A Biological Inspired Cognitive Framework for Memory-Based Multi-Sensory Joint Attention in Human-Robot Interactive...

Area of Science:

• Robotics
• Cognitive Science
• Human-Robot Interaction

Background:

• Effective human-robot collaboration requires shared attentional focus, known as joint attention.
• Current human-robot interaction often uses fixed robot behaviors, lacking dynamic adaptation to interactive scenarios.
• Joint attention is crucial for proficient human-robot collaborations.

Purpose of the Study:

• To develop autonomous attentional behavior for robots using multi-sensory perception to match human attentional focus.
• To investigate the improvement of human-robot joint attention with a novel biologically-inspired memory-based attention component.
• To assess the model's performance in real-world unstructured environments with the iCub humanoid robot.

Main Methods:

• Implemented a multi-sensory perception system for robots to autonomously track human attention.
• Integrated a biologically-inspired memory-based attention component to enhance joint attention dynamics.
• Evaluated the system using the iCub robot in a joint task with a human in an unstructured real-world setting.
• Compared robot and human attention performance across audio-visual and audio-only modalities.

Main Results:

• The autonomous attention model demonstrated robust performance in capturing stimuli, localizing targets, and executing actions.
• Robot's temporal attention dynamics were found to be compatible with human participants.
• The robot performed better in audio-visual conditions compared to audio-only conditions for localization tasks.
• Egonoise was identified as a significant factor affecting audio-only localization performance.

Conclusions:

• The developed model offers an effective solution for memory-based joint attention in dynamic, real-world environments.
• The robot's performance in audio-visual joint attention is comparable to human behavior.
• Further research may be needed to mitigate egonoise effects for improved audio-only localization in robots.