Performance trade offs in IoT-based traffic monitoring and incident detection systems.

Traffic monitoring systems are fundamental for making roads safer and more efficient. Newly proposed traffic monitoring systems enable the use of IoT devices to monitor traffic in a cost-effective way even over large distances on rural highways. These systems rely on machine learning algorithms which are too computationally expensive to run on the devices themselves, and so need to be executed in the cloud or over local processing nodes. In this work, we simulate an IoT-based traffic incident detection system deployed over a side road, comparing two different architectures: one with ad hoc communication and cloud processing, and the other based on edge computing. Results show that the networks based on edge computing have higher packet delivery rates in all scenarios and up to 30% faster round-trip time compared to the cloud-based approach.