Sebastian Rojas-Ordoñez

All five papers are open access. Press Read online to open the full text here, or download the PDF.

1. Read online
   MCMDS · Taylor & FrancisJuly 2026Journal article30 pages

   An FMI-centric open-source and interoperable framework for AI-driven robotic systems: a model-based methodology for integrating physical and AI models

   Mathematical and Computer Modelling of Dynamical Systems, 32(1), 2707718

   S. Rojas-Ordoñez, M. Segura, J. Arboleda, E. Zulueta

   The development of AI-driven cyber-physical robotic systems often depends on proprietary toolchains that limit flexibility and interoperability. This paper presents a modular, open-source, and FMI-centric methodology that integrates OpenModelica-generated Functional Mock-up Units (FMUs), ONNX-based perception pipelines, and ROS 2 middleware into a unified robotic workflow. Beyond software connectivity, the approach makes explicit how perception and control interact through supervisory logic, timing assumptions, and component encapsulation. The methodology is validated on a sign-following mobile robot against an industrial Simulink-based baseline. Both systems achieved 100% task success under nominal conditions. The Simulink baseline was slightly more path-efficient, whereas the FMI-based workflow exhibited substantially lower end-to-end latency. A controlled perception-dropout study further quantified degradation in efficiency, success rate, and latency. A qualitative TurtleBot3 deployment also supports portability beyond simulation. Overall, the results support open, standards-based robotic workflows that are transparent, portable, and reproducible.

   Read moreShow less

   • FMI / FMU
   • OpenModelica
   • ONNX
   • ROS 2
   • End-to-end latency

   Read online Download PDF5.4 MB DOI

   Open access · CC BY 4.0 · © The Author(s)

2. Read online
   ACME · SpringerJuly 2026Review article44 pages

   Bridging the Embedded Execution Gap: A Review of the Functional Mock-up Interface for Actuator-Centric Execution Workflows in Industrial Embedded Systems

   Archives of Computational Methods in Engineering

   S. Rojas-Ordoñez, M. Segura, I. Yarza, E. Zulueta

   Industrial actuators integrate multi-domain physical dynamics with embedded control algorithms under strict timing constraints. Despite mature model-based design workflows, a persistent embedded execution gap separates validated high-fidelity simulation models from deterministic, resource-bounded execution on embedded industrial platforms. This review examines the Functional Mock-up Interface (FMI) as an interoperability enabler for actuator development workflows and argues that interoperability at the model-exchange level does not automatically translate into interoperability at the execution level. The review analyzes the evolution of FMI from version 2.0 to 3.0, systematizes actuator-relevant fidelity requirements (nonlinearities, compliance/resonance, timing effects), and evaluates representative approaches for integrating FMI into industrial and robotic toolchains. Two deployment trajectories are synthesized: (i) runtime-constrained FMU execution on edge/MCU-class platforms for prototyping, monitoring, and non-safety-critical functions; and (ii) analyzable code-generation paths such as Embedded FMI (eFMI) and GALEC, which target bounded memory and timing transparency for certification-oriented workflows. Although FMI 3.0 introduces improved timing semantics through clocks and Scheduled Execution, end-to-end feasibility remains limited by toolchain maturity, scheduler integration patterns, and the scarcity of complete quantitative evidence on jitter, WCET, and memory footprint under realistic load. The review identifies key challenges in determinism, fidelity management, certification alignment, and ecosystem-level conformance/validation, clarifying the technical barriers that must be addressed to fully bridge the gap between high-fidelity simulation and embedded actuator deployment.

   Read moreShow less

   • FMI 2.0 / 3.0
   • eFMI · GALEC
   • Scheduled Execution
   • WCET & jitter
   • Industrial actuators

   Read online Download PDF4.8 MB DOI

   Open access · CC BY 4.0 · © The Author(s)

3. Read online
   MAKE · MDPIFebruary 2026Journal article23 pages

   Plug-and-Play LLM Knowledge Extraction for Robot Navigation: A Fine-Tuning-Free Edge Framework

   Machine Learning and Knowledge Extraction, 8(2), 49

   S. Rojas-Ordoñez, M. Segura, I. Yarza, V. Mendoza, E. Zulueta

   Large Language Models are increasingly used for high-level robotic reasoning, yet their latency and stochasticity complicate their direct use in low-level control. Moreover, extracting actionable navigation cues from multimodal context incurs inference costs that are challenging for embedded platforms. We present a plug-and-play framework that augments a finite-state machine with asynchronous velocity suggestions generated by a Large Language Model, using an off-the-shelf DistilGPT-2 model running on-device on a Jetson AGX Orin. The system extracts task-relevant cues from the current context and integrates them only if they satisfy deadline, schema, and kinematic validation, thereby preserving a deterministic 50 Hz control loop with a <5 ms fallback path. We compare multiple Large Language Models for embedded robot control and quantify trade-offs among model size, inference time, and output validity. To assess whether the Large Language Models add value beyond signal processing, we include an ablation against a standard smoothing baseline; the results indicate that the Large Language Models contribute anticipatory, context-dependent adjustments that are not captured by filtering alone. Experiments in Gazebo and on a real TurtleBot3 reduce the final position error from 0.246 m to 0.159 m and improve trajectory efficiency from 0.821 to 0.901 without increasing control-loop latency. Approximately 80% of the Large Language Models’ outputs pass validation and are applied. Overall, the framework reduces developer effort by enabling behavioral changes at the prompt level while maintaining interpretable, robust edge-based navigation.

   Read moreShow less

   • Large language models
   • Robot navigation
   • Edge AI · Jetson AGX Orin
   • Finite-state machine
   • TurtleBot3

   Read online Download PDF8.3 MB Code & data DOI

   Open access · CC BY 4.0 · © The Author(s)

4. Read online
   Applied Sciences · MDPIDecember 2025Journal article19 pages

   Natural-Language Relay Control for a SISO Thermal Plant: A Proof-of-Concept with Validation Against a Conventional Hysteresis Controller

   Applied Sciences, 15(24), 12986

   S. Rojas-Ordoñez, M. Segura, V. Mendoza, U. Fernandez, E. Zulueta

   This paper presents a proof-of-concept for a natural-language-based closed-loop controller that regulates the temperature of a simple single-input single-output (SISO) thermal process. The key idea is to express a relay-with-hysteresis policy in plain English and let a local large language model (LLM) interpret sensor readings and output a binary actuation command at each sampling step. Beyond interface convenience, we demonstrate that natural language can serve as a valid medium for modeling physical reality and executing deterministic reasoning in control loops. We implement a compact plant model and compare two controllers: a conventional coded relay and an LLM-driven controller prompted with the same logic and constrained to a single-token output. The workflow integrates schema validation, retries, and a safe fallback, while a stepwise evaluator checks agreement with the baseline. In a long-horizon (1000-step) simulation, the language controller reproduces the hysteresis behavior with matching switching patterns. Furthermore, sensitivity and ablation studies demonstrate the system’s robustness to measurement noise and the LLM’s ability to correctly execute the hysteresis policy, thereby preserving the theoretical robustness inherent to this control law. This work demonstrates that, for slow thermal dynamics, natural-language policies can achieve comparable performance to classical relay systems while providing a transparent, human-readable interface and facilitating rapid iteration.

   Read moreShow less

   • Natural-language control
   • Relay with hysteresis
   • Local LLM
   • SISO thermal plant
   • Safe fallback

   Read online Download PDF1.4 MB DOI

   Open access · CC BY 4.0 · © The Author(s)

5. Read online
   Modelica & FMI ConferenceSeptember 2025 · LucerneConference paper10 pages

   Integration of Physical and AI Models Using Open and Interoperable Standards: A Model-Based Methodology for Autonomous Robot Development

   Proceedings of the 16th International Modelica & FMI Conference, Lucerne, Switzerland

   S. Rojas-Ordoñez, M. Segura, E. Zulueta

   The development of Cyber-Physical Systems, particularly in the field of autonomous mobile robotics, often relies on proprietary environments, which limit flexibility and interoperability. This paper proposes a modular and open-source methodology that enables the modeling and simulation of such systems using non-proprietary tools. The methodology integrates Functional Mock-up Interface standard for model exchange, Open Neural Network Exchange standard for Convolutional Neuronal Networks algorithms integration, ROS2 as robotic middleware, and Gazebo as the simulation environment. To validate the approach, we applied it in the development of a mobile robot that navigates autonomously by following traffic signals. This implementation demonstrates that these open technologies can be effectively combined, overcoming common integration barriers among proprietary tools. The proposed workflow provides a practical alternative to proprietary solutions and demonstrates the feasibility of integrating open standards for the development of autonomous robotic systems.

   Read moreShow less

   • Autonomous robotic systems
   • FMI
   • ONNX
   • ROS 2
   • Gazebo

   Read online Download PDF8.4 MB DOI

   Open access · CC BY 4.0 · © The Author(s)