6. ASSISTIVE AND PERSONAL ROBOTICS: THE DILIGENT PROJECT

Building on the large experience of the HILARE project and on the recent Esprit project MARTHA (multi-robot courtyard servicing), LAAS launched on January 1999 an ambitious project, DILIGENT, with the theme of Robots for Public and Personal Servicing and Assistance.

The project emphasizes Human-Robot Interaction both at the communication and physical levels where the machine is in close contact with the Human, either purposefully (assistance) or by accident.

DILIGENT encompasses all the topics linked to situations where the robot performs its tasks in interaction with humans. The human can be an operator, a user of a service performed by the robot or simply a user (a passer-by) of the same environment. More generally, the project tackles problems such as navigation in presence of humans, transporting humans, handling out objects to humans, manipulating and transporting loads in cooperation with humans, assisting handicapped or elderly humans...

Its aim is to address the real scientific and multi-disciplinary challenge to devise "soft robots" able to "comply" with Humans and to effectively deal with the variety and dynamics of environments where the public is present.

The DILIGENT project is set in the experimental environment which comprises for indoors applications four fully instrumented wheeled mobile-robots, one of which carries a six d.o.f. arm.




6.1 Human-Friendly Navigation

Diligent's essential task is navigation. Depending on the context, it will be endowed with a set of complementary equipment. The target, within the framework of personal robotics, is to endow Diligent with "human-friendly navigation". This notion has still to be precisely defined. In fact, the project itself is envisioned as a means to incrementally define and investigate the main issues it involves.

However, one can already mention the following issues:

• The ability to operate in an environment open to the public with very little or no adaptation. Indeed, "human-friendly navigation" implies safe (for the human and the robot), robust and reasonably efficient (for the task itself) navigation in such an environment.

• "Human-friendly navigation" means also a mobile robot which is easy to instruct (map and locations learning), easy to use, easy to interact with.

• Another aspect should be the ability of the robot to analyze and react efficiently to human motion around it. For example, it may have to make decisions like "follow", or "pass", or "leave room" to a person or a group of persons. In other words, the robot should "comply" with human activities in a given environment.

• One should also consider more subjective issues such as human understanding and human acceptance of the robot behavior.

Hence, the Personal Robot can be envisioned as an extension and a new challenge for autonomous mobile robotics.






Diligent: a Nomadic XR4000




6.2 Results and future work

The basic function that LAAS wanted to demonstrate is the ability of the robot to navigate "safely" and repeatedly in the laboratory environment from one place to another. Indeed, a short-term goal is to allow its effective use by the lab members, through very simple and very intuitive commands.

This first objective has been satisfied. Diligent is capable of long run robust navigation in an environment that can be reliably modeled by horizontal laser scans. The navigation is robust to substantial differences between the actual environment state and the learned map. Besides, Diligent demonstrated effective autonomy in planning and executing its navigation tasks, and re-planning in case of permanent obstacle or a "non-cooperative" human.






Diligent WEB page




However, there is a great difference between running a robot in a "protected" lab environment and leaving the robot "alone" in an environment open to the public.

Concerning the navigation itself, there is clearly a limitation in the localization and obstacle modeling. LAAS intends to develop more generic multi-sensor localization with a special effort dedicated to vision-based localization in environments populated by humans. Besides, it is also necessary to devise a multi-sensor 3D-obstacle detection and avoidance allowing to effectively run in almost any-indoor environment. Another key aspect is learning capabilities and ease of programming. Indeed, substantial progresses should be devoted to map learning and maintenance.

Concerning the robot supervision and the human-robot capabilities, the following issues will be addresses. The robot (the supervisor + various communication means) should be able, whenever it is necessary, to explain its state, its current goal and how it intends to reach it.

LAAS intends to tackle Human-robot interaction in a generic way which is viewed as an incremental and interactive problem solving process. Another interesting aspect is to implement a human-robot interaction based on explicit models of the different "types" of humans: programmer, instructor, user, public. Such interactions will be closely linked to the implementation of perceptual primitives allowing the detection and localization of human, human motion detection, human recognition, and human gesture or movement interpretation.