Integrating neuromorphic action-oriented perceptual inputs to generate a navigation behaviour for a robot.

Area of Science:

• Robotics
• Computational Neuroscience
• Artificial Intelligence

Background:

• Attentional processing is crucial for robots to interact with dynamic environments.
• Existing robotic systems often lack sophisticated mechanisms for selective stimulus processing.

Purpose of the Study:

• To introduce and evaluate a novel neural network architecture, the pointer map, for implementing attentional processing in robots.
• To demonstrate the utility of pointer maps in enabling a robot to maneuver in relation to attended objects.

Main Methods:

• Developed a controller using two pointer maps and a motor map.
• The first pointer map processed one-dimensional distance data from infrared sensors to identify obstacle direction.
• The second pointer map processed two-dimensional image data from a CCD camera for obstacle detection and direction.

Main Results:

• The pointer map architecture successfully enabled selective processing of salient stimuli.
• The robot controller, utilizing pointer maps, demonstrated effective maneuvering capabilities in relation to attended objects.
• The system integrated sensor data (infrared and camera) for robust environmental perception and navigation.

Conclusions:

• Pointer map architectures offer a viable approach for implementing efficient attentional mechanisms in robotic systems.
• This architecture facilitates selective processing and vectorial representation of stimuli, crucial for complex robotic tasks.
• The findings suggest potential for advancing robot autonomy and interaction through biologically inspired neural network designs.