Home

frederic Delaunay edited this page Sep 25, 2013 · 17 revisions
Clone this wiki locally

Welcome to the wiki-based documentation of ARAS.

ARAS is an evolution of the CONCEPT project. Also, it is now the open-source version of Syntheligence ’s ARAS, which is a complete rewrite and is fully supported by Syntheligence.

Formerly, this project was split in 2 main sections:

HRI’s challenge is to bring natural interaction to robots, that is: giving robots the ability to understand and express themselves like humans do. So far, the focus has been on non-verbal communication (facial expression, emotionally driven behavior…) .
In the CONCEPT project, the robotic setup was essentially a novel robotic head (we baptised Rear-projected Animated Face) mounted on a Katana 400M robotic arm, although the software is not specific to a particular platform. Several interacting python modules allow scalability and dynamic programming of the robot.


Here’s the technical news:

Done:

  • Communication Layer : allows python servers to communicate with others using a human-readable protocol.
    • dedicated channels are created and configured from the contents of a configuration file
    • transparent address family supported (socket,named-pipe) from configuration
    • servers are aggregated in a meta server (to have a single process) introducing a new protocol keyword: origin [subserver_name]
    • support for threading.
  • Facial Animation System : enhancing FACS FACS improvements for realtime facial animation.
    • protocol: triplets of AU weights, normalized target value, and attack-time
    • blender backend: runtime detection of character abilities,
    • allows backends for mechatronic faces.
  • Vision module : provides access to cameras and higher-level information.
    • face detection using pyvision
    • other detection algorithms (edge, shape, color)
  • Global Feature Pool: allows external systems to inspect robot state.
    • All sensory (and proprioceptive) data available
    • remote query over the network.
  • Behaviour: a enhanced finite state machine for easy definition of robot behaviour
    • allows multiple state machines to work together, sharing states.
  • Body Management and Hardware Design :
    • Design complete, Laser cutting and rapid prototyping
    • Spine module in charge of spine+neck movement
    • backend available for Katana 450 robotic arm.
    • affective motor dynamics: emotional influence on motor movements and behaviours.

Work In Progress:

  • visual motoring: works but would need a friendlier interface.
  • facial emotion recognition, gaze detection, eye tracking (if possible).
  • laser sensor: support for the Hoyuko URG 04LX-UG01
  • face recognition: to be fully integrated
  • audio source localization: basic code available, to be fully integrated

To Do:

  • Audio Processing : word recognition (subset of english words), speaker recognition (if possible)
  • Speech Synthesis : PSOLA or similar base, realtime pitch modulation according to emotional dimensions,