Enabling swarm aggregation of position data via adaptive stigmergy: A case study in urban traffic flows

Urban road congestion estimation is a challenge in traffic management. City traffic state can vary temporally and spatially between road links, depending on crossroads and lanes. In addition, congestion estimation requires some sort of tuning to "what is around" to trigger appropriate reactions. An adaptive aggregation mechanism of position data is therefore crucial for traffic control. We present a biologically-inspired technique to aggregate position samples coming from on-vehicle devices. In essence, each vehicle position sample is spatially and temporally augmented with digital mark, locally deposited and evaporated. As a consequence, a marks concentration appears and stays spontaneously while many stationary vehicles and high density roads occur. Marks concentration is then sharpened to achieve a better distinction of critical phenomena to be triggered as detected traffic events. The overall mechanism can be actually enabled if structural parameters are correctly tuned for the given application context. Determining such correct parameters is not a simple task since different urban areas have different traffic flux and density. Thus, an appropriate tuning to adapt parameters to the specific urban area is desirable to make the estimation effective. In this paper, we show how this objective can be achieved by using differential evolution.