In the longer term, tiny flying robots might be used to assist seek for survivors trapped under rubble after a devastating earthquake. Like real insects, these robots could dart through tight spaces that larger robots can't reach, while avoiding stationary obstacles and falling debris.
Until now, microrobots have only been capable of fly slowly within the air on smooth trajectories, a far cry from the fast, agile flight of real insects – until now.
MIT researchers have demonstrated aerial microrobots that may fly at speeds and agility comparable to their biological counterparts. A collaborative team designed a brand new AI-based controller for the robot beetle, allowing it to follow gymnastic trajectories, equivalent to performing continuous body flips.
Using a two-part control scheme that mixes high performance with computational efficiency, the robot's speed and acceleration increased by about 450 percent and 250 percent, respectively, in comparison with the researchers' best previous demonstrations.
The speedy robot was agile enough to finish ten consecutive somersaults in 11 seconds, even when wind disturbances threatened to blow it off track.
Image credit: Courtesy of Soft and Micro Robotics Laboratory
“We need to have the opportunity to make use of these robots in scenarios where traditional quad-copter robots would struggle, but where insects could navigate. With our bio-inspired control framework, our robot's flight performance is now comparable to that of insects when it comes to speed, acceleration and pitch angle. This is a reasonably exciting step towards this future goal,” says Kevin Chen, associate professor within the Department of Electrical Engineering and Computer Science (EECS) and head of the Soft and Micro Robotics Laboratory within the Research Laboratory for Electronics (RLE) and co-senior creator of a Paper concerning the robot.
Chen is joined on the paper by co-lead authors Yi-Hsuan Hsiao, an EECS-MIT graduate student. Andrea Tagliabue PhD '24; and Owen Matteson, a graduate student within the Department of Aeronautics and Astronautics (AeroAstro); and EECS doctoral student Suhan Kim; Tong Zhao MEng '23; and co-senior creator Jonathan P. How, Ford Professor of Engineering within the Department of Aerospace Engineering and principal investigator within the Laboratory for Information and Decision Systems (LIDS). The research appears today in .
An AI controller
Chen's group has been constructing robotic insects for greater than five years.
They recently developed a more durable version of their tiny robot, a microcassette-sized device that weighs lower than a paper clip. The new edition uses larger flapping wings that allow for more agile movements. They are powered by a series of soppy artificial muscles that flap their wings extremely quickly.
But the controller – the robot's “brain” that determines its position and tells it where to fly – was adjusted by hand by a human, limiting the robot's performance.
In order for the robot to fly quickly and aggressively like an actual insect, it needed a more robust controller that would account for uncertainty and perform complex optimizations quickly.
Such a controller can be too computationally intensive to make use of in real time, especially given the lightweight robot's complicated aerodynamics.
To overcome this challenge, Chen's group joined forces with How's team and co-developed a two-stage, AI-driven control scheme that gives the robustness required for complex, rapid maneuvers and the computational efficiency required for real-time operations.
“Hardware advances have pushed the controller forward, allowing us to do more on the software side, but similtaneously the controller has evolved there have also been more possibilities with the hardware. As Kevin's team demonstrates latest capabilities, we show we will make the most of them,” says How.
In step one, the team built a so-called model predictive controller. This kind of powerful controller uses a dynamic, mathematical model to predict the robot's behavior and plan the optimal series of actions to soundly follow a trajectory.
Although it’s computationally intensive, it might plan sophisticated maneuvers equivalent to air flips, fast turns and aggressive body tilts. This high-performance planner can be designed to take note of limitations on the force and torque the robot can apply, which is crucial for avoiding collisions.
For example, to perform multiple flips in a row, the robot would should decelerate in order that its initial conditions are excellent to perform the flip again.
“If little mistakes creep in and you are trying to repeat that flip 10 times with those little mistakes, the robot will just crash. We need robust flight controls,” says How.
They use this expert planner to coach a “policy” based on a deep learning model to manage the robot in real time through a process called imitation learning. A suggestion is the robot's decision engine that tells the robot where and how you can fly.
Essentially, the imitation learning process compresses the powerful controller right into a computationally efficient AI model that may run in a short time.
The key was having a wise strategy to create barely enough training data that might teach policymakers all the pieces they needed to know to make aggressive maneuvers.
“The robust training method is the key of this method,” ​​explains How.
The AI-driven policy uses robot positions as inputs and outputs for real-time control commands, equivalent to: B. Thrust and torques.
Insect-like performance
In their experiments, this two-stage approach allowed the insect-sized robot to fly 447 percent faster while achieving 255 percent acceleration. The robot was capable of complete ten somersaults in 11 seconds, and the tiny robot never deviated greater than 4 or 5 centimeters from its planned trajectory.
“This work shows that soft and microrobots, which have traditionally been limited in speed, can now use advanced control algorithms to realize agility near that of natural insects and bigger robots, opening latest possibilities for multimodal locomotion,” says Hsiao.
The researchers were also capable of detect saccadic movements, which occur when insects flap very aggressively, flying quickly to a certain position after which flapping in the opposite direction to stop. This rapid acceleration and deceleration helps insects locate themselves and see clearly.
“This biomimic flight behavior could help us in the longer term after we start installing cameras and sensors on board the robot,” says Chen.
Adding sensors and cameras to permit the microrobots to fly outdoors without being connected to a fancy motion capture system can be a crucial area of ​​future work.
The researchers also want to analyze how integrated sensors could help robots avoid collisions with one another or coordinate navigation.
“For the microrobotics community, I hope this paper signals a paradigm shift by showing that we will develop a brand new control architecture that’s concurrently powerful and efficient,” says Chen.
“This work is especially impressive because these robots still perform precise rotations and rapid rotations despite the big uncertainties that arise from relatively large manufacturing tolerances in small-batch production, wind gusts of greater than a meter per second, and even the ability cable wrapped across the robot during repeated rotations,” says Sarah Bergbreiter, a professor of mechanical engineering at Carnegie Mellon University, who was not involved on this work.
“Although the control currently runs on an external computer slightly than onboard the robot, the authors show that similar but less precise control policies is also possible with the more limited computing power of an insect-scale robot. This is exciting since it points to future insect-scale robots with agility near that of their biological counterparts,” she adds.
This research is funded partly by the National Science Foundation (NSF), the Office of Naval Research, the Air Force Office of Scientific Research, MathWorks, and the Zakhartchenko Fellowship.
