Session 6A – Robotic Systems, Simulation and Digital Twins
Date: Thursday, 19th November 2026
Time: 16:00 – 17:45
Room: Sala d’Actes
Session Chair: TBA
Presentation format: Each presentation is allocated 15 minutes: approximately 10–12 minutes for the talk and 3–5 minutes for questions.
16:00 – MEDEA: A Mobile Robotic Platform for Georeferenced Radiological Surveying with Dual Radiation Detectors
Authors: Mateo David Alvarez Royero¹, Irene Santalices Acuña¹, Rocío Sánchez Blanco², Eva Navarro Navalón², José María Pereira Pérez², Miguel Ángel Sánchez-Urán González¹ and Manuel Ferre Pérez¹
Affiliations: ¹ Centre for Automation and Robotics (CAR), UPM-CSIC, Universidad Politécnica de Madrid, Madrid, Spain; ² DRACE GEOCISA.
Abstract
Radiological surveying in outdoor and industrial environments requires coordinated area coverage, reliable robot motion, and spatially referenced radiation measurements. This paper presents MEDEA, an integrated mobile robotic system for georeferenced radiological surveying using two radiation detectors mounted on an unmanned ground vehicle. The implemented system combines parallel-lane sweep generation, a mission-level path-tracking controller, GNSS/RTK and odometry-based pose management, and synchronized detector-data acquisition. Survey trajectories are generated from two reference points and executed using a finite-state controller that coordinates lane following, heading correction, transverse transitions, and entry alignment. Radiation measurements are associated with robot localization and motion data and transformed into detector-specific coordinates for georeferenced recording and alarm mapping. The system was evaluated through field execution of structured multi-lane sweeps, demonstrating the operation of the integrated trajectory-generation, control, and data-processing components. Spectral counting windows are considered for radiological scenarios involving Cs-137, Co-60, and K-40; however, isotope identification, activity estimation, and radiometric calibration are outside the scope of this study.
16:15 – Oliwall: Non-Invasive Automated Demolition Building Material Identification
Authors: Lluís Bonet Ortuño¹, Yeray Navarro Soler¹, Zewdie Habtie¹, Carlos Xavier Saldarriaga¹,³, David Caballero Flores¹, Alba Perez Gracia¹,² and Carles Domenech¹
Affiliations: ¹ Centre de Disseny d’Equips Industrials (CDEI), Universitat Politècnica de Catalunya, Barcelona, Spain; ² Institut de Robòtica i Informàtica Industrial (CSIC-UPC), Barcelona, Spain; ³ FIMCP, Escuela Superior Politécnica del Litoral (ESPOL), Guayaquil, Ecuador.
Abstract
The construction sector consumes about half of the materials extracted in the EU and, together with the operation of buildings, accounts for roughly 37% of global greenhouse gas emissions. Recovering materials before demolition requires knowing what a building is made of and where each material is, information rarely documented for old buildings. The DISCOVER project addresses this gap with two robots that collect on-site data for an enriched Building Information Model (BIM). One of them, Oliwall, is an omnidirectional mobile manipulator carrying non-invasive sensors: LiDAR, RGB-D cameras, ground-penetrating radar and a point hyperspectral spectrometer. Building on our previous description of the platform, this paper presents the rationale behind Oliwall’s mechanical design, in particular its compliant wheeled turret, details the finite-state machine and wall-scanning strategy that drive its missions, and reports its first deployment outside the laboratory, in a real disused building. During two days of field tests, we evaluated mobility on slopes of up to 13°, rough ground and small obstacles, autonomous navigation and mapping, and on-site data acquisition with every sensing modality. The results confirm the benefit of the push driving orientation, identify suspension preload and traction as the main mobility limitations, and yield a prioritized list of improvements towards fully autonomous operation.
16:30 – A Human-Centric Integrated Robotic Assistant for Mediating Tasks
Authors: Valerio Bo¹,², Lavinia Hriscu¹,², Alberto Sanfeliu¹,² and Anaís Garrell¹,²
Affiliations: ¹ Institut de Robòtica i Informàtica Industrial (CSIC-UPC), Barcelona, Spain; ² Universitat Politècnica de Catalunya – BarcelonaTech (UPC), Barcelona, Spain.
Abstract
Dynamic operational environments often involve continuous communication among diverse stakeholders, where repetitive information exchange increases workload and introduces workflow delays. This paper presents a demonstration of an autonomous robotic assistant designed to streamline routine communication tasks in a hospital environment. The system orchestrates a multiparty workflow in which a mobile robot proactively interacts with staff, navigates to specific locations, engages end users in conversational assessment, and relays collected data back to supervisors. The robot integrates autonomous navigation, spoken dialogue, and human engagement estimation to dynamically regulate interaction timing and support smoother communication. The core contribution of this work lies in integrating these functional modules into a coherent, end-to-end task orchestration pipeline. We report key lessons learned from the implementation and provide generalizable guidelines for researchers and practitioners deploying socially interactive robots in complex service environments.
16:45 – Digital Twins for Autonomous Drone Deployment
Authors: Juan Dendarieta, Luis Riazuelo and Eduardo Montijano
Affiliations: DIIS-I3A, Universidad de Zaragoza, Zaragoza, Spain.
Abstract
This paper presents a portable digital twin framework for autonomous quadrotors that pairs an identical software stack with real-time bidirectional communication between the virtual and physical environments. Developed to eliminate the severe physical and operational risks of testing complex autonomous maneuvers directly on hardware, this architecture provides a safe workflow that tightly bridges the sim-to-real gap. The framework integrates an offline trajectory generation pipeline with a modular ROS middleware layer and the PX4 autopilot stack. For validation, identical end-to-end flight profiles—covering takeoff, minimum-snap trajectories, and landing—were executed in both a Gazebo simulation and on real hardware. Quantitative tracking evaluation shows a close agreement between both domains, and a virtual obstacle that exists only in the twin is shown to stop the real drone through its onboard collision prevention. Transitioning the complete software stack requires only minimal, localized networking adjustments.
17:00 – Comparing Laser Cut Surface Quality Assessment Methods and Predicting Surface Roughness with Deep Learning
Authors: Ricardo Meireles¹, Levy Mitsugui¹, Jorge Gil¹, Armando Sousa² and Ana Reis¹
Affiliations: ¹ Department of Mechanical Engineering, Faculty of Engineering, University of Porto (FEUP), Porto, Portugal; ² Faculty of Engineering, University of Porto (FEUP), and INESC TEC – INESC Technology and Science, Porto, Portugal.
Abstract
Traditional surface roughness characterization methods like profilometry are time-consuming and impractical for large-scale industrial inspection. This study proposes a novel system that estimates critical surface topography metrics (Ra, Rq, Rz) from standard RGB imagery to meet ISO 9013 standards. Models built on a pretrained ResNet-18 backbone, trained on labels validated against high-precision 3D scanning and contact profilometry, achieved a positive test R² in every material regime and roughness parameter (0.37 to 0.86), whereas feature-based tabular models failed to generalize across materials. Fusing image features with material and process parameters was most beneficial for a single multi-material model, raising the R² of Rz from 0.56 to 0.67. While promising, data scarcity remains the primary limitation.
This work explores feasible approaches for vision-based quality inspection suitable for integration into smart factory production lines.
17:15 – Title to be announced
Authors: Jennifer Gago
Affiliations: MathWorks
Abstract
TBA
17:30 – Simulation-Based Validation of a Geometric-Kinematic Algorithm for Initial Configuration Selection in Robotic Hepatectomy
Authors: Carlos Fontúrbel, Ana Cisnal, Hugo García-Sanz and Juan Carlos Fraile-Marinero
Affiliations: Institute of Advanced Production Technologies (ITAP), University of Valladolid, Valladolid, Spain.
Abstract
In multi-robot systems for liver surgery, the success of the intervention depends on solving the Initial Configuration Selection (ICS) problem to prevent mechanical collisions and kinematic locking. This paper presents the simulation-based validation of a geometric-kinematic optimization framework for this problem. Candidate base positions are filtered with a tool-length-dependent threshold on the Conditioning Index of the normalized Jacobian, and the two homolateral bases are selected by simulated annealing with a cost function that rewards their separation and penalizes isolated feasible positions, while the contralateral base is placed with a centroid-based heuristic. Three UR3e manipulators were evaluated in four scenarios of advanced and expert difficulty on the Iwate scale: left-sided hepatectomy, right hepatectomy, Segment IV segmentectomy, and major hepatectomy. ROS 2/Gazebo simulations yielded collision-free configurations for the first three. In the major hepatectomy, the limited reach of the UR3e forced the right assistant and the endoscope holder into close base positions, and link collisions occurred. The framework can thus act as a preoperative digital twin to assess the technical feasibility of a robotic setup and port map before surgery.