Week 2: Research Ethics, System Architecture, and Hardware Validation

Objectives for the Week

• Complete and submit formal research ethics screening and approval documentation.
• Define the multi-tiered system architecture (Sensor, Edge AI, and Application layers).
• Establish a formal Work Breakdown Structure (WBS) and Gantt chart for project management.
• Configure the ESP32 development environment and empirically validate the external high-gain antennas.

The Ethical Framework of A.E.G.I.S.

Unlike traditional ambient monitoring projects that rely on intrusive optical cameras or physical wearables, A.E.G.I.S. (Autonomous Elderly Guardian & Intelligent Sensing) operates entirely on invisible radio frequencies. Specifically, it utilizes Wi-Fi Channel State Information (CSI) and Frequency Modulated Continuous Wave (FMCW) radar.

Completing the university's ethics screening forms was a necessary step to confirm the project's adherence to professional engineering standards:

• Data Privacy: No video, audio, or personally identifiable imagery is recorded, transmitted, or stored at any point.
• Radiological Safety: All hardware utilizes low-power, commercially safe radio waves (2.4GHz Wi-Fi and 24GHz radar), remaining well within standard consumer safety limits and posing zero radiation risk.
• Participant Safety: The research poses no physical harm to vulnerable demographics. I will act as the sole test subject for all simulated fall experiments to ensure no elderly participants are exposed to physical risk during the data collection phase.
• Edge-Computing Paradigm: By processing all signal data locally on the Raspberry Pi 5 edge device rather than offloading it to a cloud server, the system inherently mitigates the risk of external data breaches.

Breaking Down the Technical Architecture

While A.E.G.I.S. serves a healthcare application, its underlying architecture is a complex, layered Internet of Things (IoT) network. It requires the seamless integration of embedded systems engineering, edge-based artificial intelligence, and real-time network communication.

1. Sensor Layer: RF Data Extraction

The foundation of the A.E.G.I.S. framework relies on capturing physical disturbances in the ambient environment using two distinct sensor modalities:

• Wi-Fi CSI Nodes: ESP32-WROOM-32U microcontrollers equipped with external 6dBi high-gain antennas via U.FL/IPEX connectors. These nodes track the minute phase and amplitude variances in Wi-Fi subcarriers caused by human movement intersecting the line of sight.
• mmWave Radar: The LD2410C sensor, which is dedicated to detecting micro-movements. It will monitor the subtle chest displacements caused by human respiration to verify post-fall consciousness.

2. Edge AI Layer: Fall Classification and Processing

To maintain privacy and reduce latency, a Raspberry Pi 5 serves as the central processing brain of the local network:

• Signal Processing: Filtering environmental noise from the raw CSI and radar data streams using techniques such as Butterworth filters and Principal Component Analysis (PCA).
• Machine Learning Classification: Utilizing Support Vector Machines (SVM) or lightweight Decision Trees to classify distinct spatial movement signatures, differentiating a critical "hard fall" from standard activities like sitting or walking.
• Sensor Fusion Logic: A rule-based system that cross-references the fall detection trigger from the CSI network with the vital sign data from the radar.

3. Application Layer: Decision Support and Alerting

• Communication Protocol: Implementing an MQTT-based messaging architecture to ensure ultra-low latency transmission of alert states.
• Caregiver Interface: A secure dashboard that displays real-time room status (e.g., "Occupied - Normal" or "Critical Alert") without exposing visual data, preserving the dignity of the monitored individual.

Project Management: Structuring the Development Cycle

To manage the extensive workload systematically, I structured the project into core developmental phases: Initiation, Hardware Assembly, Data Collection, Edge AI Development, System Integration, and Evaluation.

These tasks have been mapped onto a comprehensive Gantt Chart spanning the semester. This visual timeline ensures that critical path milestones, such as baseline data collection and iterative model training, are achieved on schedule, allowing sufficient time for final system calibration.

Preliminary Hardware Validation: The Signal Attenuation Experiment

Before writing the complex C++ firmware required to extract raw CSI payloads, it was critical to empirically validate the hardware connections. Because the ESP32-WROOM-32U lacks an onboard antenna, a poorly seated IPEX connector would result in high packet loss and corrupt data.

To test this, I configured one ESP32 to act as a localized "Beacon" broadcasting a test SSID, and a second ESP32 as a "Watcher" programmed to continually log the Received Signal Strength Indicator (RSSI) via the Arduino IDE's Serial Plotter.

By establishing a baseline signal and then physically walking directly between the transmission path of the two antennas, I successfully observed a significant, repeatable attenuation (drop) in signal strength. This "shadowing" effect scientifically validates that the external antennas are correctly seated and highly sensitive to human presence. This successful test confirms the fundamental physical principles upon which the entire A.E.G.I.S. detection algorithm will be built.

Next Steps

• Transition from basic RSSI tracking to implementing the ESP-IDF framework for extracting the granular 52-subcarrier CSI payload.
• Wire the LD2410C mmWave radar sensor and configure its serial communication protocols.
• Establish the data logging pipeline on the Raspberry Pi 5 to begin capturing baseline datasets.