Researchers at Delft University of Technology have developed an autonomous navigation strategy inspired by insect behavior for tiny, lightweight robots. This strategy enables these robots to navigate back to their starting point after long journeys, using minimal computational resources and memory. The research, detailed in the journal Science Robotics on July 17, 2024, is expected to have practical applications in areas such as warehouse stock monitoring and industrial site inspections.
Tiny robots, often weighing between tens to a few hundred grams, possess potential for various real-world applications due to their safety, ability to navigate narrow spaces, and cost-effectiveness. However, their limited computational resources make autonomous navigation challenging. Current navigation systems for larger robots, which often rely on heavy sensors or detailed 3D maps, are unsuitable for these smaller counterparts.
To address this, the Delft researchers drew inspiration from ants, which use a combination of visual recognition of their environment and step counting to navigate. They developed a navigation strategy for a 56-gram drone equipped with an omnidirectional camera. This strategy requires only 0.65 kilobytes of memory to cover distances up to 100 meters. The visual processing is performed by a micro-controller, similar to those found in many inexpensive electronic devices.
Insects, particularly ants, use a technique known as “odometry” to keep track of their movements and “view memory” to recognize visual landmarks. The researchers’ strategy employs a “snapshot” model, where the drone periodically captures images of its surroundings. When the drone needs to return to its starting point, it compares its current view with the snapshots, adjusting its path to minimize differences. This method requires significantly less memory than creating detailed maps and relies on odometry to travel between snapshots.
The strategy allows snapshots to be spaced further apart, reducing memory consumption and enabling the drone to cover greater distances. This approach mimics the way ants use landmarks to navigate and adjust their route based on odometry, ensuring the drone remains within a navigable range.
Guido de Croon, a professor specializing in bio-inspired drones and co-author of the study, highlighted the potential applications of this technology. Although the navigation strategy does not generate maps and only allows the robot to return to the starting point, it is sufficient for tasks such as stock tracking in warehouses or crop monitoring in greenhouses. Drones can gather data and return to a base station, storing mission-relevant images for post-processing without relying on them for navigation.
Image credit Guido de Croon / TU Delft|MAV Lab
