ATMO Drone Shifts Shape Mid-Flight for Perfect Landings
ATMO Drone Shifts Shape Mid-Flight for Perfect Landings


Mimicking the flight of a bird, ATMO is a drone-like robot that can transform mid-air to move without a hitch across the ground after it lands.
Taking their cues from nature, engineers from the California Institute of Technology (CalTech) have developed a drone-like flying robot that can morph in mid-air, enabling it to go from flying to landing and then rolling seamlessly across the ground.
The team was motivated to create ATMO—the aerially transforming robot—by the well-publicized misadventures of NASA’s Spirit Rover, which got stuck on the surface of Mars in 2009. The robot spun its wheels in the sandy terrain and only dug itself deeper.
“They had a rover, and they had a helicopter,” said Morteza “Mory” Gharib, director of CalTech’s Graduate Aerospace Laboratories, who is overseeing the project. “These are two separate systems...We got the idea of ‘Why can’t the rover also be a helicopter and the helicopter be a rover?’ That motivated us to start thinking about this multi-modal drone that would be a helicopter that could land and drive.”
While specialized robots that could fly and drive on land already exist, they typically won’t transform until landing time approaches. ATMO mimics the ways certain birds transition from flying to touching down on land and smoothly moving across the terrain.
“Birds fly with their wings, but they can also tuck them under their bodies to slow and get through obstacles as they approach the ground. We wanted to use the idea of changing body type or adapting the morphology of a robot to achieve multiple tasks, in this case, flying and rolling on the ground,” said Ioannis Mandralis, a graduate student in aerospace at Caltech who is part of a team that developed ATMO and lead author of the journal Communications Engineering.
ATMO is powered by four thrusters for flight, featuring protective shrouds that convert to wheels in alternate driving mode. A single motor drives a central joint that transitions the thrusters—raising them for drone mode or lowering them for ground travel.
Creating a robot capable of transforming mid-air posed a significant engineering challenge. Gharib cited helicopters as an example: they use thrusters to direct air downward during landing. When the air strikes the ground, it rebounds upward. If the helicopter descends too quickly, it can create a vortex that traps the vehicle.
Over three and a half years engineers developed ATMO by exploring the aerodynamics of mid-air transformation. At the university’s drone lab, they conducted load cell testing to examine how altering the robot’s configuration during landing affected its thrust force.
To further investigate the underlying dynamics, they also performed smoke visualization tests to reveal the flow phenomena driving these changes.
“We observed how the smoke moved through the propellers at different configurations,” Mandralis said. “We found that as the robot is morphing mid-flight, the way that the smoke moves through changes significantly—which changed our dynamics significantly. So, we had to adapt the controllers to take that into account.”
The researchers incorporated those insights into the algorithm powering a newly developed control system for ATMO. This system employs an advanced technique known as model predictive control, which continuously forecasts the system’s future behavior and adjusts its actions accordingly to remain on course. Using a model predictive controller with an adaptive objective function, they also evaluated aerodynamic properties.
“Our method is applicable for emergency slope landings as well as for quick transitions from air to ground by landing or taking off with forward velocity,” Gharib said.
Software developed by the team is the brains behind ATMO’s ability to make smooth air to ground transitions.
Six versions of ATMO were developed over the years and researchers are working on a version that is able to carry heavier loads of eight pounds.
The engineers envision ATMO being used in various applications, including safer package delivery, general exploration, and search-and-rescue missions in disaster zones. In such scenarios, ATMO could autonomously map the terrain, identify suitable landing spots, and swiftly navigate to reach victims in need. The newest version of ATMO can climb up to five steps.
The robot’s cost is currently about $20,000 to $25,000, however, Gharib wants to bring that pricing down to $5,000.
Moving forward, Mandralis sees multiple versions of ATMO being developed.
“I don’t think there is going to be one ideal one,” he said. “There are probably going to be different versions depending on what a person’s needs would be.”
Annemarie Mannion is a technology writer in Chicago.
The team was motivated to create ATMO—the aerially transforming robot—by the well-publicized misadventures of NASA’s Spirit Rover, which got stuck on the surface of Mars in 2009. The robot spun its wheels in the sandy terrain and only dug itself deeper.
“They had a rover, and they had a helicopter,” said Morteza “Mory” Gharib, director of CalTech’s Graduate Aerospace Laboratories, who is overseeing the project. “These are two separate systems...We got the idea of ‘Why can’t the rover also be a helicopter and the helicopter be a rover?’ That motivated us to start thinking about this multi-modal drone that would be a helicopter that could land and drive.”
While specialized robots that could fly and drive on land already exist, they typically won’t transform until landing time approaches. ATMO mimics the ways certain birds transition from flying to touching down on land and smoothly moving across the terrain.
“Birds fly with their wings, but they can also tuck them under their bodies to slow and get through obstacles as they approach the ground. We wanted to use the idea of changing body type or adapting the morphology of a robot to achieve multiple tasks, in this case, flying and rolling on the ground,” said Ioannis Mandralis, a graduate student in aerospace at Caltech who is part of a team that developed ATMO and lead author of the journal Communications Engineering.
ATMO is powered by four thrusters for flight, featuring protective shrouds that convert to wheels in alternate driving mode. A single motor drives a central joint that transitions the thrusters—raising them for drone mode or lowering them for ground travel.
Engineering a multi-modal drone
Creating a robot capable of transforming mid-air posed a significant engineering challenge. Gharib cited helicopters as an example: they use thrusters to direct air downward during landing. When the air strikes the ground, it rebounds upward. If the helicopter descends too quickly, it can create a vortex that traps the vehicle.
Over three and a half years engineers developed ATMO by exploring the aerodynamics of mid-air transformation. At the university’s drone lab, they conducted load cell testing to examine how altering the robot’s configuration during landing affected its thrust force.
To further investigate the underlying dynamics, they also performed smoke visualization tests to reveal the flow phenomena driving these changes.
“We observed how the smoke moved through the propellers at different configurations,” Mandralis said. “We found that as the robot is morphing mid-flight, the way that the smoke moves through changes significantly—which changed our dynamics significantly. So, we had to adapt the controllers to take that into account.”
The researchers incorporated those insights into the algorithm powering a newly developed control system for ATMO. This system employs an advanced technique known as model predictive control, which continuously forecasts the system’s future behavior and adjusts its actions accordingly to remain on course. Using a model predictive controller with an adaptive objective function, they also evaluated aerodynamic properties.
“Our method is applicable for emergency slope landings as well as for quick transitions from air to ground by landing or taking off with forward velocity,” Gharib said.
Software developed by the team is the brains behind ATMO’s ability to make smooth air to ground transitions.
Six versions of ATMO were developed over the years and researchers are working on a version that is able to carry heavier loads of eight pounds.
The engineers envision ATMO being used in various applications, including safer package delivery, general exploration, and search-and-rescue missions in disaster zones. In such scenarios, ATMO could autonomously map the terrain, identify suitable landing spots, and swiftly navigate to reach victims in need. The newest version of ATMO can climb up to five steps.
The robot’s cost is currently about $20,000 to $25,000, however, Gharib wants to bring that pricing down to $5,000.
Moving forward, Mandralis sees multiple versions of ATMO being developed.
“I don’t think there is going to be one ideal one,” he said. “There are probably going to be different versions depending on what a person’s needs would be.”
Annemarie Mannion is a technology writer in Chicago.

