eva-opencog

eva-opencog

An image containing the OpenCog + Hanson Robotics Eva robot. This is the whole kit-n-kaboodle. In development. Probably broken.

The visual processing+animation pipeline consists of:

• A ROS node that takes video input from a UVC-compatible USB webcam (most typical desktop/laptop cameras are compatible).
• A ROS node that does face detection: it tries to find human faces in the video feed.
• A ROS node that maps the face postions to a 3D coordinate system.
• A ROS node that runs the Eva blender rig itself (i.e. uses blender to draw/animate Eva).

The speech processing pipleing consists of:

• The RelEx server parsing English.
• The OpenCog chatbot (currently not attached to any input source).

Building

Run the ./build.sh file in this directory. You need to have built the ros-incog-blender image (in the ../ros-incog-blender directory) first.

Testing

To verify that blender works, run the ./run.sh shell script. This will start blender and the half-dozen ROS nodes that control it. You should see a living, breating Eva.

You can chat to Eva by going to the guile prompt, and typing in chat text like so:

(process-query 'luser "Look left!")
(process-query 'luser "Smile!")
(process-query 'luser "Look sad!")
(process-query 'luser "Eva, express boredom")

Turn on the non-verbal behaviors by saying:

(run)

If things don't work, open a github issue or complain on the mailing list!

TODO

• Remove joint_state_publisher robot_state_publisher from the geometry startup -- we only need to track face locations via geometry, and not the rest of the stuff.