Davide Scaramuzza (University of Zurich)
- Vijay Kumar (University of Pennsylvania)
- Nicholas Roy (Massachusetts Institute of Technology)
- Todd Murphey (Northwestern University)
- Sebastian Trimpe (RWTH Aachen University)
- Wolfgang Hönig (TU Berlin)
- Mark Muller (University of California Berkeley)
- Jose Martinez-Carranza (INAOE)
- SiQi Zhou (University of Toronto, Vector Institute)
- Melissa Greeff (University of Toronto, Vector Institute)
- Jacopo Panerati (University of Toronto, Vector Institute)
- Yunlong Song (University of Zurich)
- Leticia Oyuki Rojas Pérez (INAOE)
- Adam W. Hall (University of Toronto, Vector Institute)
- Justin Yuan (University of Toronto, Vector Institute)
- Lukas Brunke (University of Toronto, Vector Institute)