A Brochette of Socially Interactive Robots

François Michaud, Dominic Létourneau, Pierre Lepage, Yan Morin, Frédéric Gagnon, Patrick Giguère, Eric Beaudry, Yannick Brosseau, Carle Côté, Audrey Duquette, Jean-François Laplante, Marc-Antoine Legault, Pierre Moisan, Arnaud Ponchon, Clément Raïevsky, Marc-André Roux, Tamie Salter, Jean-Marc Valin, Serge Caron, Patrice Masson, Froduald Kabanza, Michel LauriaAAAI 2005

Abstract

Researchers at the Université de Sherbrooke present a diverse collection of four robotic platforms—Spartacus, Tito, Roball-2, and AZIMUT-2—each designed to address specific challenges in social interaction and human-centric locomotion. This paper explores the hardware and software architectures required for high-level social integration, pedagogical therapy, and robust navigation in cluttered environments.

Introduction

Socially interactive robotics represents a multidisciplinary "grand challenge" requiring machines to perceive and respond to natural human cues while navigating unpredictable, human-centric spaces. The authors argue that no single robot design can fulfill all social roles. Instead, they demonstrate a "brochette" (a flavorful collection) of specialized platforms tailored for distinct interaction scenarios, from autism therapy to autonomous "schmoozing."

Methodology: The Robotic Platforms

The research highlights four specific platforms integrated via sophisticated software architectures:

Spartacus: The Socialite

Spartacus is designed for high-level autonomous interaction. It features an Artificial Audition System capable of separating and tracking multiple sound sources in noisy environments. The robot utilizes the MARIE (Mobile and Autonomous Robot Integrated Environment) middleware, which facilitates code reusability by allowing different programming environments (e.g., Player/Stage/Gazebo and CARMEN) to communicate seamlessly.

Tito: The Educator

Tito is a 60 cm tall humanoid robot specifically designed for children with autism. Rather than full autonomy, Tito often employs a "Wizard of Oz" teleoperation approach, allowing educators to control social cues. Its hardware is highly expressive, featuring a smiling mouth, a head capable of indicating "no" or "surprise," and fiber-optic hair that illuminates to engage users.

Spartacus (left) and Tito (right) represent different approaches to social interaction, from autonomous audition to teleoperated pedagogical engagement.

Roball-2: The Toy

Roball-2 is a spherical robot enclosed in a protective shell, designed to survive the rough play of toddlers. Its unique locomotion dynamics—rolling through obstacles—serve as a social cue in itself, generating interest and engagement through movement.

AZIMUT-2: The Navigator

AZIMUT-2 is a highly modular, omnidirectional platform with four independent articulations. It can be configured with wheels, legs, or tracks, allowing it to navigate tight human spaces and varied terrain with ease. It uses elastic actuators to sense environmental effects and a distributed processing architecture for its limbs.

Roball-2 (left) utilizes a spherical form factor for durability and engagement, while AZIMUT-2 (right) provides a modular, omnidirectional base for complex navigation.

Key Insights

  • Middleware Integration: The MARIE middleware is critical for leveraging diverse algorithms for localization, mapping, and vision without reinventing existing frameworks.
  • Form and Function: The physical design (form factor) dictates the nature of interaction. For instance, Roball’s spherical shape is both a protective measure and a source of social engagement.
  • Auditory Spatial Awareness: Real-time sound source separation is essential for robots to function in "cocktail party" scenarios, allowing them to focus on a single speaker amidst background noise.

Results and Discussion

Preliminary trials with Tito involving children with autism (approximately 5 years old) indicated that the robot successfully acted as an incentive for interaction, with observed imitation of the robot's movements. Spartacus demonstrated successful integration of a complex sensor suite (SICK laser, Sony PTZ camera, 8-mic array) and speech recognition for real-world social settings. AZIMUT-2 proved the mechanical versatility of its modular design for navigating human environments.

Conclusion and Future Work

While promising, the research notes several areas for future development. The Tito trials involve small sample sizes, and full autonomous decision-making in complex social scenarios remains a challenge. Future work aims to enhance Roball’s internal perception of child-robot interactions and complete the hardware iterations for AZIMUT-2. Ultimately, this "brochette" of robots illustrates that specialized hardware and flexible software middleware are key to advancing social robotics.

Summary updated

BibTeX
@inproceedings{michaud2005brochette,
  author = {François Michaud and Dominic Létourneau and Pierre Lepage and Yan Morin and Frédéric Gagnon and Patrick Giguère and Eric Beaudry and Yannick Brosseau and Carle Côté and Audrey Duquette and Jean-François Laplante and Marc-Antoine Legault and Pierre Moisan and Arnaud Ponchon and Clément Raïevsky and Marc-André Roux and Tamie Salter and Jean-Marc Valin and Serge Caron and Patrice Masson and Froduald Kabanza and Michel Lauria},
  title = {{A Brochette of Socially Interactive Robots}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2005},
}
Figure