BatDeck -- Ultra Low-power Ultrasonic Ego-velocity Estimation and Obstacle Avoidance on Nano-drones (2412.10048v1)

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.

• The paper introduces an ultrasonic ego-velocity estimation method that achieves an RMSE of 0.019 m/s on featureless surfaces.
• The paper demonstrates that dual ultrasonic sensors enable comprehensive obstacle avoidance over 136 meters of flight in challenging environments.
• The paper highlights that integrating ultrasonic sensing with nano-drones significantly reduces power consumption while enhancing navigation robustness.

Ultrasonic Sensing for Nano-Drone Navigation

The paper explores enhancements in nano-drone navigation through the integration of ultrasonic sensing techniques inspired by biological echolocation. Recognizing the size and payload constraints typical of nano-drones, it focuses on improving ego-velocity estimation and obstacle avoidance (OA) using ultra-low-power ultrasonic sensors. These contributions are pivotal given the limitations of conventional vision-based systems in certain environmental conditions, such as low-light scenarios or surfaces that exhibit minimal optical features.

Ultrasonic Ego-Velocity Estimation

The paper introduces a novel method for ego-velocity estimation leveraging phase differences in ultrasonic signals received from two synchronized sensors. The system utilizes ultra-compact TDK ICU-x0201 sensors mounted on a custom BatDeck platform on the Crazyflie 2.1 drone. It demonstrates the system's ability to estimate velocity with a root mean squared error (RMSE) of 0.019 m/s on featureless surfaces, outperforming traditional optical flow sensors that exhibit inaccuracies under similar conditions. This approach is particularly useful in environments where visual cues are absent or disrupted, maintaining velocity estimation accuracy irrespective of lighting differences or surface textures.

Obstacle Avoidance with Ultrasonics

The application of ultrasonic sensors for OA was tested through a series of trials where the drone navigated a complex, obstacle-rich environment. The results indicated a substantial improvement over laser-based ToF sensors, primarily in detection scenarios involving highly reflective or transparent surfaces, which are typical pitfalls for vision and certain ToF systems. The dual-sensor configuration provided a comprehensive spatial awareness capable of navigating across an average flight distance of 136 meters without collision under battery-drained conditions.

Technical and Practical Implications

From a technical standpoint, the integration of ultrasonic sensors presents a substantial reduction in power consumption with notably low processing and latency overheads, which are critical metrics for resource-constrained platforms like nano-drones. The implementation remains within acceptable computational limits on an ARM Cortex-M4 processor, which facilitates more energy-efficient operations in environments where navigation capabilities must align with constrained sensor packages and limited drone platforms.

In practical terms, the implications of such advancements extend beyond the immediate enhancement of navigation systems for nano-drones. The fusion of ultrasonic sensing with the traditional optical and inertial sensors promises to extend the operational range and effectiveness of autonomous drones in various fields, such as indoor navigation, search and rescue missions, and possibly extending applications into agriculture or infrastructure inspections where environments may not be predictable.

Future Prospects in AI-Driven Navigation Systems

The fusion of multiple sensing modalities to create more robust, adaptable, and energy-efficient sensing systems for drones could see further advances with the incorporation of AI-driven sensory processing. The ability to pre-process and infer navigation data on-chip, with minimal computational load, could enhance real-time decision-making and improve the adaptability of drones to dynamic environments. Machine learning models, such as CNNs or reinforcement learning paradigms, might streamline sensor fusion techniques further, providing enhanced navigation protocols adaptable to various scenarios beyond current capabilities.

In conclusion, the paper provides substantive work into the advancement of nano-drone technology by introducing biologically inspired, low-power ultrasonic sensors. It opens avenues for more sophisticated navigation systems, tackling limitations traditionally associated with laser and camera-based methods, thereby enriching the discourse on autonomous navigation robustness in complex environments.