Definition of the line

This research line focuses on understanding how humans perceive and act with their body to i) improve artificial intelligence and robotic systems, and at the same time ii) aid to unveil the inner mechanisms of information processing in the brain. We develop novel algorithms for learning, estimation and control of complex systems based on neuroscience findings and evaluate them in robotic platforms (humanoids, manipulators). Our models, besides being relevant for disciplines, such as computational neuroscience, robotics and cognitive science, are particularly interesting for human-centric solutions, e.g., healthcare, human-robot interaction, wearable robotics, etc. In the long run, this research pursues achieving robots with human comparable embodied intelligence. Enabling robots to act and adapt to real-world complex interactions is one of the key challenges of this century.

History/expertise of the group in this line

We have more than ten years of experience in brain-inspired machine learning models for robot perception and action. We pioneered the successful deployment of neuroscience-inspired models (e.g., predictive coding) in humanoid robots as well as replicating human perceptual experiments in robotics systems (e.g., rubber-hand illusion). We are internationally recognized for the active inference approach to robotics, a neuroscience-inspired framework that describes the brain as an inference machine. We are currently investigating new theoretical probabilistic models for practical applications in industrial robotics, environmental monitoring and healthcare. Our research includes:

  • Neuroscience-inspired Artificial Intelligence (NAI).
    • Variational Inference, (Deep) active inference.
    • On-chip spiking neural networks control.
  • Embodied intelligence: Robot learning, estimation and control
  • Computational models of human body perception and action
  • Self-perception, awareness and synthetic consciousness.

Related Projects

Relevant papers

Lanillos, P. et al (2021). Active inference in robotics and artificial agents: Survey and challenges. arXiv preprint arXiv:2112.01871.

Lanillos, P., Franklin, S., Maselli, A., & Franklin, D. W. (2021). Active strategies for multisensory conflict suppression in the virtual hand illusion. Scientific Reports, 11(1), 22844.

Lanillos, P., & Cheng, G. (2018). Adaptive robot body learning and estimation through predictive coding. IEEE/RSJ Int. Conf. on Inte. Robots and Syst. (IROS) (pp. 4083-4090).

Main researchers

Dr. Pablo Lanillos
Dr. Juan C. Moreno
Dr. Jorge García