Sensor-Driven RSSI Prediction via Adaptive Machine Learning and Environmental Sensing.

Area of Science:

• Wireless Communications
• Machine Learning
• Environmental Sensing

Background:

• Received Signal Strength Indicator (RSSI) prediction is vital for network planning, optimization, and efficient handover management.
• Traditional path loss models offer a theoretical basis but struggle with complex environmental interactions.
• Machine learning presents a data-driven approach to enhance wireless network performance.

Purpose of the Study:

• To propose an adaptive machine learning framework for RSSI prediction.
• To integrate environmental sensing parameters (temperature, humidity, pressure, particulate matter) into RSSI prediction.
• To evaluate the impact of environmental factors on RSSI prediction accuracy.

Main Methods:

• Developed an adaptive machine learning framework for RSSI prediction.
• Incorporated environmental sensing data (temperature, relative humidity, barometric pressure, particulate matter).
• Compared the proposed model against baseline models using performance metrics like variance explained and root mean squared error.

Main Results:

• RSSI values are influenced by environmental factors through complex, non-linear interactions.
• The proposed model showed a 6.02% increase in variance explained compared to a baseline model without environmental data.
• A 2.04% increase in variance explained and a reduction in root mean squared error to 1.40 dB were observed when including environmental parameters.

Conclusions:

• Environmental factors significantly impact RSSI prediction, challenging linear assumptions of traditional models.
• Cognitive methods integrating environmental data substantially enhance RSSI prediction accuracy.
• The proposed framework offers improved accuracy for wireless network planning and optimization.