Publications
Cable-Driven Exoskeleton for Ankle Rehabilitation in Children with Cerebral Palsy
Iñaki Dellibarda Varela, Pablo Romero-Sorozabal, Gabriel Delgado-Oleas, Jorge Muñoz, Álvaro Gutiérrez, Eduardo Rocon
Cerebral palsy is the leading cause of motor disability in early childhood, with no curative treatment currently available. To mitigate its effects and promote motor rehabilitation, robotic-assisted therapies have emerged as a complement to conventional physiotherapy. In particular, cable-driven exoskeletons offer notable advantages, providing patients with additional mobility and interaction with their environment while preserving motion assistance. Within this context, the Discover2Walk project introduces a modular cable-driven robotic platform designed for early-stage gait rehabilitation. This article presents a novel ankle control module capable of actuating 3 degrees of freedom: 2 translational (in the x and z directions) and 1 rotational (dorsiflexion/plantarflexion). Experimental results confirm the technical feasibility of the approach and its effectiveness in guiding motion within the targeted degrees of freedom.
Rethinking the Illusion of Thinking
Iñaki Dellibarda Varela, Pablo Romero-Sorozabal, Eduardo Rocon, Manuel Cebrian
Earlier this year, Apple ignited controversy by publishing "The Illusion of Thinking," prompting heated debate within the AI community. Critics seized upon the findings as conclusive evidence that Large Reasoning Models (LRMs) lack genuine reasoning capabilities, branding them as mere stochastic parrots. Meanwhile, defenders-spearheaded by Lawsen et al. (2025)-fired back, condemning the experimental setup as flawed and the conclusions overstated. We clarify this debate by replicating and refining two of the original study's most contentious benchmarks: Towers of Hanoi and River Crossing. By introducing incremental stepwise prompting and agentic collaborative dialogue, we show that previously reported failures solving the Towers of Hanoi were not purely result of output constraints, but also partly a result of cognition limitations: LRMs still stumble when complexity rises moderately (around 8 disks). Moreover, the River Crossing results initially heralded as catastrophic failures turn out to hinge upon testing unsolvable configurations. Once we limit tests strictly to solvable problems-LRMs effortlessly solve large instances involving over 100 agent pairs. Our findings ultimately defy simplistic narratives: today's LRMs are stochastic, RL-tuned searchers in a discrete state space we barely understand. Real progress in symbolic, long-horizon reasoning demands mapping that terrain through fine-grained ablations like those introduced here.
Sensorimotor features of self-awareness in multimodal large language models
Iñaki Dellibarda Varela, Pablo Romero-Sorozabal, Diego Torricelli, Gabriel Delgado-Oleas, Jose Ignacio Serrano, Maria Dolores del Castillo Sobrino, Eduardo Rocon, Manuel Cebrian
Self-awareness - the ability to distinguish oneself from the surrounding environment - underpins intelligent, autonomous behavior. Recent advances in AI achieve human-like performance in tasks integrating multimodal information, particularly in large language models, raising interest in the embodiment capabilities of AI agents on nonhuman platforms such as robots. Here, we explore whether multimodal LLMs can develop self-awareness solely through sensorimotor experiences. By integrating a multimodal LLM into an autonomous mobile robot, we test its ability to achieve this capacity. We find that the system exhibits robust environmental awareness, self-recognition and predictive awareness, allowing it to infer its robotic nature and motion characteristics. Structural equation modeling reveals how sensory integration influences distinct dimensions of self-awareness and its coordination with past-present memory, as well as the hierarchical internal associations that drive self-identification. Ablation tests of sensory inputs identify critical modalities for each dimension, demonstrate compensatory interactions among sensors and confirm the essential role of structured and episodic memory in coherent reasoning. These findings demonstrate that, given appropriate sensory information about the world and itself, multimodal LLMs exhibit emergent self-awareness, opening the door to artificial embodied cognitive systems.
Embodied artificial cognition: evaluating self-awareness in multimodal large language models with robotic sensory integration
Iñaki Dellibarda, Eduardo Rocon, Manuel Cebrián
Self-Awareness --- the capacity of an individual to represent and understand itself as the subject of experience and action --- is sustained as the foundation of intelligence and autonomous behavior. The most recent advances in AI have reached human-like performance in tasks that integrate multimodal information, especially in large language models (LLMs), which has raised interest in the embodiment capabilities of AI agents in non-human platforms such as robots. For centuries, different fields of study, from philosophy to neuroscience, have devoted significant efforts to the definition and characterization of Self-Awareness. In the present study, the capabilities of a LLM to develop Self-Awareness are analyzed when embedded in an autonomous mobile robot, relying solely on sensorimotor experience. By integrating a multimodal LLM into an autonomous mobile robot, we test its capacity to achieve artificial Self-Awareness. We find that the system demonstrates solid environmental awareness, self-recognition, and predictive awareness, which allows it to infer its robotic nature and movement characteristics. Structural Equation Modeling (SEM) reveals how sensory integration influences different dimensions of Self-Awareness and its coordination with past–present memory, as well as the hierarchical internal associations that drive self-identification. Moreover, through SEM we identify similarities between the cognitive constructs developed by the system and the human brain structures responsible for Self-Awareness. Ablation tests of sensory inputs identify critical modalities for each dimension, demonstrate compensatory interactions between sensors, and confirm the essential role of structured episodic memory in coherent reasoning. These findings show that, given adequate sensory information about the world and itself, multimodal LLMs exhibit emergent Self-Awareness, opening the door to embodied artificial cognitive systems.
A Cable-Driven Exoskeleton to Control Ankle Mobility During Gait in Children with Cerebral Palsy
Iñaki Dellibarda Varela, Pablo Romero-Sorozabal, Gabriel Delgado-Oleas, Álvaro Gutiérrez, Jorge Muñoz, Eduardo Rocon
Cerebral palsy is the most common neurological disorder in children. In the search for effective treatments, robot-assisted therapy techniques have emerged in recent years, combining physiotherapy with the use of exoskeletons for rehabilitation. In this context, the Discover2Walk project was developed, a cable-driven modular exoskeleton designed for early gait rehabilitation. In this paper, a novel method for foot and ankle joint control, easily implementable in the Discover2Walk, is presented. This method is capable of directing movement in three degrees of freedom: two translational (cartesian x and z axes) and one rotational (dorsiflexion and plantar flexion). The results obtained demonstrate the viability of the proposed system in controlling motion in the three degrees of freedom of interest.
Implementación y control de un exoesqueleto basado en estructura de cables para la asistencia de tobillo durante la marcha en niños con parálisis cerebral
Iñaki Dellibarda, Pablo Romero-Sorozabal, Gabriel Delgado-Oleas, Álvaro Gutiérrez, Jorge Muñoz, Eduardo Rocon
Cerebral palsy is the most common neurological disorder in children. To treat it, robot-assisted therapy techniques have emerged in recent years, combining physiotherapy with the use of exoskeletons for rehabilitation. In this context, the Discover2Walk project was developed, a cable-driven modular exoskeleton designed for early gait rehabilitation. In this article, a novel method for controlling the foot and ankle joints, easily implementable in the Discover2Walk, is presented. This method is capable of directing movement in three degrees of freedom: two translational (Cartesian x and z axes) and one rotational (dorsiflexion and plantar flexion). The results obtained demonstrate the viability of the proposed system to control motion in the three degrees of freedom of interest.
© 2025 Iñaki Dellibarda. All rights reserved.