MicNest: Long-Range Instant Acoustic Localization of Drones for Precise Landing (Best Paper Runner‑Up at ACM SenSys 2023)

The paper "MicNest: Long‑Range Instant Acoustic Localization of Drones in Precise Landing" was awarded the Best Paper Runner‑Up at the 20th ACM Conference on Embedded Networked Sensor Systems (SenSys) held in Boston, USA (Nov 6‑9, 2023). The work is a joint effort by Tsinghua University, Meituan’s drone team, and Politecnico di Milano.

Motivation: In urban low‑altitude logistics, high‑precision positioning is essential for safe drone take‑off and landing. Conventional GNSS‑based methods (RTK, visual) suffer from signal blockage and multipath in dense city canyons. The authors propose an acoustic‑based positioning system to improve robustness.

Overall framework: Drones emit acoustic pulse signals via a small speaker. A ground station equipped with an array of microphones (four anchors placed at the corners of the landing platform) captures the pulses. By detecting the pulse arrival times at each microphone, the system computes Time‑of‑Arrival (ToA) and Time‑Difference‑of‑Arrival (TDoA) to estimate the drone’s 3‑D position.

Key technical challenges addressed:

Very low signal‑to‑noise ratio (SNR) due to limited speaker power, long propagation distance (>100 m), strong urban background noise (40‑75 dB SPL), and propeller noise (up to 104 dB SPL).

Severe Doppler distortion caused by the slow propagation speed of sound compared with RF signals.

Real‑time processing requirements: positioning latency must be minimal to avoid instability in the flight controller.

Solution details:

PRN modulation and transmission: Each drone is assigned a unique ID that seeds a pseudo‑random noise (PRN) generator. The PRN sequence is up‑sampled to the speaker’s 48 kHz sampling rate, producing a long‑duration acoustic pulse that is friendly to human hearing.

Pulse detection: A matched filter using the transmitted PRN pulse as a template performs correlation on the received microphone signals. To compensate for Doppler distortion, the system resamples the template according to hypothesized radial velocities and selects the velocity that yields the highest correlation peak.

TDoA estimation and positioning: The four microphones (Mic0‑Mic3) provide ToA measurements. TDoA is computed for the diagonal pairs (Mic0‑Mic2 and Mic1‑Mic3) because they have the largest aperture, yielding the finest spatial resolution. The two TDoA values are sent to the drone via Wi‑Fi, where they are used to solve a pair of hyperbolic equations together with the known flight altitude, producing the drone’s horizontal coordinates.

Innovations compared with visual‑based methods:

Acoustic signals are unaffected by lighting conditions.

Acoustic radiation is omnidirectional, giving a larger horizontal coverage.

The approach theoretically supports simultaneous localization of multiple drones.

Results: The system achieves a positioning height of up to 120 m with a relative error of 0.5 % in complex urban environments. It has been integrated into Meituan’s drone flight‑control system and validated in diverse real‑world scenarios. The technology is slated for deployment in Meituan’s city‑wide instant delivery services.

Funding and collaboration: The research was supported by the Tsinghua‑Meituan Digital Life Joint Research Institute and the National Natural Science Foundation of China, involving researchers from Tsinghua University, Meituan, and Politecnico di Milano.