SuperBat - Advancing Obstacle Avoidance on Nano-UAVs by Fusing Ultrasonic and Laser-based Time-of-Flight Sensors

Abstract

Nano-Unmanned Aerial Vehicles (UAVs) hold significant promise in various applications, such as exploring ducts or aiding rescue missions in proximity to humans. Their small dimensions make them ideal to operate indoor but pose challenges in terms of payload capacity and onboard computation power. Establishing a robust Obstacle Avoidance (OA) onboard algorithm, able to operate reliably across diverse environments with various materials including transparent and reflective surfaces, is still an open challenge. Recent approaches have pointed to sensor fusion as a path towards reliable OA. In this paper, we fuse a 64-pixel laser-based Time-of-Flight (ToF) sensor (VL53L5CX) and an ultrasonic sensor (ICU-30201) in our computationally lightweight navigation policy to achieve robust and efficient OA. The field tests are based on the Crazyflie 2.1 platform (CF), resulting in a total flight mass of 38 g. Experimental evaluation on the CF shows a reliability of 80 % even in an unexplored indoor environment, including transparent and reflective obstacles, while flying at a maximum speed of 1 m/s. Additionally, the CF is able to fly through a 75 cm narrow corridor with 100% reliability. These findings underscore the efficacy of the fused laser-based ToF and ultrasonic sensors, even under relatively high-speed flight conditions, enabling robust OA on miniaturized robotic platforms.