LIG laboratory proposes a PHD with REETI

Laboratories LIG (teams AMA, MAGMA, GETALP) and GIPSA-lab (team AGPIG) are proposing a PHD

The purpose of the project is to design a mobile communicating companion robot, to provide assistance, and to ensure follow-up of impaired people (elderly, disabled…). The role of the robot is to perceive the situations of the “accompanied” person, his behavior and needs, based on various sensors placed on both of them. The overall objective is that the robot should be able to accompany and assist impaired persons in a “natural” way, at home or during a walk, and raise alerts in case of detected problem (fall, weakening…).

The work falls under the context of cooperation between two research laboratories: LIG (teams AMA, MAGMA, GETALP) and GIPSA-lab (team AGPIG).

The work for the proposed PhD will concentrate on the development of perception and information processing system, which involves the development of models for the dynamic analysis of complex situations.

The robot has been acquired in the framework of another university project. It is built upon a Robosoft Robulab10 mobile platform equipped with a Robopec Reeti expressive head, and video and audio sensors to perceive its environment and analyze the followed person’s movement and behavior. It is further able to capture physiological data (heart and breathing rates…) from the sensors attached to the person assisted.

Data fusion techniques will be developed to analyze this multidimensional heterogeneous information and to describe the person situation, in terms that are physiological (emotional state, stress, weakness…), behavioral (difficulty moving, fall…) and environmental (presence of crowd, obstacles, ambient noise…). To cope with uncertainty, incompleteness and inconsistency of the data, hypothesis management and uncertainty models will be designed.

Various problems are to be solved:

  •  Setting up the sensor and data transmission infrastructure for the robot and person assisted
  •  Low-level data processing (filtering, sampling), high level semantic information extraction (location characteristics, audio ambiance…)
  •  Setting up scenarios to create a learning database reflecting the various situations

From a more theoretical standpoint, the issues are:

  • Fusion of heterogeneous data streams (video, audio, physiological sensors…)
  • Modeling dynamic situations (transitions, events)
  • Developing models to cope with data uncertainty, incompleteness and inconsistency and manage multiple hypotheses
  • Development of active learning methods
  • Designing methods for the detection of abnormal a priori unknown situations
Contact :
Olivier Aycard (LIG/AMA - University of Grenoble1 (UJF) - UFR IM2AG - BP 53 - Centre Equation 4 - 38041 Grenoble Cedex 9 - FRANCE)
This e-mail address is being protected from spambots. You need JavaScript enabled to view it.