Practical Challenges And Pitfalls Of Bluetooth Mesh Data Collection Experiments With Esp-32 Microcontrollers

Abstract: Testing network algorithms in physical environments using real hardware is an important step to reduce the gap between theory and practice in the field, and an interesting way to explore technologies such as Bluetooth Mesh. We implemented a Bluetooth Mesh data collection strategy and deployed it in indoor and outdoor settings, using ESP-32 microcontrollers. This data collection strategy also covers an alternative packet routing strategy based on Bluetooth Mesh - MAM - already discussed and simulated in previous work using the OMNET++ simulator. We compared the real-world ESP-32 experiments with the past simulations, and the results differed significantly: the simulations predicted a +459\% unique message collection compared to the results we obtained with the ESP-32. Based on those results, we also identified vast room for improvement in our ESP-32 implementation for future work, including solving an unexpected packet duplication in the MAM algorithm implementation. Even so, MAM performed better than Bluetooth Mesh's default relay strategy, with up to +4.06\% more (unique) data messages collected. We also discuss some challenges we experienced when implementing, deploying, and running benchmarks using Bluetooth Mesh and the ESP-32 platform.