Sie sind hier:


Project Group Close Encounters - Robot Proxemics

Project Group WS 2012/13

Master ET/IT, Master A&R


Close Encounters - Robot Proxemics

_DSC0972 Kopie 2

 A key aspect of nonverbal social interaction is proxemics, namely the way in which the interacting partners mutually establish and control their spatial separation and posture while being involved into the communication situation. The mobile robot should not only move safely and collision-free, but also perform a legible, socially appropriate and interaction-dependent trajectory. Hence, appropriate proxemic behaviour of a robotic system is an important key to communication with humans – the human interaction partner should be comfortable with the interaction and communication situation.

The project group is a joint collaboration between the chair for control and systems engineering (RST) and the image analysis group (BV). The project is concerned with the perceptual basis of robot proxemics (BV) and proxemics control of a mobile robot (RST). The project group is open to students in the Master ET&IT as well as Master A&R.

The project seeks a principled integrated approach that combines different modalities (such as movement trajectories, gaze, head orientation, and body pose) in perception as well as in action to produce a coherent robot behaviour. The development of such a framework is the main goal of this project.

The project investigates the following research issues:

  • Which cues and signals shape human-robot proxemic behaviour and in which way?
  • Which is the appropriate integrated approach to combine different modalities (e.g., trajectories, gaze, head orientation, body pose) in perception and action to produce a coherent robot proxemic behaviour?

Perception for robot proxemics

The perception of human movement and posture rests upon the Kinect 3D sensor. The availability of Kinect and the effort that the robotics community has put into the development of algorithms for human-robot interaction shows the potential of this technology. The software developer kit including the natural user interface (NUI) library offers a variety of algorithms and methods for tracking people, recognizing poses and gestures from the depth and image stream provided by the Kinect sensor. The aim is to predict the movements and intentions from the perception of the current physical state of the interactors and the movement history.

Proxemics control

Proxemics control is concerned with the design and implementation of closed-loop robotic proxemics behaviour, modular representations of motion control, task and context specific constraints, coordinated and synchronized movements. Basic reactive robotic behaviours are designed that closely mimic the social customs and patterns of human encounters in natural environments. This observation motivates a mechanism for negotiation via non-verbal but overt signals in order to mutually agree upon a desired and comfortable formation in order to establish a formation that is consistent with the social and cultural rules and patterns of interaction.

The following basic scenarios are to be investigated:

  • Lateral passing: How does the robot approach and pass an individual or a group of individuals that humans perceive as comfortable and safe?
  • Intersection encounters: How does the robot react when it reaches an intersection, junction, or door at the same time as a human?
  • Establishing interaction and formations: How does the robot recognise the human‟s intent to establish an interaction and how does it correct its own movement in response to its partner?
Contact: Frank Hoffmann, Christian Wöhler
Participants: max 10 participants
Background: robotics, computer vision, MATLAB, ROS