View Proposal

---

Proposer

Zi Hau Chin

Title

Robust Proximity Zone Classification using Adaptive Filtering and Machine Learning for BLE Devices

Goal

To develop a robust and adaptive system for precise BLE device proximity classification by integrating real-time RSSI smoothing with machine learning.

Description

This project develops a smart sensing system that accurately classifies the proximity of Bluetooth Low Energy (BLE) devices into defined zones. It leverages an Adaptive Kalman Filter to dynamically denoise noisy Received Signal Strength Indicator (RSSI) data and employs machine learning techniques to establish robust proximity zones resilient to environmental variations. The core work involves: 1. Add Adaptive Kalman Filter (AKF): Add an AKF to smooth the noisy raw RSSI in real-time, dynamically adjusting filter parameters. A little bit of modifications to the source code to collect raw RSSI, RSSI with Kalman Filter and RSSI with AKF at the same time. 3. Train ML Classifier (Offline): Collect labelled training data using the smoothed RSSI from your AKF in various environments, then use scikit-learn to train a machine learning model to classify proximity zones. 4. Integrate & Deploy: Integrate the trained and optimised ML model into the source codebase, ensuring efficient inference. Modify the MQTT output to include the predicted proximity zone classification alongside the RSSI values. Conduct comprehensive system testing in real-world scenarios to validate the accuracy of proximity zone classification and the overall system stability. 5. Evaluation: Perform performance benchmarking on raw RSSI, KF and AKF.

Resources

Raspberry Pi 4b

Background

C Programming, Python?, Algorithms, IoT, Web app?

Url

Difficulty Level

Moderate

Ethical Approval

None

Number Of Students

1

Supervisor

Zi Hau Chin

Keywords

bluetooth low energy (ble), rssi, proximity detection, machine learning, iot

Degrees

Bachelor of Science in Computing Science