Though autonomic computing has received a lot of attention since its inception in the early 2000, to date the research has mainly considered relatively small-scale systems. But Kephart recently reminded that the main challenges in this area revolve around facing the inherent decentralisation when going large-scale, which greatly complicates often interdependent adaptation decisions. Especially as autonomic computing imposes inevitable time constraints as it de facto implies a reactive control for adapting. In this paper, we propose a decentralised approach for coordinating the decision-making that uncouples the exchange and gathering of coordination information on one hand from the coordinated local decision-making on the other hand. This uncoupling allows for these local decisions to be made without delays as they are based upon information gathered previously. This approach is illustrated by a geotracking application for large fleets of trucks from the SALTY ANR project, and validated through a series of simulations demonstrating the adequacy and the good performances of its various algorithms.