Institution profile

Politecnico di Milano

Academic institutioneurope · it
Official website
Research library823linked papers
Opportunities0open roles
Selected work

Representative Papers

AIM: Acoustic Inertial Measurement for Indoor Drone Localization and Tracking

Nov 06, 2022ACM International Conference on Embedded Networked Sensor Systems

To address the challenge of precise UAV localization in GPS-denied indoor environments—particularly under non-line-of-sight (NLoS) conditions where existing methods rely on dedicated hardware or pre-deployed infrastructure—this paper proposes a markerless, infrastructure-free acoustic-inertial fusion localization paradigm. Leveraging only the intrinsic acoustic features of an onboard sound source and a distributed microphone array, the method enables real-time, high-accuracy motion tracking even in NLoS scenarios. We introduce an IQR-enhanced extended Kalman filter to robustly suppress outliers in acoustic source localization. Experimental results demonstrate that, in complex indoor settings, our approach achieves an average positioning error 46% lower than that of commercial UWB systems. Moreover, it consistently maintains sub-meter accuracy (<0.8 m) across arbitrarily scaled and configured spaces, significantly improving NLoS robustness and deployment flexibility.

16 citationsRead paper

MultiLink: Multi-class Structure Recovery via Agglomerative Clustering and Model Selection

Jun 01, 2021Computer Vision and Pattern Recognition

This paper addresses the robust joint recovery of multiple geometric structures (e.g., planes, cylinders, homography/fundamental matrices) from noisy data contaminated with outliers. We propose an online model fitting and selection-driven agglomerative clustering framework. Our method innovatively integrates dynamic linkage criteria to enable end-to-end co-optimization of model fitting and selection. It combines online RANSAC-style fitting, information-theoretic adaptive model selection, and multi-structure consistency metrics—thereby overcoming key limitations of conventional approaches, including sensitivity to inlier thresholds and severe sampling bias. Extensive evaluations on multiple public benchmarks demonstrate significant improvements over state-of-the-art methods, achieving high accuracy, strong robustness to outliers and noise, fast runtime, and insensitivity to threshold tuning. The source code is publicly available.

14 citations1 influentialRead paper

GOLIATH: A Decentralized Framework for Data Collection in Intelligent Transportation Systems

Aug 01, 2022IEEE transactions on intelligent transportation systems (Print)

To address single-point failures and trust deficiencies inherent in centralized data exchange within Intelligent Transportation Systems (ITS), this paper proposes a decentralized traffic data collection framework deployed on In-Vehicle Infotainment (IVI) systems. The framework integrates blockchain technology, vehicle-to-vehicle (V2V) peer-to-peer communication, and a lightweight, customized Byzantine Fault Tolerance (BFT) consensus mechanism—specifically designed for resource-constrained vehicular networks. It achieves, for the first time, efficient, tamper-resistant collaborative verification in such environments. Evaluated in a high-fidelity simulation, the system processes location-based transactions on-chain within milliseconds and detects malicious behavior with 99.2% accuracy. These results demonstrate substantial improvements in distributed fault tolerance and security, effectively overcoming the trust bottlenecks of conventional crowdsourced ITS architectures.

14 citationsRead paper

Indoor Drone Localization and Tracking Based on Acoustic Inertial Measurement

Jun 01, 2024IEEE Transactions on Mobile Computing

To address the challenge of precise localization and tracking of indoor drones under GPS-denied and non-line-of-sight (NLoS) conditions, this paper proposes a hardware-agnostic acoustic-inertial fusion method that requires no drone hardware modification or large-scale infrastructure deployment. Innovatively leveraging the drone’s rotor-generated acoustic signatures as motion sources, the approach integrates a distributed microphone array with a customized extended Kalman filter (EKF) and incorporates an interquartile range (IQR)-based robust outlier rejection mechanism. This enables real-time, three-dimensional pose estimation and motion tracking in arbitrarily sized and configured indoor environments. Experimental evaluation in a complex 10 m × 10 m indoor setting demonstrates that the method achieves a 46% reduction in mean localization error compared to a commercial ultra-wideband (UWB) system; moreover, positioning accuracy remains within 0.5 m across a 20 m operational range.

10 citationsRead paper

Keyword Queries over the Deep Web

Nov 14, 2016International Conference on Conceptual Modeling

To address the challenge of keyword search over structured deep web data—particularly restricted tables—that are inherently inaccessible to conventional keyword-based retrieval, this paper proposes a keyword query modeling framework tailored for the deep web. The method comprises three key components: (1) schema-agnostic virtual document generation, which maps invisible database contents into indexable textual representations; (2) cross-table semantic matching integrated with query rewriting to enhance semantic alignment between keywords and underlying data; and (3) joint optimization of result ranking via table structure inference, query expansion, and learning-to-rank techniques. Experiments on real-world deep web datasets demonstrate substantial improvements: NDCG@10 increases by 32% on average over baseline methods, with concurrent gains in both recall and precision. This work establishes the first end-to-end, systematic modeling paradigm for deep web keyword search, advancing the discoverability of deep web data.

6 citations1 influentialRead paper
Recent publications

Latest Papers