BRAINMOVE - Neural Sensorimotor Modelling for Adaptive Movement Control

Eduardo Ros, Jesús Garrido, Ríchard R. Carrillo, Diego Navarro, Juan Helios García Guzmán, Adán Cruz, Brayan Alfonso Valencia Vidal, José Bernardo Martínez, Javier Ubago, María P. Tirado, Francisco Naveros Arrabal, Ignacio Abadia, Álvaro González

Tue, 01 Sep 2026

From neural circuit to robot — the BRAINMOVE pipeline: neural system simulation (cerebellum, basal ganglia and hippocampus), vestibulo-ocular-reflex and musculoskeletal models, and collaborative robot control.

Principal investigators: Eduardo Ros Vidal and Jesús Garrido Alcázar.

Biological motor control is far more efficient, adaptive and robust than the control strategies we can currently build into robots, and it does not rest on a single brain structure. BRAINMOVE investigates how key brain systems — the cerebellum, the basal ganglia and the hippocampus — jointly support efficient, adaptive and safe motor control, and how these principles can be translated into machine-learning controllers for collaborative robots with complex, compliant dynamics.

Biological motor control relies on distributed learning mechanisms: supervised sensory-prediction-error learning in the cerebellum, reinforcement learning and action selection in the basal ganglia, and sequential state representation and planning in the hippocampus. Their integration remains poorly understood. In parallel, the emergence of collaborative robots (cobots) with elastic actuation for passive safety compliance imposes new control challenges: elastic components improve safety, but at the cost of highly complex dynamics, which are difficult to model accurately and therefore challenging for traditional control methods. This has driven a growing interest in data-driven and machine-learning-based control, which promises to learn control laws or internal models directly from experience.

The project therefore studies:

How the computational principles of the cerebellum and the locus coeruleus support supervised sensory-prediction-error learning and the neuromodulation of learning rates.
How the basal ganglia implement reinforcement learning and action selection, and how the hippocampus sustains sequential state representation and planning.
How these subsystems complement each other in the execution of movement, with a focus on the plasticity mechanisms that underpin them.
How these computational and learning principles can be applied to efficient and accurate control of collaborative robotics with complex internal dynamics.

The central motivation of BRAINMOVE is to bridge these two worlds: on the one hand, to advance our understanding of the computational principles implemented by the cerebellum, the basal ganglia and the hippocampus in motor control; on the other, to translate these principles into bio-inspired learning architectures for the control of complex, compliant robots. A better understanding of plasticity dynamics, of how they are supported by the neurophysiological substrate, and of the construction of accurate models able to reproduce experimental results will also contribute to identifying dysfunctions associated with ageing or specific pathologies.

All of this is addressed within the framework of sensorimotor control tasks, validated both with simulated experimental set-ups — such as the vestibulo-ocular reflex (VOR) — and on collaborative robotic platforms.