engagement Documentation

engagement: Human-robot engagement

Human-robot engagement package developed at Worcester Polytechnic Institute by Prof. Charles Rich et al. Currently contains one node, for recognition of directed gaze, mutual facial gaze, adjacency pairs and backchannels.

• Author(s): Brett Ponsler
• Review Status: unreviewed
• Review Notes:
• Homepage: http://sourceforge.net/projects/ros-engagement/
• Downloadable files

ROS Message and Service Types

ROS Message Types	ROS Service Types
Boolean Engagement Empty GestureObjects Actor Objects Gesture Utterance	ObjectsBoolean ActorEngagement ActorBoolean

engagement is "the process by which two (or more) participants establish, maintain and end their perceived connection during interactions they jointly undertake" (Sidner, C.; Lee, C.; Kidd, C.; Lesh, N.; and Rich, C. 2005. Explorations in Engagement for Humans and Robots. Artificial Intelligence 166(1-2): 140-164).

ROS API

List of nodes:

• engagement/recognition

---

recognition

recognition Receives incoming messages from the collaboration and vision layers and uses these messages to recognize engagement behaviors between an actor and a robot during an interaction.

Usage

$ recognition [standard ROS args]

Example

$ recognition

ROS topics

Subscribes to:

Topics from Collaboration

• "collab/salient": [engagement::Objects] A salient object from the collaboration layer.
• "collab/utterance": [engagement::Utterance] An start or end of an utterance was detected.

Topics from Control

• "control/look": [engagement::GestureObjects] The robot is looking at a set of objects.
• "control/point": [engagement::GestureObjects] The robot pointed at a set of objects.
• "control/shake": [engagement::Empty] The robot shook his head as an acknowledgement behavior.
• "control/nod": [engagement::Empty] The robot nodded his head as an acknowledgement behavior.

Topics from Vision

• "vision/point": [engagement::GestureObjects] A human hand pointed at a set of objects.
• "vision/facial_gaze": [engagement::Gesture] A human looked at the robot's face.
• "vision/look": [engagement::GestureObjects] A human looked at a set of objects.
• "vision/shake": [engagement::Gesture] A human shook his head.
• "vision/nod": [engagement::Gesture] A human nodded his head.
• "vision/face": [engagement::Actor] A human face was detected.

Publishes to:

• "recognition/status": [engagement::Actor] The current engagement status of an actor.

ROS parameters

Reads the following parameters from the parameter server

• "recognition/conf/window_size" : [double] The window size for computing engagement statistics.
• "recognition/conf/directed_gaze/max_delay" : [double] The maximum delay for the directed gaze recognizer.
• "recognition/conf/mutual_facial_gaze/max_delay" : [double] The maximum delay for the mutual facial gaze recognizer.
• "recognition/conf/adjacency_pair/max_delay" : [double] The maximum delay for the adjacency pair recognizer.
• "recognition/conf/backchannel/max_delay" : [double] The maximum delay for the backchannel recognizer.

ROS services

• "recognition/statistics": [engagement::ActorEngagement] The goal to return the current engagement statistics.
• "recognition/directed_gaze": [engagement::ObjectsBoolean] The goal to establish directed gaze with a set of objects, with the success being returned.
• "recognition/mutual_facial_gaze": [engagement::ActorBoolean] The goal to establish mutual facial gaze, with the success being returned.
• "recognition/adjacency_pair": [engagement::ActorBoolean] The goal to establish adjacency pair, with the success being returned.