BatDeck - Ultra Low-power Ultrasonic Ego-velocity Estimation and Obstacle Avoidance on Nano-drones
Hanna Müller 1 , Victor Kartsch 1 , Michele Magno 1 , Luca Benini 1,2 · 2024
Read BatDeck - Ultra Low-power Ultrasonic Ego-velocity Estimation and Obstacle Avoidance on Nano-drones by Hanna Müller 1 , Victor Kartsch 1 , Michele Magno 1 , Luca Benini 1,2 (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-10048.txt.
Introduction
Hanna Muller, Victor Kartsch, Michele Magno, and Luca Benini presented BatDeck in 2024, an ultrasonic sensing deck for nano-drones. Tiny drones suit tight-space rescue and are safe near people, but their small payload limits sensing. Laser and camera sensors fail on clear or shiny surfaces and in the dark.
Taking a cue from bats, the team used low-power ultrasonic sensors facing down and forward to gauge speed and dodge obstacles. The system measured speed with low error and flew for eight minutes in tests, covering a good distance among clear and shiny obstacles. Comparisons with laser and optical methods showed the approaches can back each other up for safer flight.
Catalog record
| CREATOR | Hanna Müller 1 , Victor Kartsch 1 , Michele Magno 1 , Luca Benini 1,2 |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-10048
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
BatDeck - Ultra Low-power Ultrasonic Ego-velocity Estimation and Obstacle Avoidance on Nano-drones Thanks: Hanna Müller 1 , Victor Kartsch 1 , Michele Magno 1 , Luca Benini 1,2 Abstract Nano-drones, with their small, lightweight design, are ideal for confined-space rescue missions and inherently safe for human interaction. However, their limited payload restricts the critical sensing needed for ego-velocity estimation and obstacle detection to single-bean laser-based time-of-flight (ToF) and low-resolution optical sensors. Although those sensors have demonstrated good performance, they fail in some complex real-world scenarios, especially when facing transparent or reflective surfaces (ToFs) or when lacking visual features (optical-flow sensors). Taking inspiration from bats, this paper proposes a novel two-way ranging-based method for ego-velocity estimation and obstacle avoidance based on down-and-forward facing ultra-low-power ultrasonic sensors, which improve the performance when the drone faces reflective materials or navigates in complete darkness. Our results demonstrate that our new sensing system achieves a mean square error of 0.019 m/s on ego-velocity estimation and allows exploration for a flight time of 8 minutes while covering 136 m on average in a challenging environment with transparent and reflective obstacles. We also compare ultrasonic and laser-based ToF sensing techniques for obstacle avoidance, as well as optical flow and ultrasonic-based techniques for ego-velocity estimation, denoting how these systems and methods can be complemented to enhance the robustness of nano-drone operations. I Introduction The projected growth in the UAV market through 2030 reflects a rising interest among businesses and individuals in the diverse solutions that UAV technology can enable in different domains such as agriculture, public safety, urban planning, and entertainment [ 1 ] . Specifically, nano-drones (typically \sim 10\text{\,}\mathrm{cm} diameter [ 2 ] ) allow exploring constrained environments like greenhouses and buildings, while being safe for people and property. Drones require precise state information—including position, velocity, and orientation—to execute missions reliably. Achieving precision involves algorithms that integrate data from multiple sensors. Algorithms like simultaneous localization and mapping (SLAM) [ 3 ] and visual-inertial odometry (VIO) [ 4 ] combine visual data (from cameras) and inertial data to continuously track a drone’s position and orientation. The extended Kalman filter (EKF) offers a computationally efficient approach to resource-limited platforms, such as nano-drones, by fusing data from various sensors, including inertial measurement units (IMUs) and low-resolution optical flow sensors. Even though there have been advances in optical flow robustness in recent years [ 5 , 6 ] , estimates can be severely affected when optical conditions for optical flow are suboptimal, such as in low-light or feature-less settings. Ultrasound, which measures the time-of-flight (ToF) in echos of high-frequency sound, has the potential for providing robust state estimation, even without optical features [ 7 ] . A wide range of ultrasonic sensors applicable for use in robotics are currently available on the market and have been examined in the literature [ 8 , 9 , 10 , 7 ] . However, their size, weight, and energy requirements preclude their use in nano-drone applications. Fig. 1 : The Crazyflie 2.1 with the BatDeck and the Flow-deck v2, equipped for OA (top) and ego-velocity estimation (bottom). The newly introduced ICU-x0201 sensors from TDK (as shown in our setup in Figure 1 ) overcome these limitations with their ultra-low power consumption (below 1\text{\,}\mathrm{mW} ) and compact form factor ( 5.17\text{\,}\times 2.68\text{\,}\times 0.9\text{\,}\mathrm{mm} ), making them an ideal choice for integration into small-scale robotic platforms [ 11 ] . However, for using a Doppler-shift-based velocity estimation as used in [ 7 ] , a highly directed ultrasonic wave is needed, leading to bulky and heavy acoustic horns, which are infeasible on nano-drones. Hence there is a need to develop a novel ego-velocity method with ultrasonic sensors for nano-drones. Obstacle avoidance (OA) is also critical to ensure the drone platform’s safe and efficient operation. Often, OA is performed with high-resolution sensors as light detection and ranging (LiDAR) , Radio Detection and Ranging (RADAR) or cameras [ 12 , 13 ] . However, LiDAR and RADAR are bulky and power-hungry, and are thus unsuitable for nano-drones applications. Laser-based ToF sensors have also been featured in systems coupled with naturally-inspired OA algorithms, providing a low-power and robust performance [ 14 , 15 ] . Still, camera and laser-based solutions are heavily impaired when dealing with reflective surfaces or transparent barriers such as glass walls. A viable solution to address these robustness challenges is to perform sensor fusion of multiple sensors, but nano-drones that are typically below 50\text{\,}\mathrm{g} can only resort to a small and lightweight set of sensors given payload and power constraints [ 2 ] . Inspired by the echolocation techniques used by bats and dolphins, ultrasonic sensors — which measure ToF using echoes of high-frequency sound — offer a promising alternative, excelling with sound-reflective objects regardless of color, transparency, or texture [ 16 , 17 ] . This work explores equipping nano-drones with bat-like navigation capabilities by leveraging a novel, compact, and low-power ultrasonic sensor. It extends previous work [ 18 ] by introducing a novel method for ego-velocity estimation. Specifically, our contributions are: 1. Design and implementation of an ultrasonic hardware extension system (PCB) for nano-drones ( BatDeck ), integrating up to four compact and energy-efficient ICU-x0201 sensors that enable emitting and receiving ultrasonic signals. 2. Implementation and evaluation of an OA algorithm tailored for resource-constrained lightweight hardware (ARM Cortex-M4) onboard a nano-drone in terms of power and latency. 3. In-field evaluation and comparison of BatDeck with laser-based ToF sensors for OA applications, based on mission success rate, power consumption, and adaptability to various environments, providing insights into the advantages and potential limitations of both approaches in a real-world scenario. 4. Development and implementation, including the mathematical background, of a novel ego-velocity estimation algorithm relying on two ground-facing ultrasonic sensors. 5. Proof-of-concept in-field evaluation and comparison of the ultrasonic ego-velocity estimation to an optical-flow-based sensor for velocity estimation in different environments, highlighting how these sensors can be complemented to provide a more robust drone state estimation. II Related Work Ego-velocity estimation involves determining an object’s speed and movement direction relative to its environment, essential for navigation, obstacle avoidance, and maintaining stability in dynamic settings. Numerous studies have tacked the estimation of accurate ego-motion parameters based on various sensing modalities. For instance, Shan et al. [ 19 ] proposed a probabilistic framework utilizing lidar and IMU data to correct motion-induced distortions in lidar point clouds, thereby improving ego-velocity reliability. Camera-based systems, often when combined with IMUs , have proven effective for ego-motion and ego-velocity estimation as a part of SLAM , VIO [ 3 , 4 ] or convolutional neural networks (CNNs) . In [ 20 ] , authors address ego-motion and depth estimation for UAV navigation in dynamic environments by introducing Cowan-GGR, a deep learning-based framework integrating optical flow, monocular depth, and IMU-based pose updates to enhance ego-motion estimation, particularly by using geometric relationships between depth and translation. Similarly, radar sensors have demonstrated reliability in velocity estimation; for instance, Park et al. [ 21 ] introduced a mmWave Doppler radar system with two orthogonal radar modules to capture 3D motion, supporting accurate measurements in foggy or unstructured outdoor environments. However, these solutions are generally too bulky and power-intensive for application in nano-drones. Cameras have also