Imagine a robot that can find its way around using only a tiny sip of power, navigating the world with efficiency inspired by the human brain. Researchers at Queensland University of Technology (QUT) have developed just such a system, called LENS (Locational Encoding with Neuromorphic Systems). This breakthrough in low-energy robot navigation uses brain-inspired computing to dramatically reduce energy consumption, opening doors for robots to explore longer and further in challenging environments.
Contents
Key Takeaways:
- A new system, LENS, allows robots to navigate using visual information.
- It mimics neural processes, requiring up to 99% less energy than traditional methods.
- The system uses an “event camera” and brain-like computing.
- This could enable robots to operate much longer on limited battery power.
The Challenge: Hungry Robots
Modern robots are powerful, but they often consume a lot of energy, especially when processing complex visual information to understand their surroundings and figure out where they are – a crucial task called visual place recognition. This high energy demand limits how long robots can operate, which is a major hurdle for missions in remote areas like deep space, underwater, or during search and rescue operations where power sources are scarce.
Traditional systems capture full images, processing every pixel, every frame, constantly. Think of it like taking thousands of detailed photographs every second, even when nothing in the scene changes. This requires significant processing power and, consequently, energy.
Introducing LENS: Navigating Like a Brain
The QUT team, including neuroscientist Dr. Adam Hines, Professor Michael Milford, and Dr. Tobias Fischer, tackled this problem by looking to nature. Their LENS system is built on “neuromorphic computing,” which is designed to function more like biological brains.
“To run these neuromorphic systems, we designed specialised algorithms that learn more like humans do, processing information in the form of electrical spikes, similar to the signals used by real neurons,” explained Dr. Hines. Instead of processing full images continuously, LENS uses a special piece of hardware: an “event camera.”
Researchers at QUT with a robot using the new LENS low-energy navigation system
An event camera doesn’t capture frames like a regular camera. It only reacts to changes in brightness at each pixel, moment by moment. This is similar to how our own eyes and brains are highly sensitive to movement and changes in our visual field. This “change-aware” sensing combined with a low-power, brain-inspired chip and specialized algorithms allows the system to be incredibly efficient.
Remarkable Efficiency and Potential
The results are striking. In tests, the LENS system was able to recognize locations along an 8km journey while using under 10 per cent of the energy needed by traditional visual navigation systems. Dr. Hines notes this represents up to a 99 per cent reduction in energy requirements for the visual localization part of the task.
Furthermore, the system required only 180KB of storage for the 8km path – nearly 300 times less data than comparable methods. This low data footprint is another advantage for systems with limited memory.
“This system demonstrates how neuromorphic computing can achieve real-time, energy-efficient location tracking on robots,” Dr. Hines said.
What This Means for Future Robots
This development is more than just a technical achievement; it has practical implications for the future of robotics. By drastically cutting the power needed for navigation, LENS could enable robots to:
- Explore Deeper and Further: Underwater vehicles or space probes could cover much larger distances on the same battery charge.
- Operate Longer in Emergencies: Search and rescue robots could stay active for extended periods in disaster zones.
- Become More Accessible: Lower energy needs can mean smaller, cheaper batteries, potentially making sophisticated navigation available for a wider range of robotic applications.
The QUT researchers believe their new system could change how neuromorphic computing is integrated into real-world robotics, pushing the boundaries of what’s possible for robots operating far from a power outlet.
This energy-saving approach, mimicking the efficiency of the brain, represents a significant step towards creating more capable and enduring robotic explorers.
