Control semántico de un robot rehabilitador mediante modelos de lenguaje
DOI:
https://doi.org/10.17979/ja-cea.2026.47.13823Palabras clave:
IA generativa/LLMs para robótica y control , Interacción humano-en-el-bucle mejorada con LLMs, Interfaces inteligentes, Ingeniería de rehabilitación y prestación de servicios sanitarios , Control supervisor y autómatasResumen
Los modelos de lenguaje (LLMs) ofrecen nuevas posibilidades para simplificar la interacción con plataformas robóticas de rehabilitación, cuya parametrización de bajo nivel constituye una barrera reconocida de usabilidad clínica. Este trabajo propone una arquitectura para el control semántico de la plataforma Discover2Walk, un robot de rehabilitación de la marcha infantil accionado por cables. Una capa agéntica basada en LLM interpreta órdenes en lenguaje natural emitidas por el clínico y genera un plan de acciones estructurado. Una capa de seguridad determinista valida algorítmicamente cada acción frente a un catálogo de rangos admisibles antes de su ejecución.
Finalmente, una capa de actuación materializa las consignas validadas como comandos ROS~2 sobre los módulos de la plataforma. La arquitectura se valida en simulación bajo tres condiciones experimentales y de forma preliminar sobre el robot real, mostrando su viabilidad preliminar para traducir la intención clínica en consignas seguras sobre el hardware robótico.
Referencias
Chen, W. et al., 2025. Llm-enabled incremental learning framework for hand exoskeleton control. IEEE Transactions on Automation Science and Engineering 22, 2617–2626. DOI: 10.1109/TASE.2024.3382679
Driess, D. et al., 2023. PaLM-E: An embodied multimodal language model. In: Proceedings of the 40th International Conference on Machine Learning. Vol. 202 of Proceedings of Machine Learning Research. PMLR, pp. 8469–8488. DOI: 10.5555/3618408.3618748.
Emken, J. L. et al., 2005. Robot-enhanced motor learning: Accelerating internal model formation during locomotion by transient dynamic amplification. IEEE Transactions on Neural Systems and Rehabilitation Engineering 13 (1), 33–39. DOI: 10.1109/TNSRE.2004.843173
Huang, W. et al., 2023a. VoxPoser: Composable 3D value maps for robotic manipulation with language models. In: Proceedings of the 7th Conference on Robot Learning. Vol. 229 of Proceedings of Machine Learning Research. PMLR, pp. 540–562.
Huang, W. et al., 2023b. Inner monologue: Embodied reasoning through planning with language models. In: Proceedings of the 6th Conference on Robot Learning. Vol. 205 of Proceedings of Machine Learning Research.PMLR, pp. 1769–1782.
Ichter, B. et al., 2023. Do as i can, not as i say: Grounding language in robotic affordances. In: Proceedings of the 6th Conference on Robot Learning. Vol. 205 of Proceedings of Machine Learning Research. PMLR, pp. 287–318.
Kim, E. et al., 2025. Clinician-directed large language model software generation for therapeutic interventions in physical rehabilitation. arXiv preprint arXiv:2511.18274. DOI: 10.48550/arXiv.2511.18274
Klaic, M. et al., 2024. Application of the extended technology acceptance model to explore clinician likelihood to use robotics in rehabilitation. Disability and Rehabilitation: Assistive Technology 19 (1), 52–59. DOI: 10.1080/17483107.2022.2060356
Liang, J. et al., 2023a. Code as policies: Language model programs for embodied control. In: 2023 IEEE International Conference on Robotics and Automation (ICRA). IEEE, pp. 9493–9500. DOI: 10.1109/ICRA48891.2023.10160591
Liang, Y. et al., 2023b. Taskmatrix.ai: Completing tasks by connecting foundation models with millions of apis. arXiv preprint arXiv:2303.16434.
Marchal-Crespo, L. et al., 2009. Review of control strategies for robotic movement training after neurologic injury. Journal of NeuroEngineering and Rehabilitation 6, 20. DOI: 10.1186/1743-0003-6-20
Meyer, J. T. et al., 2021. An analysis of usability evaluation practices and contexts of use in wearable robotics. Journal of NeuroEngineering and Rehabilitation 18 (1), 170. DOI: 10.1186/s12984-021-00963-8
Nicora, G. et al., 2025. Healthcare practitioners and robotic-assisted rehabilitation: Understanding needs and barriers. Journal of NeuroEngineering and Rehabilitation 22, 78. DOI: 10.1186/s12984-025-01593-0
Romero-Sorozabal, P. et al., 2024. Discover2walk: A cable-driven robotic platform to promote gait in pediatric population. In: 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). pp. 4678–4685. DOI: 10.1109/IROS58592.2024.10802838
Schick, T. et al., 2023. Toolformer: Language models can teach themselves to use tools. In: Advances in Neural Information Processing Systems. Vol. 36, pp. 68539–68551. DOI: 10.52202/075280-2997.
Singhal, K. et al., 2023. Large language models encode clinical knowledge. Nature 620 (7972), 172–180. DOI: 10.1038/s41586-023-06291-2
van Dellen, F. et al., 2022. Settings matter: A scoping review on parameters in robot-assisted gait therapy identifies the importance of reporting standards. Journal of NeuroEngineering and Rehabilitation 19, 40. DOI: 10.1186/s12984-022-01017-3
Varela, I. D. et al., 2026. POIROT: Interrogating agents for failure detection in multi-agent systems. arXiv preprint arXiv:2606.02282. DOI: 10.48550/arXiv.2606.02282
Xia, Y. et al., 2025. Control industrial automation system with large language model agents. In: 2025 IEEE 30th International Conference on Emerging Technologies and Factory Automation (ETFA). pp. 1–8. DOI: 10.1109/ETFA65518.2025.11205539
Yao, S. et al., 2023. React: Synergizing reasoning and acting in language models. In: International Conference on Learning Representations.
Zitkovich, B. et al., 2023. Rt-2: Vision-language-action models transfer web knowledge to robotic control. In: Proceedings of the 7th Conference on Robot Learning. Vol. 229 of Proceedings of Machine Learning Research. PMLR, pp. 2165–2183.
Descargas
Publicado
Número
Sección
Licencia
Derechos de autor 2026 Pablo Romero-Sorozábal, Iñaki Dellibarda Varela, Rafael Sendra-Arranz, Juan Miguel Valverde-García, Annemarie F. Laudanski, Eduardo Rocon
Esta obra está bajo una licencia internacional Creative Commons Atribución-NoComercial-CompartirIgual 4.0.
