Real-World Transformers? Engineers Make Bots That Morph Mid-Air

The world of robotics is advancing at a quick pace. From worker robots to humanoid robots that are more dexterous and capable of complex movements, we are entering the new age of robotics. 

Just last month alone, researchers created interactive robots that act like therapy horses, responding to human emotions; a soft but smart robot that moves and grips objects by sensing its environment, much like an octopus; and a robotic dog that mimics mammals for superior mobility on both land and in water.

Engineers have even invented a self-healing muscle for robots that can detect injury, heal it, and then reset to detect future harm. The ability to control robots remotely while feeling the interaction at fingertips has also been introduced.

Amidst all this, a team of engineers has now developed an actual Transformer that can change its shape while in the air. This mid-air transformation enables the robot to roll away effortlessly and then initiate its operations on the ground without pause. 

With this capability, Caltech engineers have overcome the challenge with specialized robots that can both drive and fly without getting stuck in rough terrain. The enhanced flexibility of these robots can be especially beneficial for robotic explorers and deliveries.

Why Ground and Aerial Robots Struggle in Real-World Environments

Effective ground-aerial movement is crucial for a wide range of robotic applications; however, neither ground nor aerial robots are yet capable of operating reliably in the real world. 

While ground robots are limited by their range of operation, which makes it impossible for them to move across high obstacles or perform inspection tasks, aerial robots face the problem of limited battery performance due to payload requirements and safety concerns when flying in an urban setting.

These challenges that today’s autonomous systems face can be overcome by combining aerial and terrestrial capabilities. Hence, the focus of the Caltech team of engineers is on the development of ground-aerial robots.

The designs of many such robots tend to depend on the philosophy of redundancy and the use of multiple actuators to meet their bi-modal movement requirements.

These redundant robot designs, however, often end up using more actuators and components than are necessary, resulting in increased weight and cost.

Here, morphobots or robots that reuse the same appendages for different tasks via shape change can generate different modes of locomotion while decreasing both the system complexity and weight.

These kinds of robot designs often take inspiration from animals’ multi-functional locomotion behaviors and are expected to boost the efficacy of mobile autonomous robots that have to face changing, unstructured environments.

For instance, a study from Colorado State University researchers from a couple of years ago presented¹ an embedded shape-morphing scheme for morphologically adaptive robotic systems.

The researchers developed three robots that can morph their legs and bodies as needed to move through difficult terrain. These systems were designed to mimic the way biological organisms, such as frogs, adapt their shape depending on their environment and life cycle. To develop these robots, the researchers used materials that can become soft or rigid with changes in temperature and move without bulky power systems.

Its embedded morphing scheme utilized a lightweight artificial muscle, much like a human one, which contracts when electricity is applied, allowing the researchers to achieve a variety of shape types and making them more versatile and better equipped to navigate difficult environments. 

Recent research has used multi-functional appendages and body shape changes to improve movement, allowing maneuvers that were not possible before. But one ability of morphobots that has not been studied as much is their mid-air shape-shifting to improve both ground and aerial movement.

This can provide morphobots with the capability to bypass the need for ground-vehicle interaction during the transformation. 

The mid-air change can offer a reliable path to behavioral agility and mission safety in scenarios where ground morphing may not be possible due to rough terrain obstructing the robot’s appendages’ ground motion.

So, the Caltech engineers presented their study, which visualizes an aerial transition maneuver that connects flying and driving.

This maneuver is called the dynamic wheel landing, where the goal is to have a smooth transition from flying to driving by transforming near the ground and landing on dual-purpose wheel-thruster appendages with as close to a drive configuration as possible, which means the largest possible tilt angle, while accomplishing the desired impact velocity. 

Unlike the conventional quadrotor landing maneuvers, wherein the robot generally lands by vertical, non-transforming descent, the maneuver presented in the study involves morpho-transitioning, which means shifting between two modes via near-ground morphing.

But achieving this kind of maneuver isn’t an easy task; rather, it is a challenge from a design, modeling, and control perspective. 

Not only does the maneuver require increased torque to withstand the thrust forces consistently, but it also introduces new dynamic couplings between actuator limits and the robot’s degrees of freedom. Autonomous near-ground aerial operation is already a known challenging problem due to the effects of ground aerodynamics. On top of that, the aerodynamics of morphing flight and near-ground transformation are largely unknown. 

To address these challenges, Caltech researchers have designed the Aerially Transforming Morphobot (ATMO) specifically to solve the problem of mid-air transformation.

Inside ATMO: The Real-World Transformer Robot Explained

Published in the journal Communications Engineering, the study, supported by funding from the Center for Autonomous Systems and Technologies at Caltech, addresses the challenge of aerial transformation for Morphobots by designing a flying-driving robot called ATMO.

This robot is specialized for mid-air transformation through a morphing mechanism that enables changing the body shape in mid-flight while requiring minimal actuation.

It uses four thrusters to fly while the shrouds that protect the thrusters become the wheels of the system in an alternative driving configuration. This entire transformation relies on a single motor to move a central joint, which pushes the thrusters up into drone mode or down into drive mode.

The new robotic system is inspired by nature, with lead author Ioannis Mandralis, a graduate student in aerospace at Caltech, illustrating how birds fly and adjust their body morphology to slow themselves down and avoid obstacles. 

“Having the ability to transform in the air unlocks a lot of possibilities for improved autonomy and robustness.”

– Mandralis

And while seeing a bird land and run seems pretty simple, it isn’t.

“In reality this is a problem that the aerospace industry has been struggling to deal with for probably more than 50 years,” said Mory Gharib, the Hans W. Liepmann Professor of Aeronautics and Medical Engineering and director and Booth-Kresa Leadership Chair of Caltech’s Center for Autonomous Systems and Technologies (CAST), where researchers collaborate work on advancing drone research, autonomous exploration, and bio-inspired systems.

All flying vehicles have to deal with complicated forces close to the ground. 

In the case of helicopters, when they come in for a landing, their thrusters push lots of air downward. Here, lift and thrust are supplied by the spinning rotors. As the airflow hits the ground, some of it circulates back up. So, if the chopper descends too quickly, it can get sucked into this vortex of air and lose its lift.

When it comes to ATMO, things become even more complicated because it must contend with near-ground forces while having four jets that continuously change the extent to which they are shooting toward each other. This creates more turbulence and, in turn, instability.

In order to get a better understanding of the aerodynamic force, the engineers conducted experiments in the drone lab of CAST.

To investigate how altering the robot’s configuration during landing affects its thrust force, the team conducted load cell experiments, which involve measuring the force applied to an object using a load cell, a device that converts mechanical force into an electrical signal.

The researchers also conducted smoke visualization experiments, which are used to make airflow patterns visible, to discover the underlying situation that leads to these changes in dynamics.

Once gathered, the insights were then fed into the algorithm behind the new control system that the researchers created for ATOM.

This system utilizes an advanced control technique called model predictive control, which constantly predicts the way the system will behave in the near future and then adjusts its actions to stay on track.

According to Mandralis:

“The control algorithm is the biggest innovation in this paper. Quadrotors use particular controllers because of how their thrusters are placed and how they fly. Here we introduce a dynamic system that hasn’t been studied before. As soon as the robot starts morphing, you get different dynamic couplings — different forces interacting with one another. And the control system has to be able to respond quickly to all of that.”

Testing ATMO: How Engineers Validated the Mid-Air Transformation

The ATMO from Caltech engineers has accomplished both driving and flying using the dual-purpose appendages through shape-shifting. But what makes ATMO different from other such kinds of robots is the ‘self-locking tilt actuator mechanism’ that allows for transformation mid-air with a simpler design, lower cost, and minimal actuation requirements.

When in flight mode, the robot is configured as a standard quadcopter and uses its wheel-thruster attachments for propulsion. In drive mode, these very same appendages are reused for wheeled locomotion. 

The resulting compact robot has a total weight of 5.5 kg, which also includes the battery. As for its dimensions, the robot stands 33 cm tall and 30 cm wide in ground configuration and 16 cm tall and 65 cm wide in aerial configuration.

For driving, ATMO uses two belt-pulley systems located on either side, which are operated by driving motors, enabling differential drive steering.

In addition to having a computer onboard that runs a custom controller, the robot is also equipped with onboard sensors for state estimation and fusion. All of the communication is accomplished through the advanced software ROS2. 

To validate the system, the controller was applied to a dynamic wheel landing in the CAST flight area using a motion capture system to enable state estimation.

In this experiment, the controller was used to track a reference trajectory in space that comprised a descent with some forward motion while angling the wheel-thrusters, landing on the wheels, and then driving forward.

The model-based control scheme is developed to cover the full operational package of flying, driving, and transitioning. To address the actuator saturation problem that occurs as the robot tilts its thrusters to land on wheels, the team “used a decomposition of the control objective function into a convex combination of specialized objective functions for each locomotion mode.”

This provided a flexible framework for controlling the systems during the transition from ground to air.

The developed controller enabled landings with tilt angles past the actuator saturation limits. This allows the new robot to clear bumpy terrain. 

With a final angle of tilt at landing of 65°, the robot demonstrated that it’s successfully able to land at a tilt angle that exceeds the critical angle. This, the study noted, is achieved because of the change in the cost function during the transition phase, and as a result, ATMO can continue to tilt its wheel-thrusters while maintaining the desired attitude. 

To validate the control method, the team performed a driving takeoff, which was followed by a dynamic wheel landing.

They also showcased an important use case of mid-air transformation, an inverse maneuver consisting of rapid takeoff along with forward driving movement, in addition to landing on a slope.

In the experiment, ATMO was able to land smoothly on a slope of known height and position, which may be dangerous due to the risk of tipping over, and can be avoided by transforming prior to landing, and continuing driving. 

Overall, experimentally validating the functioning and viability of these robots shows that “using mid-air robotic transformation can result in dynamic ground-aerial transition maneuvers that enhance robot agility and expand operational range – paving the way for greater autonomy in future mobile robotic missions,” noted the study.

While the team has successfully demonstrated dynamic transition maneuvers, the conditions here were controlled to facilitate rapid development. For instance, a motion capture camera system was used to accurately and rapidly estimate the position and orientation of the robotic system, surpassing what can be achieved by existing onboard sensors.

So, further investigation is needed to determine how these maneuvers work in the real world, where robots have to face more complex, unstructured terrain and make decisions based on partial sensor information, which is subject to noise.

 Investing in Robotics: Why Amazon (AMZN) Stands Out

When it comes to a prominent name in the robotics industry, the e-commerce giant Amazon (AMZN ) has been making a lot of advances here. To lead robotics, Amazon first acquired Kiva Systems in 2012 for $775 million, which was later rebranded to Amazon Robotics LLC. The company then unveiled its first-ever autonomous mobile robot (AMR) called Proteus in 2022.

Amazon (AMZN ) 

As of May 2025, Amazon reports having more than 750,000 robots deployed across its operations that sort, lift, and carry packages. 

“Years of innovation have allowed us to build, test, and scale this unique, highly integrated suite of robotics systems that work to support employees fulfilling customer orders.”

– Scott Dresser, Vice President of Amazon Robotics

According to him, advancements in AI have allowed for their seamless integration, which drives an estimated 25% productivity improvement at its fulfillment facilities.

There are a total of nine robots. This includes Proteus, Amazon’s proprietary autonomous mobile robot designed to work around people using sensors and a mix of AI-based and ML systems. 

Robin is a robotic arm that is responsible for sorting packages and has successfully finished more than three billion package moves. Yet another robotic arm is Cardinal, which puts packages into carts. Sparrow is also a robotic arm that picks up and moves individual items.

Sequoia uses robotics, AI, and computer vision systems to consolidate inventory. Hercules finds and brings pods of items to employees, with Titan also tasked to do the same, but with the capability to lift twice as much as Hercules. Then there’s Vulcan, which is Amazon’s first robot with a sense of touch that works alongside employees. 

Furthermore, a variety of packaging innovation systems are used to pack customer orders, with a Packaging Automation machine used to create made-to-fit paper bags.

Amazon now boasts a market cap of $2.18 trillion, with its shares trading at $205.8 as of this writing, down 6.24% year-to-date. It has an EPS (TTM) of 6.13, a P/E (TTM) of 33.55, and an ROE (TTM) of 25.24%.

As for financials, Amazon reported net sales of $155.7 billion in the first quarter ended March 31, 2025. Sales increased 8% YoY in North America to $92.9 billion and 5% YoY internationally to $33.5 billion.

For this period, Amazon reported an operating income of $18.4 billion, a net income of $17.1 billion or $1.59 per diluted share, and an operating cash flow of $113.9 billion. The company’s free cash flow decreased to $25.9 billion.

“We’re pleased with the start to 2025, especially our pace of innovation and progress in continuing to improve customer experiences,” said CEO Andy Jassy, who noted next-gen Alexa (Alexa+) becoming “meaningfully smarter” as well as more capable, new Trainium2 chips and Bedrock model expansion making it easier for AWS customers to train models and run inference cost-effectively, and first Project Kuiper satellites successfully launching into low earth orbit to provide masses broadband access.

Click here for a list of top robotics companies.

Latest Amazon (AMZN) Stock News and Developments

Conclusion: Why ATMO Marks a New Era in Robotics

The world of robotics is harnessing bio-inspired engineering, mid-air transformation, and intelligent control systems to design ground-aerial robots that have been challenging due to the increased actuation demands, which can add weight and reduce the efficiency of their locomotion.

Caltech engineers have achieved this through ATMO, a robot that transforms near the ground with a smooth transition between aerial and ground modes by leveraging the near-ground aerodynamics and stabilizing the system using a model-predictive controller.

ATMO marks a key step in bridging the gap between aerial and terrestrial mobility, which is validated through numerous experimental demonstrations. With its real-world transforming capabilities, the robot showcases huge potential in redefining autonomous operations across industries and paving the way for more agile, resilient, and adaptive machines!

Click here to learn how robots can take a cue from nature.

Studies Referenced:

^{1. Sun, J., Lerner, E., Tighe, B., Middlemist, C., & Zhao, J. (2023). Embedded shape morphing for morphologically adaptive robots. Nature Communications, 14(1), 6023. https://doi.org/10.1038/s41467-023-41708-6}