Color Shirt
In this task, the robot needs to take a picture of a person who
wears a shirt in a specific color.
The color can be
red,
blue,
green or
yellow, and is provided as part
of this task. Since the process of interpreting a color, say blue, can be very
subjective, we will accept a color to be blue, if most people
would identify the color as blue rather than another solid color
such as, red, green or yellow.
The format of "color shirt" is jpg, that saying the completion
certificate of this task is an image in jpg format. The score of an
in-time, correct completion is 100. This task is mainly used for
demonstrating the vision module of intelligent robots. It
makes sense for robots to actively search areas with more human
activities, so robots with a good navigation capability can perform
better in this task.
Target Search
This task requires a robot to visually localize a target object
in a 3D environment.
We maintain a list of objects that are easily accessible anywhere
around the world, such as a can of coke. An instance of this task is
specified by giving the name of a target object (from this list) in
plain text. Therefore, each team needs to maintain a mapping from
the name of any object on the list to an set of visual
representations. A representation can be an image or a point cloud
(in case of robot using RGB-D sensors).
The format this this task is jpg. This task is mainly used for
evaluating the vision and navigation capabilities, as robots need to
simultaneously move its platform from point to point and visually
analyzing scenes being visited.
Human Following
This task requires a robot to follow a human walker for at least
a predefined distance.
The completion certificate of this task includes a picture of the
human walker and an image showing the trajectories traveled by the
human walker and the robot.
Object Delivery
The format of completion certificate is txt. The score of an in-time, correct completion is 150. This task is used for demonstrating human-robot interactions, navigation and vision. The robot needs to find a person P1 who is willing to help specify the object name O, initial room R1, and goal room R2. To avoid language ambiguities, object names and room numbers can be selected through drop-down menus, or the robot can suggest humans to follow well-specified script (e.g., "Please input the room number. You can type 'r101', 'r102', 'r103', or 'r200'"). In so doing, the robot may want to actively seek human input in busy areas. After that, the robot needs to move to R1 to load object O (possibly with human P2's help). Finally, the robot moves object O to room R2 and asks the last human P3 to describe the object with a few words. The completion certificate includes the time-stamped object name given by P1 and the time-stamped words P3 used for describing the object.