Hitachi Ltd announced that it has added "object recognition" and "object search" functions to its "EMIEW2" humanoid robot.
The object recognition function uses a database based on Web information to look for images similar to an object shown by the user with a high-speed similar image search function. Also, it analyzes texts accompanying the images to deduce the name of the object and link the object's name and image.
With the object search function, the robot can search for an object in a building when requested by the user. It compares the image of an object mentioned by the user (the image and the object are linked by using the object recognition function) with images stored in a database about the locations of objects in the building to determine the location of the object.
Moreover, by using the map of the building, the robot determines an optimal route to the location and guides the user to it. The database about the locations of objects is updated as needed by using the images of objects taken by networked cameras installed in the building.
This time, Hitachi improved the movement of the robot guiding the user to a location by employing the "model-prediction posture control technology" to smoothly take a curve. The technology calculates the optimal posture in real time to cope with change in centrifugal force.
By combining the technology with the active suspension mechanism, which is also used for the previous model of the robot, it became possible to make a sharp turn and make several turns in succession without losing speed. The previous model decelerates before turning a corner so that a centrifugal force is not applied to the robot.