Artificial intelligence continues to provide innovations across the aeronautic and robotics industries. Recent developments involving integrated AI-powered strain receptors could make it possible to create more agile and lightweight options in the coming years. Here’s everything you need to know.
Since the dawn of time, man has looked towards nature to gain a better understanding of flight. However, creating robots that flap their wings to fly has proven to be much more default than traditional winged craft. Sadly, this scenario has left aerial robots without some key capabilities that their natural counterparts have such as quickly converting between hovering and optimized flight. Thankfully, this scenario may be about to change.
Flapping Wing Aerial Robots
Today, aerial robots are prevalent across multiple sectors, and their influence, capabilities, and availability are on the rise. People often think of only propellor-powered and winged craft when discussing aerial robotics. However, there are several other options that may not get the attention but definitely have unique benefits that make them stand out.
Flapping Wings
Flapping wings provide the best of both worlds. They enable birds to quickly gain vertical lift and stabilize to glide for long distances. Winged insects can hover and change direction quickly. Think of how a bumble bee or hummingbird zips around an area, or how a moth will circle a light bulb.
To date, there have been some major advancements in flapping-winged robot design. However, the flight controllers needed to make these craft operate reliably in changing conditions rather than in a lab has proven to be difficult to create. However, these designs still capture the imagination of developers and creators alike, with the movie Dune recently showcasing an Ornithopter that relies on flapping wings resembling a dragonfly.
Source – Fandom
AI-Powered Strain Receptors Study
A recent study, “Machine Learning-Based Wind Classification by Wing Deformation in Biomimetic Flapping Robots: Biomimetic Flexible Structures Improve Wind Sensing,” draws on natural inspiration to enhance the capabilities of flapping winged robots. Specifically, researchers examined multiple creatures to determine how their senses enable them to optimize their flight patterns accurately.
AI-Powered Strain Receptors Take Natural Inspiration.
The team noticed that all flapping winged birds and insects have some sort of sensory organ located in their wings. They conceived that this organ does different tasks in different animals which enables them to correct their flight characteristics to improve their results. The team noticed that grasshoppers had strain receptors located in the veins of their wings. Whereas many birds, like chickens, have sensors near their feather follicles.
Until this study, little understanding existed of exactly what data these sensors provided to the animal. However, researchers deduced that the sensory info allowed the animals to detect wind, body movements, and changing environmental conditions in real-time. Seeking to give robots the same capabilities, the team set off to create a reliable AI-powered strain sensor that could mimic its natural counterparts, enabling the robot to “feel” its environment and conditions, adjusting accordingly.
Wing Design
The team took inspiration from one of nature’s most agile flyers, the hummingbird. They set off to create Hummingbird-mimetic wings that have a similar structure to the bones found in the bird. The shafts taper at the ends and act as wing veins, adding another layer of stability to the wing’s structure.
These flexible wings were 3D printed utilizing a dual-nozzle fused deposition modeling 3D printer. This approach allowed the team to print using a 12.5 μm-thick copolyester polymer and carbon-fiber-reinforced polyethylene terephthalate. This approach provided the characteristics of a natural wing that could flex and move along its path.
Free Movement
Specifically, the wing could freely feather until an angle of ±23°. The wing would also twist along the leading edge during each flap. This movement provided additional power by maximizing the lift force, similar to insects. The engineers set the wings flapping amplitude to 158°, and the flapping frequency was adjusted to ≈12 Hz for the experiments.
Source – Advanced Intelligent Systems
AI-Powered Strain Receptors
The team integrated strain gauges within the hummingbird-like wing structure. Specifically, seven commercially available low-cost strain receptors that had base widths and lengths of 1.4 and 4.2 mm were glued to specific locations on the test wings. These sensors were then used to measure the wing’s pressure and strain across seven different wind directions. The directions used included 0°, 15°, 30°, 45°, 60°, 75°, and 90°.
Motor
In order to make the wings flap, a DC motor was attached. The motor utilized a Scotch yoke mechanism and reduction gears to provide realistic flapping motions. The device was set to 12 cycles per second, and sensory wires were run through connectors on the wings to a data register. Notably, the engineers used a TEXIO TECHNOLOGY device with a constant voltage power supply to ensure uniformity and measurability.
AI-Powered Strain Receptors Convolutional Neural Network Model
One of the main components of the experiment was the utilization of a convolutional neural network. This model allowed researchers to register, classify, and train a flight controller that was capable of making adjustments on the fly using data gathered from the strain sensors and matched against the CNN model.
The strain sensing data lets the machine learning algorithm classify wind conditions accurately. As part of the training, the sensor’s data was obtained to emulate hovering flight in a wind tunnel. Notably, 720 strain and phase datasets were obtained for each wind condition. This data was broken down into individual flaps of the wing.
AI-Powered Strain Receptors Test
The team began the testing phase by registering the sensor data of the wings with zero wind. The lack of airflow allowed the sensors to zero out and make accurate comparisons as conditions were enhanced. Also, the team tasted three different wings with the same strain gauge data and compared the results.
A magnetic rotary encoder was used to capture the wing state accurately during different conditions. The device sat directly on top of the wings, allowing for a resolution of 0.703° during the flapping phase. Interestingly, the team initiated the process by setting a single rotation of the encoder to a single flapping cycle.
Wind Tunnel
The wind tunnel was a crucial part of these experiments. It allowed the team to simulate hovering flight under gentle to harsh wind conditions. Specifically, eight alternating wind conditions were used in the testing phase. Each condition had 3 measurements taken during a single flap cycle.
AI-Powered Strain Receptors Test Results
The results of the study were impressive. The team was able to determine with 99% accuracy the wind conditions. Impressively, the determination only took a single flap and in some instances, as little as a 0.2 flapping cycle provided very accurate results. Additionally, the study found that the sensors closest to the wing shafts provided the fastest results.
Cycle Time Matters
The cycle time of each measurement made a major difference in the results. The team noticed that at under 0.2 cycles, data reliability fell sharply. However, at 0.2, the sensors achieved an 85% accuracy. This accuracy could be improved or reduced based on the number of sensors in the wing.
Biomimetic Wing Shaft Structures Improve AI-Powered Strain Receptors Results
The testing found that the wing shaft structure plays a vital role in data retrieval and accuracy. As such, the structured wings tested could determine wind conditions much faster than a non-structured test subject. This discovery led engineers to determine that enhancing wing structure and sensor placement could yield even more accuracy in the future.
AI-Powered Strain Receptors Benefits
There’s a long list of benefits that this study brings to the market. For one, it provided robotic engineers with a simple wing strain sensing capability that relied on commercially available off-the-shelf and affordable pieces. These low-cost and low-power-consuming devices are easily integrated into flying bots without the need to make any major changes.
Agility
The agility that bumble bees achieve is almost otherworldly. These winged animals can rapidly stop, hover, and change direction without much strain. Scientists hope to create drones with the same capabilities, enabling a new level of integration.
Adaptability
Nobody can tell you which way the wind will blow all the time. However, the sensory input of the wings tested can directly recognize flow conditions without the assistance of any additional devices. This data can be used to improve environmental awareness, providing better control and quick information encoding based on environmental conditions.
Simplistic Approach
Another major benefit of flapping wings versus other hovering technologies is simplicity. Hovercrafts require a lot of airflow and can only hit a certain height. Reversely, helicopters are massively complex, requiring thousands of moving parts to be calibrated perfectly to achieve a hover state. This latest study could make it possible to 3D print vehicle wings that are capable of stable hover and fast directional changes without tons of moving and intricate parts.
AI-Powered Strain Receptors Use Cases
There are several use cases for flapped-winged robots. These devices could help to reach hard-to-find locations or provide smooth scanning of natural disasters or warzones. Small aerial robots currently suffer from severe limitations in weight and size. The use of flapping wings could enhance their payload by reducing the weight needed for flight apparatuses.
AI-Powered Strain Receptors Researchers
This study was put forth by researchers at the Institute of Science Tokyo. The report was led by Associate Professor Hiroto Tanaka and included work from Hiroto Tanaka. Additionally, Tomoya Fujii helped with the wing design. Notably, the researchers received support from JSPS KAKENHI Grant-in-Aid for Scientific Research on Innovative Areas “Science of Soft Robots” under grant no. JP18H05468.
Companies that Could benefit from the AI Powered Strain Receptors
The ability to accurately and quickly determine wind conditions is an option that many companies could use to improve their offerings. The use of these sensors on flapped winged robots opens the door for drone makers to expand on this tech to create more agile and unique options. Here’s one company that can accomplish the task in the coming months.
Kratos Defense & Security Solutions Inc (KTOS +5.07%) originally entered the market in 1994 as a telecommunications infrastructure provider before switching its mission and goals over to drone manufacturing. The company is based out of San Diego CA.
Kratos Defense & Security Solutions, Inc. (KTOS +5.07%)
In 2004, Kratos Defense & Security Solutions Inc. began making high-level acquisitions across the market. These acquisitions provided the company access to advanced technologies and led to the firm shifting its name and overall focus over to military defense technologies.
Today, Kratos is recognized as a leading provider of military drones and software. The company’s stock, KTOS, has seen steady growth over the year due to a variety of factors including the company continually innovating its offerings alongside growing demand for automated and AI-powered warfare drones.
Kratos Defense & Security Solutions Inc.’s deep ties with institutional investors, governments, and its proven track record make it the perfect company to integrate this tech in the coming months.
Future of AI-Powered Strain Receptors
The engineers behind the AI-powered strain sensors study believe that there is much more work to be done to ensure this technology hits its peak. Currently, the winged drone sector is still a fledgling market.
However, as the advantages of winged flight, such as stable hovering and quick direction changes, create unique opportunities, you can expect more demand for these bots to appear. As such, the team intends to conduct further studies on more complex wind conditions and combinations of different strain-sensing locations in order to optimize their design.
AI-Powered Strain Receptors – Making Wings Smart
The introduction of reliable and affordable AI-powered strain receptors into winged robots is sure to enhance performance across the board. This study takes the industry one step closer to mimicking nature and unlocking age-old mysteries surrounding flight. In the coming months, this study could lead to the creation of many new and capable flapping winged craft.
Learn about other cool robotics projects here.
