Cloud-based simulation for services based on crowd-sensing

DFRC AG is a a Swiss SME and a leader in the field of crowd monitoring and smart cities. Its main commercial offering is LBASense, a package which uses a combination of short- and long-range sensors, mapping algorithms, and a data fusion algorithm, to anonymously map the behaviour of crowds on a city-wide scale. There are a wide variety of potential uses for LBASense, such as understanding the behaviour of tourists, planning new transportation routes, or even assisting emergency services in serious incidents.

However, before DFRC can place and operate the sensor network that underpins LBASense, it needs to ‘fingerprint’ the cellular network in the target city. The fingerprinting process is a significant bottleneck in the deployment of the service, as the computational power required to speed up the fingerprinting process, particularly in order to allow DRFC to offer LBASense in large cities, is many times higher than DRFC’s current capabilities. The LBASense mapping algorithm therefore needed to be ported to a cloud-based HPC system to enable the level of computation required. The use of HPC was necessary to make the process sustainable and scalable, as opposed to simply speeding up the process.

With the algorithm and related software successfully ported to the HPC system, a single iteration of the simulation runs in just 500 seconds, an approximately 86% decrease from the original computation time of 3,500 seconds.