In our lab, we investigate how robots can operate safely and reliably in unpredictable, hazardous, or structurally uncertain environments. Our goal is to equip autonomous systems with the ability to sense, evaluate, and adapt to risk in real time, enabling informed navigation decisions in complex indoor spaces.
A major research effort in this area is led by Erik Parker. This project leverages high-frequency vibration data collected from distributed accelerometer modules to construct a dynamic risk-tracking map of a building. These compact sensor modules are deployed throughout an environment by the Spot robot and communicate wirelessly with a centralized base station. The base station runs custom signal-processing and estimation algorithms that analyze each location’s impulse response and predict the relative risk associated with potential navigation paths.
The graph above displays the frequency-domain response of a sensor module mounted on an actuator excited with pure tones at 1 kHz, 1.5 kHz, 2 kHz, and 3 kHz. The resulting FFT (Fast Fourier Transform) confirms that the system accurately captures and preserves key frequency components, even after signal conditioning and filtering. This validates the integrity of the sensing and processing pipeline.
Several pre-processing techniques are currently implemented, including wavelet-based denoising and band-pass filtering to isolate structurally relevant frequency bands. Additional comparative algorithms are under active development to improve robustness, classification accuracy, and real-time performance.
The image above shows the custom-designed PCB developed for this project. At its core is an ESP32-S3 microcontroller, which manages data acquisition, wireless communication, and onboard processing. The primary sensing element is an IIS3DWB high-bandwidth accelerometer capable of capturing detailed vibration signatures. The board also integrates impulse actuator control, a self-test signal generator, and a lithium battery charge management circuit, enabling fully autonomous field deployment.
The render above illustrates the proposed experimental test environment used to validate system performance. The setup consists of wooden platforms with varying levels of mechanical stability, along with wall structures to simulate realistic indoor conditions. The system is designed to distinguish between stable and unstable surfaces by analyzing their vibrational characteristics.
The image above shows one of the constructed wooden platforms prior to wall installation. These platforms serve as controlled test surfaces for collecting baseline and comparative vibration data under different structural conditions.
Ongoing work focuses on expanded testing, refinement of risk-estimation algorithms, and integration of the full system with autonomous navigation routines. The ultimate objective is to enable robots to assess structural integrity in real time and make safer path-planning decisions in complex built environments.