IoT Sensing for Reality-Enhanced Serious Games, a Fuel-Efficient Drive Use Case

Area of Science:

• Computer Science
• Machine Learning
• Internet of Things

Background:

• Internet of Things (IoT) technologies enable new instructional games called reality-enhanced serious games (RESGs) for field-based training.
• RESGs can incorporate real-time user performance evaluation using sensor data.
• Fuel-efficient driving is a key area where RESGs can be applied for performance monitoring.

Purpose of the Study:

• To investigate user performance evaluation in RESGs using real-world data.
• To develop and validate a smart sensing dataflow for real-time driver performance assessment and feedback.
• To promote fuel-efficient driving through a novel RESG application.

Main Methods:

• Proposed a reference model for a smart sensing dataflow combining two parallel, real-time machine learning modules.
• Compared machine learning algorithms (support vector regression, random forest, artificial neural networks) for quantitative performance assessment.
• Implemented an instant recommendation module using fuzzy logic for verbal feedback on driving patterns.
• Utilized the enviroCar public dataset with On-Board Diagnostic II (OBD II) data for system testing.

Main Results:

• Random forest showed slightly better performance assessment correlation (R² = 0.99) but higher inference time compared to other algorithms.
• The fuzzy logic module provided timely verbal recommendations for inefficient driving.
• The system demonstrated robustness across various driving environments and vehicle models.
• Achieved high estimation correlation and punctual verbal feedback, confirming the approach's feasibility and effectiveness.

Conclusions:

• The proposed machine learning-based dataflow is effective for real-time driver performance evaluation in RESGs.
• RESGs can successfully promote fuel-efficient driving through data-driven feedback.
• Privacy concerns regarding the use of sensitive personal driving data must be addressed.