Pervasive presence: passive detection and active guidance of people

Privacy-preserving pervasive computing requires context-awareness mechanisms that operate without dedicated infrastructure, intrusive user interaction, or the collection of personally identifiable information. This dissertation addresses this challenge through a research trajectory combining passive Bluetooth Low Energy (BLE) sensing on commodity smartphones with context-aware mobile sensing, structured interaction, and AI-based reasoning, enabling privacy-conscious presence monitoring, mobility-context recognition, and active spatial assistance. The first contribution investigates Indoor Occupancy Estimation through passive BLE sensing. The proposed framework addresses three fundamental challenges that limit practical deployment: the multi-device phenomenon, cross-room signal propagation, and persistent non-human wireless emitters. A processing pipeline combining temporal aggregation, Received Signal Strength Indicator (RSSI) based filtering, automatic environmental baseline estimation, and context-aware semantic calibration supports occupancy estimation without requiring dedicated sensing infrastructure or active user enrollment. Extensive multi-week deployments across heterogeneous indoor environments demonstrate reductions in estimation error compared with na\"ive counting approaches while preserving user privacy through a zero Media Access Control (MAC) address storage policy. The second contribution extends the same passive sensing paradigm to outdoor Intelligent Transportation Systems (ITS) by exploiting variations in the surrounding wireless environment to infer transportation modes. High-level statistical features extracted from opportunistic BLE observations are processed by an on-device machine learning pipeline to distinguish urban mobility modalities including walking, private-car travel, bus-tram travel, and subway travel. Experimental validation on real-world commuting data demonstrates the feasibility of the proposed approach, while also showing that classification performance varies across transportation modes, with subway travel remaining particularly challenging. The approach maintains infrastructure independence and edge-based processing. The third contribution expands the research from passive context perception to Active Spatial Guidance in Global Positioning System (GPS) denied environments. This work investigates the use of generative AI, memory-grounded reasoning, and structured human-computer interaction to support indoor localization and navigation through technical building layouts. The studies investigate standardized user input derived from evacuation-map semantics alongside vision-capable spatial reasoning, examining the feasibility and limitations of zero-shot visual localization, while memory-grounded mechanisms are explored for familiarity-aware route guidance. These investigations highlight both the potential of AI-supported spatial reasoning and the importance of structured user input when unconstrained descriptions do not provide sufficient information for reliable localization. This research represents a natural extension of the dissertation's overarching objective of transforming passive environmental awareness into intelligent user assistance. Beyond these core contributions, the dissertation presents several complementary collaborative studies that further explore the broader design space of pervasive and ubiquitous computing. Overall, this dissertation demonstrates how passive wireless sensing, privacy-preserving software engineering, and AI-driven interaction techniques can contribute to a coherent research trajectory for context-aware systems operating with limited reliance on specialized infrastructure. The proposed methods contribute practical approaches to pervasive computing, indoor intelligence, and intelligent mobility while preserving user privacy by design.