Gaze Control of a Robotic Head for Realistic Interaction With Humans

Area of Science:

• Robotics
• Artificial Intelligence
• Cognitive Science

Background:

• Gaze control is crucial for effective human-robot interaction (HRI) in face-to-face communication.
• Robots interacting with multiple people require sophisticated mechanisms to manage attention and social cues.
• Existing methods often struggle with dynamic environments and multiple interlocutors.

Purpose of the Study:

• To present a novel neurorobotics method for robots to determine whom to attend to in multi-person interactions.
• To enable robots to dynamically shift focus based on various social and conversational cues.
• To replicate human-like conversational behavior and attention-shifting in robotic systems.

Main Methods:

• A competitive neural network architecture was developed to process multiple stimuli (e.g., gaze, speech, pose).
• The network dynamically weighs competing inputs to decide the focus of attention.
• A robotic head with an integrated virtual agent and gaze control system was implemented within a ROS architecture.

Main Results:

• The proposed method allows for smooth transitions in the robot's focus of attention.
• The system effectively handles multiple interlocutors appearing and disappearing from the robot's view.
• Experiments demonstrated the successful integration and functionality of the gaze control system in a robotic head.

Conclusions:

• The novel competitive network approach provides an effective solution for robot gaze control in multi-person HRI.
• The dynamic attention mechanism enhances the naturalness and effectiveness of human-robot conversations.
• This work contributes to more sophisticated and socially aware robotic interaction capabilities.