Abstract
A recurrent task in coordinated systems is managing (estimating, predicting,or controlling) signals that vary in space, such as distributed sensed data orcomputation outcomes. Especially in large-scale settings, the problem can beaddressed through decentralised and situated computing systems: nodes canlocally sense, process, and act upon signals, and coordinate with neighbours toimplement collective strategies. Accordingly, in this work we devisedistributed coordination strategies for the estimation of a spatial phenomenonthrough collaborative adaptive sampling. Our design is based on the idea ofdynamically partitioning space into regions that compete and grow/shrink toprovide accurate aggregate sampling. Such regions hence define a sort ofvirtualised space that is "fluid", since its structure adapts in response topressure forces exerted by the underlying phenomenon. We provide an adaptivesampling algorithm in the field-based coordination framework, and prove it isself-stabilising and locally optimal. Finally, we verify by simulation that theproposed algorithm effectively carries out a spatially adaptive sampling whilemaintaining a tuneable trade-off between accuracy and efficiency.