VIViorela Ila

Viorela Ila

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Viorela Ila has a total of 64 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Online Dynamic SLAM With Incremental Smoothing and Mapping

    by Jesse Morris, Yiduo Wang, Viorela Ila

    Dynamic SLAM methods jointly estimate for the static and dynamic scene components. However, existing approaches, while accurate, are computationally expensive and unsuitable for online applications. In this work, we present a novel factor-graph formulation and system architecture for Dynamic SLAM that inherently supports incremental optimisation and online estimation. This represents the first formulation explicitly designed to leverage incremental inference methods in the dynamic setting. On multiple datasets, we demonstrate that our method achieves camera pose and object motion accuracy equal to or better than state-of-the-art. We further analyse the structural properties of our approach to demonstrate its scalability and provide insight regarding the challenges of solving Dynamic SLAM incrementally. Finally, we show that our formulation leads to problem structure well-suited to incremental solvers, and our system architecture further enhances performance, achieving a $\mathbf {5\times }$ speed-up over existing methods.

  • DynoSAM: Open-Source Smoothing and Mapping Framework for Dynamic SLAM

    by Jesse Morris, Yiduo Wang, Mikolaj Kliniewski, Viorela Ila

    Traditional Visual Simultaneous Localization and Mapping systems focus solely on static scene structures, overlooking dynamic elements in the environment. Although effective for accurate visual odometry in complex scenarios, these methods discard crucial information about moving objects. By incorporating this information into a Dynamic SLAM framework, the motion of dynamic entities can be estimated, enhancing navigation whilst ensuring accurate localization. However, the fundamental formulation of Dynamic SLAM remains an open challenge, with no consensus on the optimal approach for accurate motion estimation within a SLAM pipeline. Therefore, we developed DynoSAM, an open-source framework for Dynamic Objects SLAM that enables the efficient implementation, testing, and comparison of various Dynamic SLAM optimization formulations. We further propose a novel formulation that encodes rigid-body motion model in object pose estimation as well as an error metric agnostic to object frame definition. DynoSAM integrates static and dynamic measurements into a unified optimization problem solved using factor graphs, simultaneously estimating camera poses, static scene, object motion or poses, and object structures. We evaluate DynoSAM across diverse simulated and real-world datasets, achieving state-of-the-art motion estimation in indoor and outdoor environments, with substantial improvements over existing systems. Additionally, we demonstrate DynoSAM's contributions to downstream applications, including 3D reconstruction of dynamic scenes and trajectory prediction, thereby showcasing potential for advancing dynamic object-aware SLAM systems. Code is open-sourced at https://github.com/ACFR-RPG/DynOSAM


  • Bias-Eliminated PnP for Stereo Visual Odometry: Provably Consistent and Large-Scale Localization

    by Guangyang Zeng, Yuan Shen, Ziyang Hong, Yuze Hong, Viorela Ila, Guodong Shi, Junfeng Wu

    In this letter, we first present a bias-eliminated weighted (Bias-Eli-W) perspective-n-point (PnP) estimator for stereo visual odometry (VO) with provable consistency. Specifically, we develop a $\sqrt{n}$-consistent PnP estimator that accounts for 3D point uncertainties, ensuring that the relative pose estimate converges to the true value as the feature number increases. Next, on the stereo VO pipeline side, we propose a framework that triangulates current features for tracking new frames, decoupling temporal dependencies between pose and 3D point errors. We integrate the Bias-Eli-W PnP estimator into the proposed pipeline, creating a synergistic effect that enhances the accuracy of pose estimation. We validate the performance of our method on the KITTI and Oxford RobotCar datasets. Experimental results demonstrate that our method: 1) achieves significant improvements in both relative pose error and absolute trajectory error in large-scale scenarios; 2) provides reliable localization under erratic motions. The successful implementation of the Bias-Eli-W PnP in stereo VO indicates the importance of information screening in robotic estimation tasks, shedding light on diverse applications where PnP is a key ingredient. Our code is open-sourced at https://github.com/LIAS-CUHKSZ/CurrentFeature-Odometry.


  • PEBO-SLAM: Observer Design for Visual Inertial SLAM With Convergence Guarantees

    by Bowen Yi, Chi Jin, Lei Wang, Guodong Shi, Viorela Ila, Ian R. Manchester

    This paper introduces a new parameterization for the problem of visual inertial simultaneous localization and mapping (VI-SLAM), i.e. for a robot only equipped with a single monocular camera and an inertial measurement unit (IMU). In this problem, the system state evolves on the nonlinear manifold $\mathit{SE}(3)\times \mathbb {R}^{3\mathit{n}}$ , on which we design dynamic extensions such that the deterministic VI-SLAM problem can be reformulated—without any approximation—into online constant parameter identification, expressed as a linear regression. This demonstrates that deterministic VI-SLAM can be translated into a linear least squares problem globally and exactly . Based on this observation, we propose a novel SLAM observer, following the recently established parameter estimation-based observer (PEBO) methodology. A notable merit of the proposed observer is its almost global asymptotic stability. Unlike most existing methods, its convergence does not rely on persistency of excitation or uniform complete observability—assumptions commonly used in stability proofs that can be challenging to satisfy in real-world applications.


  • A Geometric Observer for Scene Reconstruction Using Plenoptic Cameras

    by Sean G. P. O'Brien, Jochen Trumpf, Viorela Ila, Robert Mahony

    This paper proposes an observer for generating depth maps of a scene from a sequence of measurements acquired by a two-plane light-field (plenoptic) camera. The observer is based on a gradient-descent methodology. The use of motion allows for estimation of depth maps where the scene contains insufficient texture for static estimation methods to work. A rigourous analysis of stability of the observer error is provided, and the observer is tested in simulation, demonstrating convergence behaviour.

  • Calibrating Light-Field Cameras Using Plenoptic Disc Features

    by Sean O'BRien, Jochen Trumpf, Viorela Ila, Robert Mahony

    This paper proposes a new method for estimating calibration parameters of plenoptic cameras by minimizing the nonlinear plenoptic reprojection error. Novel plenoptic feature types are proposed as data for the calibration method. These plenoptic disc features are in a natural one-to-one correspondence with physical points in front of the camera. We exploit the intrinsic geometry of plenoptic cameras in a novel projection model that relates the plenoptic disc features to physical points. The resulting calibration quality, as quantified by mean reprojection error and 3D reconstruction error, outperforms recently published results.


  • Robust Visual Odometry in Underwater Environment

    by Jun Zhang, Viorela Ila, Laurent Kneip

    The accurate estimation of pose and velocity of an autonomous underwater vehicle (AUV) is critical to ensure the repeatability and validity of scientific data that is captured using sensors onboard the AUV. A low-cost and effective way is by using stereo camera sensors to perform visual odometry (VO). However, this is a difficult problem in underwater due to poor imaging condition and inconsistent motion caused by water flow. This paper proposes a robust and effective stereo underwater VO system that can overcome aforementioned difficulties and accurately localize the AUV. Experimental results demonstrate that the proposed pipeline outperforms existing VO systems in underwater environment, as well as obtains a comparative performance on the KITTI benchmark dataset.


  • Fast Incremental Bundle Adjustment with Covariance Recovery

    by Viorela Ila, Lukas Polok, Marek Solony, Klemen Istenic

    Efficient algorithms exist to obtain a sparse 3D representation of the environment. Bundle adjustment (BA) and structure from motion (SFM) are techniques used to estimate both the camera poses and the set of sparse points in the environment. Many applications require such reconstruction to be performed online, while acquiring the data, and produce an updated result every step. Furthermore, using active feedback about the quality of the reconstruction can help selecting the best views to increase the accuracy as well as to maintain a reasonable size of the collected data. This paper provides novel and efficient solutions to solving the associated NLS incrementally, and to compute not only the optimal solution, but also the associated uncertainty. The proposed technique highly increases the efficiency of the incremental BA solver for long camera trajectory applications, and provides extremely fast covariance recovery.


  • Exploring the effect of meta-structural information on the global consistency of SLAM

    by Mina Henein, Montiel Abello, Viorela Ila, Robert Mahony

    Accurate online estimation of the environment structure simultaneously with the robot pose is a key capability for autonomous robotic vehicles. Classical simultaneous localization and mapping (SLAM) algorithms make no assumptions about the configuration of the points in the environment, however, real world scenes have significant structure (ground planes, buildings, walls, ceilings, etc.) that can be exploited. In this paper, we introduce meta-structural information associated with geometric primitives into the estimation problem and analyze their effect on the global structural consistency of the resulting map. Although we only consider the effect of adding planar and orthogonality information for the estimation of 3D points in a Manhattan-like world, this framework can be extended to any type of geometric, kinematic, dynamic or even semantic information. We evaluate our approach on a city-like simulated environment. We highlight the advantages of the proposed solution over SLAM formulation considering no prior knowledge about the configuration of 3D points in the environment.


  • Mission-time 3D reconstruction with quality estimation

    by Klemen Istenic, Viorela Ila, Lukas Polok, Nuno Gracias, Rafael Garcia

    Accurate and detailed 3-dimensional (3D) models of the underwater environment are becoming increasingly important in modern marine surveys, since they convey immense information that can be easily interpreted. Techniques such as bundle adjustment (BA) and structure from motion (SfM), which jointly estimate sparse 3D points of the scene and camera poses, have gained popularity in underwater mapping applications. However, for large-area surveys these methods are computationally expensive and not intended for online application. This paper proposes an SfM pipeline based on solving the BA problem in an incremental and efficient way. Furthermore, the new system can provide not only the solution of the optimization (camera trajectory along time and the 3D points of the environment), but also the estimate of the uncertainty associated with the 3D reconstruction. This system is able to produce results in mission-time, i.e. while the robot is in the water or very shortly afterwards. Such quick availability is of great importance during survey operations as it allows data quality assessment in-situ, and eventual replanning of missions in case of need.