|
|
|
|||
|
|||||
|
|
|
|
|
User Login |
|
|||||
|
Vision Robotics Technology
Mapping The mapping consists of the creation of a 3-D grid that represents the area in the environment of interest and a landmark and object database that contains information about generic landmarks (e.g. lines and corners) or specific objects (e.g. plug that the robot is plugged into). The 3-D grid has an X-Y extent for the area that the robot operates in and a height that also covers the robot height. The grid contains voxels, which are typically the same size in all 3 dimensions and, for example are 2 cm in all dimensions for the vacuum cleaner. The 3-D grid is initialized to completely unknown and information is added to it using "ray throwing." Ray throwing is the algorithm where a piece of an object as identified by the stereo matching algorithm as located relative to the camera in 3-D space and an imaginary ray is thrown from the camera to the object. Between the camera and the object, those elements are marked as being empty with the final element at the location of the object is marked as full. The 3-D map is converted into a 2-D map using the height of the robot, and this 2-D map is used for path planning, task planning (e.g. where to clean) and object avoidance. The 3-D grid can be easily modified by un-throwing rays to support the recursive nature of the localization used in VRC's SLAM algorithm. The stereo matching algorithms also identify the 3-D coordinates of lines, corners, and other objects relative to the controller robot. These landmarks are added to a 3-D database, which combines similar objects to calculate a more accurate estimate of the 3-D position of these objects in the world. |
|
| Contact Legal Privacy |