Navigating Like Ants
We know insects are not especially smart in the human sense. However, despite their limitations, they are able to perform remarkable feats of organization and orientation. This makes them an inspiration for researchers who are working on microrobots and lightweight drones, which are dealing with similar constraints.
For example, the desert ant Cataglyphis can forage over long distances and then walk straight back to its nest, with journey lengths of up to 1 km.
The low carrying capacity, as well as the power available, mean that solutions used by heavier autonomous systems like self-driving cars cannot be used. For example, LIDAR (“laser radar”) is great for creating 3D maps of the environment, but they are too heavy and power-hungry. They also require a lot of computing, itself requiring a lot of memory and processing which are power-hungry and heavy.
Beacons and GPS signals are alternatives, but they require expensive setups, can be unreliable, or are outright impossible. So, understanding how insects like ants and bees manage to navigate the world with only the most limited “hardware” and energy supply could help us replicate it with our own creations for robots and drones.
This is the general idea of using bio-inspired robots, a topic we explored further with octopus, salamander, snake, and dog-inspired robots in our article “How Robotics Can Take a Cue From Nature”.
Why Using Microrobots?
Smaller robots and drones are cheaper to build and can cover more surfaces at once for the same cost. Being smaller, they can also observe things in more detail, without risking colliding with their environment. For example, they fly inside a greenhouse and scan for early signs of diseases or pests on the plants.
Or they can be deployed for search and rescue missions, investigating ruins or wilderness for people needing help. Such swarms of robotic “birds/ants/dragonflies” could quickly detect survivors after an earthquake, for example.
Source: TU Delft
How Ants Navigate The World
A method is using vision, something insects are especially good at as they have an almost omnidirectional visual system (seeing in all directions at once). However, this vision has a relatively low resolution.
Some of the most ancient and established theories on how insects use vision to orient themselves are in the “snapshot model.”
The idea is that the insect brain takes snapshots of its environment regularly. When it needs to navigate back to “home”, it compares the current environment to recently stored snapshots.
This concept is now well understood, as far as at the neuronal level. So it could be relatively easily replicated in robots.
In theory, this method alone could be enough. But in practice, it suffers from a few limitations:
- To work well, it requires a very tight series of snapshots, with even one missing data leading to disorientation and potentially the robot getting completely lost.
- As it requires a lot of snapshots, it would prove overwhelming for both ant brains and robots’ memory.
Adding Odometry
Another method used by ants, and insects in general, is keeping track of their movement, a method called odometry. This is a method also used in robotics, but the problem is the lack of precision. Each step is estimated from motion sensors (or subjective perception in the case of ants), but never perfectly reflects the real movement.
This led to a progressive drift in the accuracy of the odometry-based estimate of the current location, becoming progressively more inaccurate over time.
Combining the 2 methods was the key insight for researchers at the University of TU Delft, Netherlands. In a scientific paper called “Visual route following for tiny autonomous robots”, they combined both visual snapshots with odometry to increase microrobots’ autonomy.
Greater Performance
This allowed the robot to regularly reset the odometry drift whenever it found back one of its landmark snapshots.
Source: Science Robotics
At the same time, relying mostly on odometry reduces the demand for ultra-close snapshots, giving the microrobots the ability to go quicker between points, without having to check all the time for visual clues of its trajectory.
“The main insight underlying our strategy is that you can space snapshots much further apart if the robot travels between snapshots based on odometry.
Homing will work as long as the robot ends up close enough to the snapshot location, i.e., as long as the robot’s odometry drift falls within the snapshot’s catchment area.”
Professor Guido de Croon.
The research team used their new orientation software combining snapshot and odometry to test how little data could be used to orient a robot weighing only 56g over 100m.
Source: Science Robotics
It can be at an extremely low size, only 1.16 kilobytes. For reference, an average image taken by a smartphone will be in the thousands of kilobytes per picture, and most online images are in the tens or hundreds of kilobytes.
Even better, all the image processing could be performed by a lightweight mini-computer called a “micro-controller”, which can be found in many cheap electronic devices.
Applications
Industry
Such microrobots and drones will be very limited in their data processing capacity, with most of the onboard micro-controller processing power busy managing the navigation and data collection.
Such drones could however be used to track inventory in warehouses or monitor crops in greenhouses. It would work by having them walk or fly around and collect data like pictures, code bars, or RFID tags. This data point can be saved on a small SD card.
These recordings would then be transferred to a larger computer or server that could post-process and translate them into useful data.
Military
Another likely field of application could be military technology, especially considering the growing importance of drones on the modern battlefield, as illustrated by the war in Ukraine.
Small flying drones that are light enough to fit in an infantryman pack could be sent ahead for reconnaissance and bring back images of enemy positions to sheltered soldiers.
Because the area would likely be heavily jammed by electronic warfare (EW) and in ever-changing areas, autonomous navigation of the drones will be a must. Lightweight and low power consumption are likely also going to be key features. In the research discussed here, a drone was able to navigate a trajectory of 300 m in a simulated forest environment.
Source: Flir
Further Research
The strategy of combining odometry and snapshots is very efficient and can be made even more efficient by improving the precision of the odometer. The algorithm used can also likely be tweaked to be even more memory power efficient.
Another improvement would be to add to the robot’s collision-avoidance capabilities, especially as it already has an omnidirectional vision.
A solution will still need to be found for when somehow the robot still gets lost. For example, the researchers propose that “the robot could estimate the catchment area size online and be endowed with a search procedure when losing the route“.
This procedure is especially fit for small robots that are usually struggling with navigation using other methods. But it could start to be applied to larger robots as well, in a bit to reduce the need for expensive equipment like LIDAR, and reduce computing and power requirements.
Drones & Robot Companies
1. AutoStore Holdings Ltd. (AUTO.OL)
Autonomous vehicles like self-driving cars might be around the corner, but they have been a difficult technology to develop, even for tech leaders like Google and Tesla. But there is a sector that is already being revolutionized by autonomous driving and robotics: logistics.
The Norwegian AutoStore provides automated warehouses for industries as diverse as pharmaceutics, clothing, groceries, aviation, logistics, or industrial manufacturers. Apparel, industrial, and third-party logistic companies make up the three largest segments of AutoStore’s business.
The company’s warehouses rely on autonomous robots that can autonomously identify and pick up the parcels or products and carry them to where they should go. You can see them in action in this video:
The company is quickly expanding, thanks to more and more major companies realizing the advantage of creating more efficient, resilient, and quicker logistical systems after the pandemic. On average, it takes only 1-3 years for the upgrading to autonomous warehouses to pay back the initial investment.
AutoStore is active in 50 countries, operating 58,500 robots for 900 different customers. It grew its revenues by 50% CAGR since 2017. This is 2-3x quicker than the automated warehouse market yearly growth, estimated at 15%.
Source: AutoStore
Like many European tech companies, AutoStore provides very advanced solutions that are also somewhat invisible to the greater public.
Most warehouses will move toward automation. Leaders in this sector are likely to outperform the sector growth, as it makes sense to rely on the provider able to deploy these solutions at scale and at a cheaper price.
Robots that are more autonomous and more efficient at finding their way could be both an opportunity and a threat for AutoStore. Currently, you need to fully redesign a warehouse to use the company’s robotic solutions.
In the future, the robots could be able to find their way without the need for the currently used grid, making adoption much easier, less disruptive to ongoing operations, and the initial investment much smaller, resolving what are still the main obstacles to the mass adoption of the technology.
Source: Autostore
2. Zebra Technologies Corporation (ZBRA)
Zebra Technologies produces tracking labels and scanners that allow for the monitoring of every component of a “smart” factory. It includes mobile computers, barcode scanners, machine vision, location technology, tags, and RFID (Radio-frequency identification).
This level of data gathering and analysis is a key component in implementing robotics, especially of the more mobile and flexible kind, out of an assembly line.
The company has been at the origin of the code bar popularization and has since 2018 been on an acquisition spree to bring together all the technologies required for the “robotization” and digitalization of modern warehouses & factories.
Source: Zebra
Currently, the company’s main segments are e-commerce & retail and transportation/logistics, followed by manufacturing.
Source: Zebra
As robots are becoming increasingly the center of e-commerce and logistics, Zebra tracking systems are getting more in demand.
Until now, it is still required to prepare the place for relatively large robots.
If a microrobot weighing only a few dozen grams can now go around and scan RFID tags, we could soon see continuous monitoring of all the activities in a factory floor or a warehouse to be handled autonomously, through a hive of bee-like drones.