Sticking Up for the Little Guy
Robots weighing in the tens of grams to a few hundred grams are realistic tiny robots for practical uses because of their modest weight and size. Far from bumping into people and objects, they maneuver through the aisles even if they are narrow. Furthermore, and most importantly, they could be produced relatively cheaply and in good numbers to quickly monitor large areas, for example, greenhouses for detecting the first signs of pest or disease.
However, these microbots are not able to work independently because these are much smaller in mass and have much lesser resources than that of the macrobots. A big challenge is their need to move around on their own. It should be noted that if there is no visual way to navigate, additional facilities are sometimes helpful, for instance GPS satellites when one is in the territory or the wireless communication beacons if one is indoors. GPS is not available indoors, and the signals are extremely poor in built-up areas such as the urban canyon. Furthermore, setting and maintaining beacons inside also can be expensive or impossible in some cases, such as the search-and-rescue mission.
Sensors for current AI for autonomous navigation are currently designed for large equipments like self-driving cars, utilizing bulky, heavy and power consuming instruments like the laser ranging instruments- the LiDARs that small robots can hardly afford. While VBA are more power efficient, the process usually involves accurate, large-scale models of the environment, demanding significant processing and memory, which small robots cannot handle.
Counting Steps and Visual Breadcrumbs
To address these adversities, academicians have considered nature as a source of inspiration as the focus is often on insects. Ants and many other insects rely on odometry or the counting of steps and other movements of their own bodies, and ‘view memory’, or a low resolution, almost 360 degree visual field. As for the odometry processes, these are relatively easier to formulate even if they are not optimal, whereas as view memory is not entirely clear. A theory sometimes referred to as the “snapshot” theory posits that insects, for example ants, occasionally take snapshots. When near a snapshot location, they align the current view with a snapshot to return to that location, so they do not get lost using only odometry.
Snapshot-Based Navigation
These instructions are similar to Hansel’s advice in the fairytale of Hansel and Gretel when they got lost in the forest, hence the name – snapshot-based navigation. Hansel used stones to find his way back home, while birds ate the bread crumbs, misleading them. “In our case, the stones are the snapshots,” notes Tom van Dijk, the study’s first author. This approach relies on the robot being close to the snapshot location. If the snapshot and the robot’s environment differ too much, the robot may deviate from the correct path. Therefore, enough snapshots are needed, but not too many, as taking a large number of snapshots will significantly increase memory use. Prior methods placed snapshots very close together, allowing the robot to move smoothly from one snapshot to the next.
‘The main insight, which underlines the strategy is the fact that one can space the snapshots significantly further apart if the robot uses odometry to travel between the snapshots,’ said Guido de Croon, the Full Professor in bio-inspired drones and the author of the article. Homing will work, so long as the robot ends up within a fixed distance from the snapshot location, which is defined by the snapshot’s catchment area that can tolerate the robot’s odometry drift This doubles the robot range since the robot can fly much slower when homing in on a snapshot than when simply snapping from odometry from one snapshot to the next snapshot.
The insect like like flight plan allowed a 56 gram “CrazyFlie” quad copter equipped with an omnidirectional camera to travel distances of up to 100 meters with only 0. 65 kilobytes of memory. Every rendering was performed on a small scale processor, a micro-controller, which is normally used in most low-cost electronics.
Putting Robot Technology to Work
In this regard, the invention of the insect-inspired abstract navigation benchmark is a leap to employing pin-size robots in real life’, points out Guido de Croon. But this can be just enough to meet most of the requirement where a small degree of differentiation has to be made. For example in stock taking where Stocks are piled in a warehouse or in a crop evaluation where crops are grown in a compound, the drones could have taken off, work out and come back to a predefined ground station. They may store new data concerning the mission-related identification images on an SD card and copy them to a server for additional processing; however, such images were not necessary for navigation per se.
