Teresa Escrig

News and oppinion about Cognitive AI & Robotics

Cognitive Robots is collaborating with Robosoft to enhance Kompai companion robot’s capabilities

one comment

Our company Cognitive Robots in Spain, is collaborating with French company Robosoft to provide more intelligence to their Kompai companion robot for the elderly.

Since February 2011, Cognitive Robots and Robosoft have been working together to enhance the “intelligence” of the Kompai companion robotics platform.  The project is funded by the European Commission as part of a set of demonstrations, of the current capabilities of the robots (ECHORD Project called C-Kompai).

The Kompai’s capabilities prior to  incorporating our ‘ Cognitive Brain for Service Robotics’ into the platform is shown in the current GUI control panel.  The robot is controlled using a push button interface as well as voice commands.

The main functions that Robosoft asked Cognitive Robots to improve with the Cognitive Brain were:

  • Previously when a robot was purchased, the Robosoft technician needed to go to the elderly persons home and spend most of the day creating a map of the space that the robot would operate in.  Unfortunately, due to the maps limitations, the elderly couldn’t move any furniture around without the technician coming back again and remapping the environment.  This is a problem that remains in industrial applications as well.
  • Kompai had a limited perception of the plane provided by the laser sensor at a certain height. That was a big problem because any house could have plenty of obstacles that would remain unseen by Kompai.

Cognitive Robots proposed to include two new features:

  • The robot would vacuum the house.
  • To include a more dynamic and proactive behaviour by the robot, than merely waiting to be called to do something.

The Kompai’s capabilities after the Cognitive Brain is fully incorporated will be:

  • Automatic map creation.  Any furniture can be moved around without any technical assistance.
  • 3D obstacle detection using the Kinect sensor
  • ‘Autonomous vacuum cleaner’ capability
  • Proactive behaviour: Kompai will engage actions and interaction with the elderly.

These enhanced capabilities are summarized with the addition of the following three new buttons to the GUI control panel.

The scope of this particular project doesn’t go further, but we’re curious to learn your thoughts on how the behaviour of the Kompai could be further enhanced?

Comments are welcome to contribute to the development of the companion robots industry!

One Response to 'Cognitive Robots is collaborating with Robosoft to enhance Kompai companion robot’s capabilities'

Subscribe to comments with RSS or TrackBack to 'Cognitive Robots is collaborating with Robosoft to enhance Kompai companion robot’s capabilities'.

  1. Proactive robot:
    I think prediction, evaluation and control of outcomes in the world are important concepts to anchor our understanding of how a useful ‘proactive’ robot might operate.
    Mykola Rachevskiy, elsewhere in the discussion, usefully asks if an artificial neural network can be made more like the mammalian neocortex. Yes, indeed biological learning inn the neocortex can be interpreted as a predictive mechanism, not yet perfectly understood, in which there is constant interaction between time-dependent sensations and experience-based predictions.
    In this abstract model, recently explored in the literature, when an incoming sequence of sensations is matched to part of a previously experienced temporal pattern, a prediction can be made which speeds up and, if properly applied, enhances recognition. Note that the model is a hierarchical one in which predictions (known as feed-back) flow outwards from more abstract levels towards more distal ones. Each level is divided into a grid of nodes. Each node takes information from several nodes in the next most distal level in the hierarchy (feed-forward) and passes predictions (feed-back) to that same ‘patch’ of nodes.
    (http://www.numenta.com/htm-overview/education/DileepThesis.pdf – this also covers learning)
    Perhaps this predictive mechanism is a generally adopted feature of the brain that enables it to take information, predict outcomes based on experience or instinct and exert control over those outcomes by means of muscle movements.
    By implication, a robot that harnessed this type of mechanism could predict and control outcomes in the physical world (because that is what brains do).
    A proactive robot could generate predictions based on the current likelihood and desirability of built in or learned outcomes. It would be proactive because it would be sensitive to internal goals and therefore not strictly a stimulus-response system.

    David Green

    29 Apr 12 at 2:40 pm

Leave a Reply