Ein integrates with ROS (Robot Operating System). It takes input on feeds such as ROS Image topics. It outputs on standard ROS topics such as object_recognition_msgs.

Ein does not have to use ROS; indeed one of the first robots we integrated it with besides Baxter was the Aibo, and it connects over a TCP/IP connection. However data structures such as ros::Time and the like are embedded into Ein’s code, so it must have a ROS environment to compile, currently ROS Indigo.

Sending Commands to Ein

You can send commands to Ein by publishing on the topic /ein/right/forth_commands or /ein/left/forth_commands for each hand. These topics take a std_msgs/String which is a valid Back program. You can use any ROS program to publish to this topic, including rostopic pub. For an example, see sendToBothArms.sh which sends commands to each arm.

Document color calibration

Write about the file system and where objects are stored

which one is focused?

setTargetClass (but change it to focused class)

Write a word to map the whole table

tableMap

add pictures of predicted map

scenePredictBestObject tableUpdateMaps

To add objects to a map:

tableTakeScene (move the arm to take the map) scenePredictBestObject (predict the best object) tableUpdateMaps (visualize what you predicted)

(random things I don’t understand that seem to result in a recognized object array being published) 0 sceneMapSceneObject recordPreTargetLock recordPostTargetLock publishRecognizedObjectArrayFromBlueBoxMemory (rename this)

Topic: /ein_right/blue_memory_objects

Bug: predict orientation in recognized object array

Write about Ein’s architecture (ros, not ros, qt, everything else)