AURORA

Acoustic UAV Reconnaissance & Orientation for Rescue Assistance

UAV-assisted disaster response that localizes human distress calls in low-visibility environments.

FPGASignal ProcessingBeamformingEmbedded SystemsAcoustic Localization

<100 ms

Latency

2nd Runner-Up

Result

2026

Year

Overview

AURORA is a UAV-mounted acoustic sensing system designed for search-and-rescue scenarios where conventional vision-based approaches become unreliable due to smoke, dust, darkness, or collapsed structures. Instead of relying solely on cameras or thermal imaging, the system uses a synchronized microphone array to capture spatial audio and estimate the direction of human distress signals in real time.

The project combines FPGA-assisted preprocessing, embedded computing, beamforming, and sound source localization techniques to transform raw multi-channel audio into actionable navigation guidance, enabling faster and more reliable acoustic-assisted rescue operations.

Problem

In disaster environments such as collapsed buildings, smoke-filled spaces, or low-visibility conditions, vision-based search systems often become ineffective. Human voices and distress sounds, however, remain valuable indicators of survivor locations. AURORA addresses this challenge by enabling UAVs to detect, localize, and estimate the direction of human-generated sounds using acoustic signal processing, allowing rescue teams to navigate more effectively in environments where traditional imaging techniques fall short.

Process

Mic array capture

A multi-microphone array mounted on the UAV captures spatial audio across all channels in tight sync — timing alignment across channels is the foundation everything downstream depends on, since DOA/TDOA estimation is only as accurate as the sample-level sync between microphones.

FPGA preprocessing

Multi-channel audio acquisition and early-stage filtering are handled through FPGA-based preprocessing to support low-latency signal handling before higher-level processing on the embedded processor. This hardware-software partitioning improves overall responsiveness while reducing the computational load on the embedded system.

DOA / TDOA estimation

Direction-of-Arrival (DOA) and Time-Difference-of-Arrival (TDOA) techniques are applied to estimate the direction of incoming sound sources. Combining both approaches improves localization accuracy while maintaining real-time performance.

Beamforming & noise rejection

Beamforming and frequency-selective filtering enhance human speech while suppressing UAV motor noise and surrounding environmental interference, improving the reliability of acoustic localization.

Navigation command generation

The estimated source direction is translated into UAV orientation guidance — a heading a rescue operator (or, eventually, an autonomous flight controller) can act on directly, rather than a raw angle number that still needs interpretation.

Core Concepts

Sound Source LocalizationDirection of Arrival (DOA)Time Difference of Arrival (TDOA)BeamformingFPGA-Based ProcessingUAV Navigation Assistance

Key Challenges Solved

UAV motor ego-noise suppressionReal-time acoustic processing under a strict latency budgetHuman voice isolation in a noisy, moving-platform environmentAutonomous sound-guided orientation output

Results

Demonstrated real-time acoustic localization with end-to-end processing latency below 100 ms.

Validated the effectiveness of beamforming and noise suppression techniques for improving human voice detection in UAV operating conditions.

Successfully integrated FPGA-assisted preprocessing with embedded signal processing for efficient real-time performance.

Received recognition as a national-level engineering project for its innovative application of acoustic localization in disaster response.

Learnings

Suppressing UAV motor noise proved to be one of the most challenging aspects of the system, requiring noise-aware signal processing rather than relying solely on localization algorithms.

FPGA-assisted preprocessing highlighted the importance of hardware acceleration for low-latency embedded applications involving high-bandwidth sensor data.

Acoustic sensing can effectively complement vision-based systems in disaster response, particularly in environments where visibility is severely limited.

Developing a multidisciplinary system reinforced the importance of defining clear interfaces between embedded hardware, signal-processing algorithms, and software components.