Don't call it K-9. NSK developes robotic guide dog. Now with Kinect-O-Vision!

Japanese manufacturer NSK, maker of robot-related parts including ball bearings, has announced that they are developing a quadrupedal robot that could serve as a guide dog for the blind. Now working closely with the UEC, they are hard at work on a prototype, which uses a Microsoft Kinect sensor to detect stairs and other obstacles. New features include voice recognition, so that the robot can be easily commanded to start, stop, and move up and down the stairs.

NR002 model taking on some stairs

Earlier prototypes were much less advanced. The original NR001 had been completed in 2005 by a small team of engineers who had graduated from the University of Electro-Communications. Next up, developed in 2007, was NR002 model. This model was able to recognize stairs without the need for special markers and change its posture to successfully move up and down each step. At that time, three of the four legs had to remain in contact with the ground to support the body.  Each leg had 4 degrees of freedom and a wheel on each foot so that on flat terrain it could roll rather than walk at speeds up to 3.7 kph. Even when rolling on its wheels, it could easily change direction by swiveling its legs.

Aside from Kinect, in the new model, the legs have been completely redesigned to move almost as quickly as a person up and down stairs, as opposed to the earlier crab-like configuration which was much slower.  Additionally, they’ve begun to add other features, such as a longer leash, and each foot now has a bumper sensor to avoid crashing into things.

There are a few advantages to having a robot instead of a flesh and blood MBF (man's best friend). For satarters, a robot equipped with GPS and google maps could give you directions on your way to a destination, while also providing instructions such as when to stop. Hoping to commercialize the robot by 2020, they seem determined to put out a convincing product out of all this research and development.

Obsticle 

Automatic level detection

Source: Plastic Pals