In-Vehicle and In-Room Mobility and Health Monitoring: A Comparative Simulation Study of Communication Between Edge Devices and Servers
Open Access DepositedThe rapid evolution of Internet of Things (IoT) technologies is transforming intelligent systems, particularly in domains where real-time monitoring and communication are critical—such as transportation safety and healthcare. IoT health monitoring systems promise to revolutionize healthcare delivery by enabling continuous, real-time surveillance of patient vital signs in both mobile (vehicular) and stationary (indoor) environments. However, the reliability of these systems fundamentally depends on communication infrastructure performance, which varies significantly across deployment contexts. Despite the growing adoption of IoT-based health monitoring, major gaps remain in understanding how environmental conditions, device configurations, and connectivity choices influence end-to-end communication performance.This thesis presents a comprehensive simulation-based comparative analysis of IoT health monitoring communication systems across 9,000 vehicular simulation runs (96 scenarios with 100 random seeds each) and 24 indoor scenarios using the OMNeT++ discrete-event simulation framework. We evaluate vehicular mobility scenarios in both urban and rural environments (48 urban and 48 rural), incorporating variations in vehicle speed, cellular tower distance, data rate, message size, and edge device type. We additionally assess 24 indoor smart-room scenarios featuring different camera connection technologies, WiFi configurations, and packet sizes.
Through rigorous statistical analysis, including independent-samples t-tests (t-statistics ranging from 38 to 143 with p-values below 10^-300), ANOVA, multiple linear regression, correlation analysis, coefficient of variation analysis, and 10-fold cross-validation, we quantify performance across 15 metrics such as end-to-end delay, packet loss rate, success rate, throughput, signal strength, and buffer utilization. Our findings reveal five major results with significant practical implications.
First, urban vehicular deployments significantly outperform rural deployments across all metrics, with urban environments achieving 21.41% lower delay (229.2 ms vs. 290.2 ms), 16.64% lower packet loss (41.07% vs. 49.02%), and 21.67% higher success rates (49.14% vs. 40.52%). Effect sizes range from Cohen’s d = 2.786 for delay to d = 9.149 for packet loss, indicating substantial real-world impact. These findings highlight an urban–rural digital health divide with serious implications for healthcare equity.
Second, wired camera connections dramatically outperform Bluetooth Low Energy alternatives in indoor environments, showing orders-of-magnitude differences in delay and throughput.
Third, vehicle speed emerges as the strongest predictor of system performance, with exceptionally strong correlations. Speed has a correlation of 0.970 with packet loss (explaining 96% of variance) and -0.993 with signal strength (nearly perfect negative correlation). Increasing vehicle speed from 25 to 65 mph results in a 34.59% increase in packet loss and a 24.20% decrease in success rate. These results reveal that mobility-induced effects, such as Doppler shift, handover frequency, and rapid channel variation, represent the primary challenge for vehicular health monitoring, contradicting assumptions that modern LTE systems fully compensate for mobility.
Fourth, WiFi band selection and data rate parameters show no statistically significant impact in single-user indoor health monitoring scenarios (p ≈ 1.0). This indicates substantial over-provisioning and opportunities for cost reduction.
Fifth, multi-objective optimization analysis identifies distinct Pareto-optimal configurations for different applications, with performance varying by a factor of 2,442 between the best and worst configurations. This demonstrates that application-specific system design is essential; generic deployment strategies are insufficient.
We develop and validate three multivariate predictive models that achieve high accuracy: packet loss (R² = 0.961), success rate (R² = 0.873), and delay (R² = 0.538). For indoor monitoring, throughput analysis confirms that camera connection type, not WiFi configuration, is the dominant factor influencing performance.
This research provides several novel contributions. We present the first comprehensive comparison of IoT health monitoring systems across urban, rural, and indoor environments using consistent metrics and rigorous statistical methodology. We quantify the performance implications of various connectivity technologies and establish evidence-based thresholds for selecting wired versus wireless solutions. We develop validated predictive models that enable rapid performance forecasting without extensive prototyping. We introduce the first multi-objective Pareto frontier analysis for health IoT, characterizing delay, throughput, reliability, and trade-offs. Finally, we establish a large-scale simulation benchmark (120 scenarios and 15 metrics) that supports future research and reproducibility.
The practical implications are substantial. For vehicular health monitoring, we show that embedded-PC edge devices with wired sensor connections offer an 84.9 ms delay advantage that is critical for real-time arrhythmia detection, while maintaining packet loss performance statistically identical to smartphone-based systems (p = 0.984). This eliminates the perceived latency–reliability trade-off suggested by earlier analyses and positions embedded PCs as the preferred choice for time-critical applications. Rural deployments must either accept a 21.67% lower success rate or invest in improved cellular infrastructure, raising important policy considerations for healthcare equity.
For indoor monitoring, wired cameras are essential for any time-sensitive application. The 1,206-times performance difference observed indicates that wireless cameras introduce delays exceeding 100 seconds, making them unsuitable for fall detection or telemedicine. WiFi over-provisioning means that 40 Mbps is sufficient for typical two-sensor deployments, enabling cost savings through lower-tier equipment. Application-specific optimization frameworks show that real-time monitoring requires wired connections with 4 KB packets (resulting in 42 ms delay), while high-throughput recording benefits from 60 KB packets (achieving 2.78 Mbps throughput).
We acknowledge several limitations. Although simulation-based evaluation allows systematic parameter variation and avoids ethical concerns, real-world validation is needed to verify performance under operational conditions. Our single-user scenarios do not capture multi-user interference common in hospitals or ambulance fleets. Future work should focus on field validation of predictive models, extension to next-generation wireless technologies, and clinical studies linking communication performance to patient outcomes.
This thesis provides healthcare technologists, system designers, and policymakers with empirical evidence and quantitative tools for designing evidence-based IoT health monitoring systems. By characterizing performance across diverse operational contexts, identifying critical bottlenecks, and developing predictive models, we advance the design of reliable and equitable remote patient monitoring systems. The findings clearly demonstrate that one-size-fits-all designs are inadequate; optimal deployment must align with clinical requirements, environmental constraints, and available resources. As healthcare increasingly incorporates IoT-enabled continuous monitoring, the systematic understanding developed in this work becomes foundational for successful and equitable adoption.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
Talukder_gwu_0075M_17645.pdf | 2025-12-15 | Open Access |
|