Towards a Multi-Pixel Time-of-Flight Indoor Navigation System for Nano-Drone Applications

Unmanned aerial vehicles (UAVs) have recently attracted the industry's attention due to their numerous civilian and potential commercial applications. A promising UAV subclass includes nano and micro UAVs, characterized by centimeter size, few grams of payload and extremely limited on-board computational resources. Those features pose major challenges to enable autonomous navigation or even more basic relevant subtasks, such as reliable obstacle avoidance. This paper explores and characterizes a multi-zone Time of Flight (ToF) sensor to enhance autonomous navigation with a significantly lower computational load than most common visual-based solutions. In particular, the state-of-the-art integrated ToF sensor is characterized for the first time in literature in-field using an ad hoc lightweight PCB and the Crazyflie nano-UAV. The paper focuses, on the 8x8 pixel configuration, to detect obstacles up to 3m with centimeter accuracy and a frame rate up to 15 fps. The paper presents a solution for computing the approaching angle, crucial for many UAV tasks, with a maximum error of +/- 6 degrees. Furthermore, relying on empirical data, the paper proposes a lightweight approach to calculate collision probability from each ToF sensor frame. This work aims to pave the way for future nano-UAV camera-less compact navigation solutions capable of extracting complex environmental features directly on-board.