Detecting urban road congestion via parametric adaptation of position-based stigmergy

Urban traffic management requires congestion detection. Traffic shape changes over time and location in which it is observed. Moreover it depends on roads, lines and crossroads arrangement. In addition, each congestion event has its own peculiarities (e.g. duration, extension, flow). Therefore, to give correct responses any detection model needs some kind of parametric adjustment. In this paper, we present an adaptive biologically-inspired technique for swarm aggregation of on-vehicle GPS devices positions, able to detect traffic congestion. The aggregation principle of the position samples is based on a digital mark, released at each sample in a digital space mapping the physical one, and evaporated over time. Consequently, marks aggregation occurs and stays spontaneously while many stationary vehicles are crowded into a road. In order to identify actually relevant traffic events, marks aggregation has to be correctly configured. This is achieved by tuning the mark’s structural parameters. Considering that each urban area has a specific traffic flow and density, determining a proper set of parameters is not trivial. Here, we approach the issue using different differential evolution variants, showing their impact on performance.