A Framework for Distributed Interaction in Intelligent Environments.

Ubiquitous computing is extending its applications to an increasing number of domains. "Monolithic" approaches use centralised systems, controlling devices and users' requests. A different solution can be found in works proposing "distributed" intelligent devices that communicate, without a central reasoner, creating little communities to support the user. If the former approach uses all the available sensors being more easily context-aware, the latter is scalable and naturally supports multiple users. In this work we introduce a model for a distributed network of entities in Intelligent Environments. Each node satisfies users' requests through Natural User Interfaces. If a node cannot produce the expected output, it communicates with others in the network, generating paths where the final target is undetermined and intermediate nodes do not understand the request; this is the focus of our work. The system learns parameters and connections in the initial topology. We tested the system in two scenarios. Our approach finds paths close to the optimum with reasonable connections.