Robotic navigation, Abstract: The navigation capabilities of insects, specifically ants, have inspired advancements in the AI algorithms for tiny, autonomous robots. Researchers at TU Delft have utilized biological findings on ant navigation—where ants visually recognize their environment and count their steps to return home safely—to develop an insect-inspired autonomous navigation strategy for lightweight robots. This strategy allows these robots to navigate long distances and return to their starting point with minimal computational and memory requirements.
Introduction
Insects, particularly ants, exhibit remarkable navigational abilities, often traveling considerable distances from their nests and successfully finding their way back. Understanding these mechanisms is valuable to biology and has profound implications for developing AI systems for small, autonomous robots. TU Delft drone researchers have harnessed these biological insights, specifically how ants use visual recognition of their environment and step counting, to create an autonomous navigation strategy for tiny, lightweight robots. This approach requires minimal computational power and memory (0.65 kilobytes per 100 meters), paving the way for numerous practical applications in the future.
The Potential of Tiny Robots
Tiny robots, ranging from tens to a few hundred grams, have potential applications in various real-world scenarios. Their lightweight nature ensures safety in case of accidental collisions, and their small size allows them to navigate through narrow spaces. Moreover, suppose these robots can be produced cost-effectively. In that case, they can be deployed in large numbers, enabling rapid coverage of extensive areas, such as greenhouses, for early pest or disease detection.
However, the operation of such small robots poses significant challenges due to their limited resources compared to larger robots. Autonomous navigation is particularly problematic, as external infrastructure like GPS is often unavailable or unreliable indoors and in cluttered environments while installing and maintaining indoor beacons is expensive and sometimes impractical.
AI systems designed for autonomous navigation are typically tailored for larger robots, such as self-driving cars, which rely on heavy, power-intensive sensors like LiDAR. Due to their size and power requirements, these sensors are unsuitable for tiny robots. Although energy-efficient, alternative approaches using visual sensors often create detailed 3D maps of the environment, necessitating substantial computational and memory resources beyond the capacity of tiny robots.
Insect-Inspired Navigation
AI, Robotic
To overcome these challenges, researchers have turned to nature for inspiration. Insects, particularly ants, are interesting as they navigate significant distances with minimal sensing and computing resources. Biologists have increasingly understood insects’ strategies, combining odometry (tracking their motion) with visually guided behaviors based on their low-resolution yet omnidirectional visual systems. While odometry is well understood, the mechanisms behind view memory are unclear. One early theory, the “snapshot” model, suggests that insects occasionally take visual snapshots of their surroundings, which they later use to navigate by minimizing visual discrepancies between their current view and the snapshot.
This snapshot-based navigation can be likened to Hansel and Gretel’s strategy of leaving stones to find their way home. For the robot, snapshots serve as these “stones.” However, the robot must be close enough to the snapshot location for effective navigation. Too many snapshots can lead to excessive memory consumption, while too few can cause the robot to get lost.
The innovative aspect of the TU Delft strategy is the integration of odometry with spaced snapshots. This method allows the robot to travel longer distances as it only homes to snapshots occasionally, reducing the need for frequent visual comparisons. This balance minimizes odometry drift while maintaining low memory usage.
Robotic
Experimental Results
The insect-inspired navigation strategy was tested on a 56-gram “CrazyFlie” drone equipped with an omnidirectional camera. The drone successfully navigated distances up to 100 meters, using only 0.65 kilobytes of memory. All visual processing was performed on a micro-controller, a small and inexpensive computer commonly found in various electronic devices.
Applications and Future Work
The proposed navigation strategy represents a significant step toward deploying tiny autonomous robots in real-world applications. Although the functionality is more limited than state-of-the-art navigation methods—providing only return-to-base capability without generating a map—it is sufficient for many practical uses. For example, drones can track inventory in warehouses or monitor crops in greenhouses, collecting data and returning to a base station for post-processing.
This technology’s future potential is vast. As the understanding of insect navigation deepens and computational technologies advance, tiny autonomous robots will become increasingly capable, opening up new possibilities in various industries.
Conclusion
Insights from ant behavior have enabled significant advancements in the autonomous navigation of tiny robots. TU Delft researchers have developed a navigation strategy requiring minimal computational resources by mimicking how ants use visual snapshots and step counting. This breakthrough could lead to practical applications in numerous fields, enhancing efficiency and safety while minimizing costs. The research represents a promising step in integrating biological principles into robotic technology.
