Pre-Conference Talk by SHESHADRI Smitha | Enhancing Smartphone-based Inertial Indoor Tracking with Conversational User Input

IMU-based tracking enables infrastructure-independent indoor tracking using smartphones but suffers from error accumulation, limiting its usability. We present a system that dynamically improves real-time IMU tracking by strategically requesting and integrating user-provided positional inputs. Our approach addresses the key challenge of integrating sparse, asynchronous conversational user inputs with continuous IMU data by incorporating user input into the weight adjustment process of a Particle Filter. Evaluation on two indoor routes with 24 participants, each carrying a smartphone in one of three conditions (pocket, backpack, handheld), show that our system maintains RMSE below 10 meters in 85% of trials, compared to 6% for the inertial-only baseline, reducing RMSE by 61% and maximum error by 50%. Analysis of user interaction variables indicates that higher response rates, better input quality, and shorter search distances significantly improve tracking, while response delays have negligible effects. Additionally, while tracking accuracy varied slightly across smartphone carrying conditions, differences were not statistically significant. Cadence analysis confirms no significant disruption to walking patterns, demonstrating usability during natural movement. These findings highlight the effectiveness of user-informed IMU tracking across different smartphone-carrying conditions and diverse interaction parameters. To evaluate performance beyond controlled settings, we tested the system in two public indoor environments—a shopping mall and a museum. Eight participants performed navigation and exploration tasks using applications built on top of our tracking system. The system consistently supported both tasks and improved tracking performance, achieving a 62.6% reduction in RMSE.

This is a Pre-Conference talk for ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp/ISWC 2026).
 

Smitha SHESHADRI is a PhD candidate in Computer Science at Singapore Management University, specializing in Human–Computer Interaction and indoor spatial intelligence. Her research explores conversational and user-as-sensor approaches to indoor localization and tracking, with the goal of developing lightweight, infrastructure-free positioning systems. Her work has been published at ACM IMWUT/UbiComp and CHI, and investigates how natural language interaction can serve as a spatial signal.