Privacy-Preserving Group Coordination Sensing

This capstone project with The Ohio State University ECE department explored how to measure and interpret group coordination in shared spaces without relying on cameras. The goal was a privacy-preserving approach that could still capture meaningful motion patterns and support real-time, on-device processing.

‍

The first semester focused on problem framing and system design rather than full model performance. The team aligned on what “group coordination” should mean at a high level, what outputs would be useful, and what constraints mattered most for a deployable solution.

‍

On the hardware side, the concept centered on an mmWave radar sensor that produces anonymous motion signals (3D point clouds and velocity) instead of identifiable imagery. A small edge computer (Raspberry Pi) served as the local processing hub, handling ingest, filtering, and preparation of features for downstream inference.

‍

On the software side, the project outlined an end-to-end pipeline: edge anonymization and preprocessing, secure handling and storage of processed data, and a simple interface to review outcomes. The modeling plan separated the problem into (1) classifying overall group interaction states (for example, coordinated vs. neutral vs. disrupted) and (2) recognizing lower-level collaborative movements that help explain those states.

By the end of the fall, the team had narrowed the scope, selected the core sensing and compute components, and validated early data flow and preprocessing needs. In the spring, the work will focus on completing the full hardware-to-dashboard loop, collecting and labeling data in realistic scenarios, training and validating the models, and hardening the system to meet requirements like accuracy, security, and scaling to multiple groups.