Robot End Effector Tracking Using Predictive Multisensory Integration.

Area of Science:

• Robotics
• Computational Neuroscience
• Artificial Intelligence

Background:

• Human eye-hand coordination relies on integrating visual and proprioceptive information.
• Robots often require explicit markers or known kinematics for precise arm tracking.
• Learning complex sensorimotor tasks remains a challenge in humanoid robotics.

Purpose of the Study:

• To develop a biologically inspired model for humanoid robots to learn end-effector tracking.
• To investigate the role of sensorimotor prediction in multisensory integration for robotic control.
• To enable robots to track their end-effectors without explicit markers or known arm kinematics.

Main Methods:

• A novel model integrating visual and proprioceptive cues for sensorimotor prediction.
• The robot learns to predict sensory consequences of its motion from proprioceptive feedback.
• The system learns visual feature descriptors during environmental interaction.

Main Results:

• The robot learns smooth pursuit eye movements to track its hand, with or without visual input.
• The model successfully tracks external visual motions.
• The framework demonstrates improved multisensory integration through predictive processing.

Conclusions:

• Sensorimotor prediction enhances multisensory integration in robots.
• The proposed model achieves human-like eye-hand coordination characteristics.
• This approach advances robotic perception and control by learning without explicit prior knowledge.