A unified framework for joint mobility prediction and object profiling of drones in UAV networks

Han, Peng; Razi, Abolfazl; Afghah, Fatemeh; Ashdown, Jonathan · 2018 · OpenAlex-citations

DOI: 10.1109/jcn.2018.000068

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of managing dynamic network topologies in autonomous Unmanned Aerial Vehicle (UAV) networks, particularly in scenarios like search-and-rescue where ground station command is limited. The authors identify that efficient control and communication in such networks require proactive decisions based on the predicted mobility and capabilities of neighboring nodes. To solve this, they propose a unified framework called Joint Mobility Prediction and Profiling (JMPP), which simultaneously predicts the future locations of surrounding flying objects and classifies them into distinct groups based on their maneuverability profiles (e.g., rotary vs. fixed-wing) without prior knowledge of these classes. The method employs an unsupervised online learning algorithm grounded in Kalman filtering with intermittent observation. The framework operates in three main stages. First, it uses Kalman filtering to estimate the state vector (location and velocity) and denoise observations, handling unknown input forces. Second, it extracts mobility parameters by modeling acceleration and angular velocity as Gaussian mixture models, using the Expectation-Maximization (EM) algorithm to estimate object-specific motion parameters. Third, it performs object profiling by applying Bayesian inference to classify objects into groups sharing similar hyper-parameters. Crucially, the system includes an online self-tuning module that uses parametric clustering (K-means) to dynamically adjust the number of mobility classes and refine hyper-parameters as new data arrives, allowing the network to adapt to heterogeneous and emerging node types. Simulation results demonstrate the effectiveness of the JMPP framework. The Kalman filtering stage achieved a mean squared error ratio of less than 1% for trajectory estimation. In terms of object profiling, the method achieved an average classification success rate of approximately 91% across three distinct mobility classes. This performance significantly outperformed baseline methods, including Gaussian Process classification (80.33% success rate), fuzzy c-means, and K-means applied directly to driving force data. The study also highlighted that prediction accuracy is sensitive to the measurement update rate, declining significantly when observation opportunities are sparse. Furthermore, the online module successfully recognized new object types and refined class parameters over time, validating its flexibility in open environments. The significance of this work lies in its ability to facilitate autonomous decision-making in Flying Ad-hoc Networks (FANETs) by providing both short-term trajectory predictions and long-term mobility profiles. By enabling UAVs to identify neighbor types and predict their movements without centralized coordination or prior labeling, the framework supports proactive routing and efficient task allocation. This approach enhances network connectivity and operational performance in time-sensitive applications, offering a scalable solution for heterogeneous UAV networks where node capabilities and behaviors may vary widely.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction