Mission-level Robustness with Rapidly-deployed, Autonomous Aerial Vehicles

by Carnegie Mellon Team Tartan at MBZIRC 2020

Anish Bhattacharya    Akshit Gandhi    Lukas Merkle    Rohan Tiwari    Karun Warrior    Stanley Winata    Andrew Saba    Kevin Zhang

Oliver Kroemer    Sebastian Scherer

The published paper can be found here, on the Field Robotics website. We present the following contributions:

• strategies for simple, yet robust algorithms in computer vision, vision-based control and guidance, and mission planning;

• an autonomous system-startup and mission-launching architecture that facilitates the rapid deployment of our robots in time-constrained and WiFi-denied settings;

• show that mission plans containing self-check cycles helped us overcome faulty hardware and failed subtasks and still succeed in the overall mission;

• thorough assessments of failure cases at MBZIRC 2020 and describe critical lessons learned from field development and testing.

In the supplement, we present work done beyond the competition: a post-competition simulation analysis of our ball-catching strategy and the development and assessment of a better control strategy, and a real-world UAV-UGV cooperative block stacking strategy.

Feel free to contact Anish Bhattacharya with any questions! This is a collaboration between the AirLab and the Intelligent Autonomous Manipulation Lab.