Distributed Computing for Intelligent Buildings in Communities: The Caravels Approach

The transition toward decentralized energy systems requires intelligent, resilient, and user-centered infrastructures capable of supporting both autonomy and cooperation across buildings. This paper presents Caravels, which is a distributed system architecture designed to enable intelligent communities through container-based orchestration, focusing in two new features: personalized user modeling, and inter-building service and data sharing. Building upon a modular microservice deployment paradigm, Caravels integrates a graph-based user preference module for context-aware personalization and a data sharing mechanism that facilitates peer-to-peer cooperation. This paper presents a three-part case study evaluating the system across heterogeneous environments deployed in a community setting. Results demonstrate real-time adaptability through dynamic service deployment based on user-defined preferences, effective inter-building collaboration via shared IoT access and service consumption, and personalized automation through reinforcement learning embedded in graphs. The system maintains low-latency operation and minimal resource usage, validating its applicability to low-cost, edge-constrained environments. Caravels innovations enable distributed cooperation on edge, sharing data while preserving sovereignty, and context aware personalization at the building-level. • Organic intelligent framework for communities. • Distributed intelligent framework with central orchestration for dynamic operation. • A novel distributed container-based orchestration infrastructure for communities. • User- and Human-centric solution to enable the deployment of services in communities.