Pervasive service ecosystems are a new paradigm for the design of context-aware systems featuring adaptivity and self-awareness. A theoretical and practical framework has been proposed for addressing these scenarios, taking primary inspirations from natural ecosystems and grounding upon two basic abstractions: “live semantic annotations” (LSAs), which are descriptions stored in infrastructure nodes and wrapping data, knowledge, and activities of humans, devices, and services; and “eco-laws”, acting as system rules evolving the population of LSAs as if they were molecules subject to chemical-like reactions. In this paper, we aim at deepening how self-organisation can be injected in pervasive service ecosystems in terms of spatial structures and algorithms for supporting the design of context-aware applications. To this end, we start from an existing classification of self-organisation patterns, and systematically show how they can be supported in pervasive service ecosystems, and be composed to generate a self-organising emergent behaviour. A paradigmatic crowd steering case study is used to demonstrate the effectiveness of our approach.