Institut des Systèmes Intelligents
et de Robotique





Tremplin CARNOT Interfaces



Thesis title : Contribution to the design of interfaces and interactive behaviors for personal robots

Keyword : social robots, interfaces, interactive behaviors, decision making engine, person detection, person following.

30 March à Paris
UPMC – Campus Jussieu, 4 Place Jussieu - 75005 Paris
Tour 55, 2ème étage, couloir 55-65, salle 211

Abstract :

This thesis falls within the context of a number of projects aimed at the design, the development and the evaluation of mobile robotics systems for assistance of people at home. Two different yet complementary aspects are considered:
1.    The definition of a graphical interface that allows users to easily access to the robot functions.
2.    The mobility of the system that should be set tacking into account the environment, the state and activity of the person and the specific context.  

The type of interfaces proposed for the interaction is a very important point for this kind of robot because of the target population: older and non-expert people, often suffering from cognitive disorders. With the aim to make the robot interface perfectly adapted to this kind of population, we propose a method for the icons selection and for testing the usability of the interface. This method has been experienced with some patients of the geriatric hospital Broca.

Problems relating to robot mobility, in cluttered and shared with people environments, are particularly complex. They are based on 2 central issues: the location and the activity of the person in the space and the definition of interactive behaviors. For the detection and localization of the person, a 2D laser-based detector has been implemented. The speed and the trajectory of the person are computed using an extended Kalman filter on the laser data. In order to have a flexible and robust robot behavior, adapted to the context at to the events that occur unpredictably, we have implemented a decision making engine exploiting the SpirOps software. The inputs of this engine are the information extracted from the different robot sensors. These information are treated by using fuzzy logic rules and the engine computes the interest values associated to each action that the robot can execute. The robot decides dynamically witch is the most appropriate action to execute in the different situations.