By Soham NandiReviewed by Frances BriggsJul 3 2025
Scientists have developed a brain-inspired navigation system for robots that cuts energy use by 99 %. The system enables lightweight, real-time operation in challenging environments like disaster zones and space.

Image Credit: R Mendoza/Shutterstock.com
Researchers have unveiled a new navigation system for robots that mimics the way the human brain processes visual information. This system dramatically improves energy efficiency with minimal storage required and without compromising accuracy.
The system, known as LENS, Locational Encoding with Neuromorphic Systems, integrates spiking neural networks, event cameras, and neuromorphic chips to deliver real-time place recognition over an eight-kilometre route using just 180 kilobytes of memory.
The work, published in Science Robotics, enables energy-efficient navigation for resource-constrained robots in applications such as search and rescue or planetary exploration missions.
Brain-Inspired Computing For Robots
Robotic navigation is an energy-intensive process that demands large computational power. Conventional visual place recognition methods process full-frame image data continuously, which requires large amounts of memory and drains power rapidly. This is particularly problematic for smaller robots designed for long-endurance missions.
Neuromorphic computing, which models the brain’s neural processes by sending information as short electrical spikes, offers a promising solution as an efficient alternative. However, most neuromorphic systems struggle to combine low power use with real-world applicability.
The LENS system addresses these challenges by bringing together three complementary, biologically inspired technologies.
The key players are spiking neural networks (SNNs), an event camera that detects pixel-level brightness changes, and a neuromorphic chip. The SNNs replicate the brain’s event-driven approach by activating only when input changes occur, drastically reducing energy consumption. This enables efficient, sparse computation compared to traditional artificial neural networks.
The event camera, also known as a dynamic vision sensor, captures changes in brightness at the pixel level with microsecond precision, rather than recording full images at fixed intervals. This mimics how human vision prioritises moving or changing elements in a scene and avoids unnecessary processing of static information.
A neuromorphic chip, the SynSense Speck, ties the system together, carrying out efficient computation. By performing all processing on board the robot, it avoids the latency and additional energy costs associated with cloud-based computation.
Efficient And Compact Performance
The researchers tested LENS in real-world conditions across an 8 km route and found the robot achieved comparable place-recognition accuracy to conventional methods, such as the sum of absolute differences technique, while consuming just 8 % of the energy.
Its model contains only 44,000 parameters and fits in just 180 kilobytes of memory, a fraction of what traditional systems require. It's ideal for use on edge devices and smaller autonomous platforms.
The system’s microsecond response time and parallel processing allow it to operate in real time, maintaining accuracy in dynamic environments where fast decision-making is critical.
This capability could be used in multiple applications: a robot equipped with LENS could navigate through a collapsed building, responding quickly to shifting debris while conserving enough battery to continue operating for hours longer. It could also prove valuable in underwater or space missions, where communication delays and power constraints make autonomous, efficient operation essential.
Download your PDF copy now!
Future Applications
LENS represents an important step towards making neuromorphic computing practical for real-world robotics. The research team sees potential for extending its use to swarm robotics, where groups of lightweight robots coordinate autonomously, or for developing adaptive learning systems that can improve in the field.
For now, LENS provides a compact, scalable way to bring energy-efficient, brain-inspired navigation to robots. With its 99 % reduced energy consumption, it could be used in some of the most challenging and power-sensitive settings.
Journal Reference
Hines, A. D., Milford, M., & Fischer, T. (2025). A compact neuromorphic system for ultra–energy-efficient, on-device robot localization. Science Robotics, 10(103). DOI:10.1126/scirobotics.ads3968
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.