Optimally-Self-Healing Distributed Gradient Structures Through Bounded Information Speed

With the constant increase in the number of interconnected devices in today networks, more and more computations can be described by spatial computing abstractions. In this context, distances can be estimated in a fully-distributed way by the so-called gradient self-organisation pattern: it is a basic building block also for large-scale system coordination, frequently used to broadcast information, forecast pointwise events, as carrier for distributed sensing, and as combinator for higher-level spatial structures. However, computing gradients is very problematic in a mutable environment: existing algorithms fail in reaching adequate trade offs between accuracy and reaction speed to environment changes. In this paper we introduce a new gradient algorithm, BIS (Bounded Information Speed) gradient, which uses time information to achieve a smooth and predictable reaction speed, which is proved optimal for algorithms following a single-path-communication strategy. Following a proposed methodology for empirical evaluation of performance of spatial computing algorithms, we evaluate BIS gradient and compare it with other approaches. We show that BIS achieves the best accuracy while keeping smoothness under control