Towards a Higher Level of Human-Robot Interaction and Integration

François Michaud, Dominic Létourneau, Maxime Fréchette, Eric Beaudry, Carle Côté, Froduald KabanzaAAAI 2006

Introduction

Operating an autonomous robot in a public, "open" setting represents one of the most significant challenges in modern robotics. Beyond the basic requirement of moving from point A to point B, a truly functional "robot servant" must overcome four major bottlenecks: system integration, natural interaction in noisy environments, environmental adaptation to dynamic changes, and transparency of action. This paper details the development of Spartacus, an autonomous mobile robot designed for the AAAI Mobile Robot Challenge, which aims to move beyond individual capabilities toward a cohesive, integrated social entity.

System Architecture and Methods

Spartacus is built upon a custom hardware platform equipped with a SICK laser range finder, a pan-tilt-zoom camera, and a specialized eight-microphone array for sound source localization. The software "brain" of the robot relies on three key architectural components designed to manage complexity:

  • MARIE (Middleware for Robotic Integrated Systems): A framework that integrates disparate software tools—such as the CARMEN navigation suite and the CLSU dialogue toolkit—into a unified system.
  • MBA (Motivated Behavioral Architecture): A goal-management system that allows the robot to balance its own intrinsic needs with external human requests.
  • Reactive HTN (Hierarchical Task Network) Planner: Unlike traditional static planners, this system breaks tasks into sub-tasks in real-time, allowing the robot to adapt to blocked paths or changing priorities while respecting temporal constraints (e.g., scheduled meetings).

Interaction and Transparency

A central tenet of the Spartacus project is the "Robot Servant" paradigm. To simplify high-level decision-making, the robot is given only one intrinsic "survival" goal: keeping its battery charged. All other tasks—such as delivering messages or "schmoozing" with conference attendees—are generated by human requests, creating a clear hierarchy for autonomous decision-making.

To facilitate better human-robot cooperation, the authors developed LogViewer. This tool tracks and visualizes the robot's internal states, including active tasks, processing loads, and planned paths. By making the robot's "thought process" visible, the researchers aim to provide the transparency necessary for humans to understand and trust autonomous behavior.

The LogViewer interface provides a comprehensive view of the robot's internal state, including task hierarchies and spatial localization.

Results and Key Insights

The deployment of Spartacus at the AAAI Mobile Robot Challenge yielded several key results:

  • Successful Multi-Modal Integration: The team successfully coordinated audio (sound localization), visual (camera tracking), and graphical (touch screen) capabilities to create robust interactions in a crowded environment.
  • Autonomous Recharging Strategy: The robot utilized a vision capability to recognize electrical outlets. When its battery ran low, it would locate an outlet and ask a human for assistance in plugging in, successfully fulfilling its primary survival goal.
  • Dynamic Planning in Practice: The HTN planner allowed the robot to manage multiple, overlapping human requests simultaneously while navigating the unpredictable environment of a busy conference hall.

Conclusion and Future Work

While Spartacus demonstrates a high level of functional integration, the authors identify several areas for continued development. Future work will focus on transitioning the LogViewer from an offline debugging tool to a real-time, onboard interface for users. Additionally, the team aims to perfect Simultaneous Localization and Mapping (SLAM) for entirely unknown environments and further refine the social nuances of robot interaction to make the experience more human-centric.

Summary updated

BibTeX
@inproceedings{michaud2006towards,
  author = {François Michaud and Dominic Létourneau and Maxime Fréchette and Eric Beaudry and Carle Côté and Froduald Kabanza},
  title = {{Towards a Higher Level of Human-Robot Interaction and Integration}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2006},
}
Figure