TJ Thomas

Lenworth Thomas, Tjaden Bridges, Sarah Bergbreiter

Abstract—Rapidly locating airborne chemicals is critical
in hazard response and environmental sensing. Robotic odor source localization (OSL) is one approach to solving this challenge, but implementing these algorithms on micro-quadrotors (∼ 50 g) is difficult due to the lack of payload capacity for larger sensor suites. As a result, systems at this scale have had
to assume known wind direction, use directional gas sensor
arrays to infer wind direction, or use swarms of quadrotors to
effectively localize odor sources. In this work, we introduce a
lightweight, crash-resistant, multi-axis wind sensor integrated
with a metal oxide gas sensor on the 32 g Crazyflie 2.1
Brushless. The wind sensor measures airflow direction, enabling
a new anemotaxis approach, Wind Servoing, which converts
wind direction to in-plane velocity commands for rapid plume
tracking. A series of characterization experiments verified
that the sensor has similar angular resolution to state-of-the
art whisker-inspired sensors but can also withstand crashes
common to micro-quadrotors. Across 45 different trials and
3 starting positions, Wind Servoing was compared to existing
baseline algorithms seeking an ethanol odor plume. Results
from these trials demonstrate that our approach is up to 50%
faster than existing OSL algorithms. To our knowledge, this is
the first demonstration of multi-axis wind sensing with wind
servoing on a Crazyflie-class (∼ 50 g) micro-quadrotor, opening
the door to agile, single-quadrotor chemical detection and rapid
hazard response.