Opportunistic Sensing in Train Safety Systems

Train safety systems are complex and expensive, and changing them requires huge investments. Changes are evolutionary and small. Current developments, like faster high speed trains and a higher train density on the railway network, have initiated research on safety systems that can cope with the new requirements. This paper presents a novel approach for a safety subsystem that checks the composition of a train, based on opportunistic sensing with a wireless sensor network. Opportunistic sensing systems consist of changing constellations sensors that, for a limited amount of time, work together to achieve a common goal. Such constellations are selforganizing and come into being spontaneously. The proposed opportunistic sensing system selects a subset of sensor nodes from a larger set based on a common context. We show that it is possible to use a wireless sensor network to make a distinction between carriages from different trains. The common context is acceleration, which is used to select the subset of carriages that belong to the same train out of all the carriages from several trains in close proximity. Simulations based on a realistic set of sensor data show that the method is valid, but that the algorithm is too complex for implementation on simple wireless sensor nodes. Downscaling the algorithm reduces the number of processor execution cycles as well as memory usage, and makes it suitable for implementation on a wireless sensor node with acceptable loss of precision. Actual implementation on wireless sensor nodes confirms the results obtained with the simulations. Keywords-opportunistic sensing; wireless sensor network; context awareness; activity recognition; train safety.