Wireless multipath routing for IoT multimedia: a comprehensive testbed analysis

Abstract Multipath routing is a key strategy for meeting the Quality of Service (QoS) requirements of video applications in the Internet of Things (IoT). However, the objective functions used in heuristic-based multipath selection mechanisms are often defined through mathematical models or network performance estimators and typically validated only in simulators, which may fail to capture the full complexity of real-world wireless environments. This paper presents the design and validation of a physical testbed composed of low-cost, single-board computers for wireless multipath video transmission. Beyond evaluating hardware resource usage (CPU, memory, and thermal load), we conduct a comprehensive cross-layer case study to assess the predictive reliability of the FITPATH and QSOpt routing heuristics. Transitioning from traditional QoS metrics to perception-based Quality of Experience (QoE), we evaluate the Structural Similarity Index (SSIM) across video sequences with varying degrees of motion and network load. Our findings reveal that while simulators capture general trends, high-motion video significantly degrades latency, suggesting the occurrence of intermediate queue bottlenecks. Furthermore, our video quality gap analysis demonstrates that while FITPATH provides highly accurate relative path rankings, both heuristics struggle to predict absolute performance drops caused by real-world interference. Ultimately, this work underscores the necessity of physical testbed validation for interference-aware objective functions in IoT multimedia networks.