Have you ever considered how insects can venture far from their habitat and yet navigate their way back home with precision? The solution to this query holds significance not just in the biological realm but also in the development of artificial intelligence (AI) for miniature, autonomous robots.
Recently, a team of drone-researchers from TU Delft has drawn inspiration from an unlikely source. They studied how ants use visual cues to familiarize themselves with their surroundings and match this knowledge with their step-counting ability to return home safely. Using these insights, they developed an insect-inspired autonomous navigation strategy that is ideally suited for small, lightweight robots. The novel technique lets these robots trace their route back home after prolonged journeys while demanding extremely little computational power and memory. Robots employing this strategy would need only 0.65 kilobytes of memory for every 100 meters travelled.
These autonomous robots may soon find extensive utility in various industries. From monitoring stock levels in warehouses to detecting gas leaks in industrial sites, their potential applications are vast. This groundbreaking research was published in their findings in Science Robotics on July 17, 2024.
The AI required for independent navigation with limited onboard resources was primarily developed with large robots in mind, such as self-driving cars. Certain approaches depend on bulky, power-consuming sensors like LiDAR laser rangers, which simply can't be carried or powered by small robots. Other alternative strategies rely on vision, which is power-efficient and provides rich data about the environment. However, these methods usually aim to create detailed 3D maps of the environment, necessitating intensive computation and memory—resources small robots lack.
This predicament led researchers to seek inspiration from nature, particularly insects. Operating over distances relevant to many real-world applications while using scarce sensing and computing resources, insects indeed make compelling study subjects. Specifically, insects combine keeping track of their own motion (known as 'odometry') with visually guided behaviors based on their low-resolution, but almost omnidirectional visual system (termed 'view memory')
The proposed insect-inspired navigation strategy allowed a 56-gram "CrazyFlie" drone, fitted with an omnidirectional camera, to cover distances of up to 100 meters using only 0.65 kilobytes of memory. The drone performed all the necessary visual processing on a puny computer chip, known as a "micro-controller," commonly found in low-cost electronic devices.
The application of tiny autonomous robots in real-world scenarios could be a significant step forward, thanks to this new strategy. The proposed strategy's functionality is more limited compared to advanced navigation methods because it doesn't create a map and only enables the robot to return to its starting point. Yet, it may be more than sufficient for various applications such as stock tracking in warehouses or crop monitoring in greenhouses. These bots could vent out, collect data, and return to the base for further processing, negating the need for elaborate navigation schemes.
Disclaimer: The above article was written with the assistance of AI. The original sources can be found on ScienceDaily.
