Resource-efficient bio-inspired visual processing on the hexapod walking robot HECTOR

Area of Science:

• Robotics and Artificial Intelligence
• Bio-inspired Computing
• Neuroscience and Neuromorphic Engineering

Background:

• Insects exhibit remarkable efficiency in processing visual motion for navigation.
• Existing robotic navigation systems often require significant computational resources.
• There is a need for resource-efficient navigation solutions for mobile robots.

Purpose of the Study:

• To develop a bio-inspired collision avoidance and navigation controller.
• To implement this controller on a novel System-on-Chip (SoC) hardware module.
• To evaluate the performance of this system on a hexapod robot (HECTOR).

Main Methods:

• Emulation of insect visual motion processing using bio-inspired algorithms.
• Implementation on a dynamically reconfigurable logic-based SoC hardware module.
• Control of a stick insect-like hexapod robot (HECTOR) for visually-guided navigation.

Main Results:

• HECTOR successfully navigated to predefined goals while avoiding obstacles.
• The SoC-based system demonstrated superior speed and resource efficiency compared to CPU and GPU implementations.
• The system's efficiency makes it suitable for fast-moving robots, including drones.

Conclusions:

• Bio-inspired visual motion processing on dedicated hardware offers a highly efficient solution for robotic navigation.
• The developed SoC module significantly advances the capabilities of visually-guided robots.
• This technology has the potential to enable autonomous navigation in complex environments for various robotic platforms.