This post is also available in:
עברית (Hebrew)
A team of researchers from Queensland University of Technology (QUT) has introduced an innovative approach to robotic navigation, inspired by the brains of insects and animals. This breakthrough could lead to more energy-efficient autonomous systems, especially in environments where power conservation is critical.
According to TechXplore, the team’s work, published in IEEE Transactions on Robotics focuses on a unique place recognition algorithm using Spiking Neural Networks (SNNs). These networks mimic the way biological brains process information through brief, discrete signals, similar to how neurons communicate in animals. postdoctoral research fellow Somayeh Hussain explained that this design makes the system particularly well-suited for neuromorphic hardware, which emulates biological neural systems and allows for faster processing with much lower energy consumption. This feature is crucial for autonomous robots, which often rely on computationally intensive and energy-consuming systems for navigation in dynamic and complex environments.
While robotics has made great strides, modern robots still face challenges in navigating unknown spaces, often depending on traditional AI systems that require significant energy. However, Dr. Tobias Fischer from the QUT Center for Robotics explained that animals excel at efficiently navigating vast, dynamic environments. This natural efficiency served as inspiration for the QUT team’s work, which aims to replicate some of these biological navigation abilities in robots.
The QUT system uses small neural network modules to recognize specific locations from sequences of images, rather than relying on individual static images. This method improved the system’s place recognition accuracy by 41%, allowing robots to better adapt to environmental changes such as varying seasons or weather conditions. By combining these modules into a scalable ensemble of spiking networks, the system can effectively navigate large and changing spaces.
The team’s solution was Demonstrated on a robot and showed real promise in practical applications where energy efficiency is critical, such as space exploration and disaster recovery operations.
