Effective Collective Summarisation of Distributed Data in Mobile Multi-Agent Systems

One of the key applications of physically-deployed multi-agent systems, such as mobile robots, drones, or personal agents in human mobility scenarios, is to promote a pervasive notion of distributed sensing achieved by strict agent cooperation. A quintessential operation of distributed sensing is data summarisation over a region of space, which finds many applications in variations of counting problems: counting items, measuring space, averaging environmental values, and so on. A typical strategy to perform peer-to-peer data summarisation with local interactions is to progressively accumulate information towards one or more collector agents, though this typically exhibits several sources of fragility, especially in scenarios featuring high mobility.

In this paper, we introduce a new multi-agent algorithm for dynamic summarisation of distributed data, called "parametric weighted multi-path", based on a local strategy to break, send, and then recombine sensed data across neighbours based on their estimated distance, ultimately resulting in the formation of multiple, dynamic and emergent paths of information flow towards collectors. By empirical evaluation via simulation in synthetic and realistic case studies, accounting for various sources of volatility, using different state-of-the-art distance estimations, and comparing to other existing implementations of aggregation algorithms, we show that parametric weighted multi-path is able to retain adequate accuracy even in high-variability scenarios where all other algorithms are significantly diverging from correct estimations.