How can we help robots?

A principal question in the field of robotics is how can robots help us?, but to make real progress perhaps we should wonder ourselves how can we help robots?

Robots are nowadays really efficient in repetitive industrial tasks. However, at present they are not suited to share with us our physical and cognitive world in a natural way: there are still many open questions to be solved in order to make robots to 'live' and cooperate with people.

The most adapted agent to this scenario of co-living and cooperation with humans, as a result of about a hundred thousand  years of evolution and adaptation, is not other than ... the human being. So, why do not we try to learn from humans how they solve the different challenges where robots are not so efficient? Perhaps with this approach we could devise how to make better robots. 

In short, with the intriguing topic above we are proposing a research area with the goal of studying human beings in order to learn from them how they perform in those areas were robots are really inefficient. Our field of study will be primarily concerned with cognitive processes from the  perception-action loop, with the aim of devising behavioral human-like strategies for artificial robots.


Human Senses

There are many human senses that allow human minds to detect relevant information from the body and the environment. Such senses go from five to even twenty one depending on the source of information (
  • Basic Human Senses
    • Ophthalmoception (Eyes) Sight or visual perception
    • Audioception (Ears) Hearing or auditory sensations
    • Gustaoception (Tongue) Sense of taste
    • Olfacoception or Olfacception (Nose) Sense of smell
    • Tactioception (Skin)
  • Four Internal Human Senses
    • Thermoception (Skin) Lack or increase of heat (temperature)
    • Proprioception (Body Parts) Awareness of body parts without visual input
    • Nociception (Whole Body) Sensation of pain in the body (skin, body organs, etc.)
    • Equilibrioception (Whole Body) Sense of balance (determined by ear fluid)
  • Additional Human Senses
    • Kinesthetic Sense (Whole Body) Sense of acceleration
    • Tactility (Mostly the Skin) Perception of pressure
    • Chemoreception (Blood and Brain) Sensation of hunger, thirst, vomiting and suffocation
    • Stretch Reception (Muscles, Joint and Skin) Sense of gag reflex, gas distension, excretion, etc.
    • Cutaneous Reception (Skin) Sense of skin vasodilation (like flushed skin)
    • Synaesthesia (Body Parts) Combination of senses (like smiling at someone's voice)
    • Sixth Sense (Small Brain) Sense of intuition (gut feeling)
    • Premonition (Paranormal) Subconscious sense of future events (usually danger)
    • Telepathy (Paranormal) Auditory perception of a person's (near or far) thoughts
    • Precognition (Paranormal) Visual perception of future events
    • Clairvoyance (Paranormal) Visual perception of invisible objects or events
    • Clairaudience (Paranormal) Auditory perception of the invisible

To be able to infer cognitive processes from the perception-action loop in humans we must be able in first instance of measuring what human senses are detecting. So we must develop sensing devices and from them models of the perceptual capabilities of humans. Our primary concern will be with the development of:
  • Models for the visual system, specially for attentional procedures emcompassigg bottom-up (saliency-based) and top down (goal oriented).
    • Development of hardware devices and models for the separation of central (fovea, point of sight) and peripheral vision.

  • Models for the hearing system.
  • Models for the kinestesic, propioceptive and equilibrioceptive systems.

The methodoly for the development will be basically the study of the literature about human perception and its translation to computational models from sensors. Once developed they will be used to infer what the human is "sensing" from sensors placed in his body.

From these we will develop an integrated framework useful to process incoming information as real senses do. These models will be used as inputs for the develppment of cognitive models for the perception-action loop. However, once available, why not to think about other promissing application, such as for instance the replacement of damaged senses of people,or the augmentation of natural senses? although ... perhaps better to defer this for the future ... .

Related information

  1. "From the human visual system to the computational models of visual attention: a survey". Silvio Filipe, Luis A.Alexandre. Artificial Intelligence Review. 2013.
  2. "Computational models of the auditory system.

    Propioceptive system for robot imitation

    Imitation of complex movements may be difficult for robots. In this project we plan to teach a robotic arm complex movements for a human arm. We will place sensors in the human arm to detect the output of the propioceptive uman system. These sensors will also be place in the robotic arm with the purpose of reproducing the movement.
    Also, maybe it would be necessary to answer how do we learn complex movements?
  1. Koenemann, J.; Bennewitz, M., "Whole-body imitation of human motions with a Nao humanoid," Human-Robot Interaction (HRI), 2012 7th ACM/IEEE International Conference on , vol., no., pp.425,425, 5-8 March 2012
  2. Evaluation Metrics and Results of Human Arm Movement Imitation.
  3. Brain Res Cogn Brain Res. 1998 Oct;7(2):191-202. Fixation behavior in observation and imitation of human movement. Matarić MJ, Pomplun M.Source Neuroscience Program and the Computer Science Department, University of Southern California, 941 West 37th Place, Los Angeles, CA 90089-0781, 
  4. Cortical Mechanisms of Human Imitation. Marco Iacoboni, Roger P. Woods, Marcel Brass, Harold Bekkering, John C. Mazziotta, and Giacomo Rizzolatti. Science 24 December 1999: 286 (5449), 2526-2528.

Evaluación de la carga cognitiva

La carga cognitiva en una determianda tarea puede indicar el esfuerzo computacional o la complejidad de procesamiento inherente al bucle percepción-acción. En este apartdo se deberían proponer mecanismos para evaluar la carga computacional, posiblemente aquellos basados en el procesamento de señales de EEG portables que hasta donde parece podrían ser aplicable (google cognitive load eeg).