CHANNELS

Aerospace

Automotive

Automated & Autonomous Vehicles

Battery & Electrification Technology

Defense

Drone TV

Electronics

Manufacturing & Prototyping

Materials

Medical

Motion Control

RF & Microwave Technology

Robotics & Automation

Sensors & IoT

Test & Measurement

Aerospace
Automotive
Automated & Autonomous Vehicles
Battery & Electrification Technology
Defense
Drone TV
Electronics
Manufacturing & Prototyping
Materials
Medical
Motion Control
RF & Microwave Technology
Robotics & Automation
Sensors & IoT
Test & Measurement
Aerospace & Defense
Search Results

Robotic Hummingbird Could Use Machine Learning to Go Where Drones Can't

Purdue University  researchers have created small flying robots that act like hummingbirds; they are trained by machine learning algorithms based on various techniques that the bird uses naturally every day. After learning from a simulation, the robot learns how to move around on its own like a hummingbird would. Artificial intelligence, combined with flexible flapping wings, also allows the robot to teach itself new tricks. Even though the robot can't see, it senses by touching surfaces, and can basically create a map without seeing its surroundings.

Topics:
Automation Motion Control Robotics Software

More From SAE Media Group

    Transcript

    00:00:00 The hummingbird is very unique, in the sense that they have the advantages of both the birds and insects. The birds can articulate their wing. They can move their wing actively, because they have muscle on their wing. But the hummingbird can hover. And most birds, they have to fly forward. So basically they have the advantages of both. First of all, you decode hummingbird flight, find out the aerodynamic principles. How do they generate lift? How do they do controls? What kind of wing kinematic patterns that make sense for your controls of your robot. The hummingbird inspired robot, these are like 12 grams, and the lift it can generate is up to 27 grams, which is very

    00:00:41 close to the size of a hummingbird. And wing beat frequency is also very similar, around 40 hertz. We do have smaller models, so this is less than 1 gram. 3 to 4 centimeters, wingtip to wingtip. Wing beat frequencies about 160 hertz. So of course the smaller you get, the higher the wing beat frequency. We use all different control algorithms, reinforcement learning, AI. So we train the robot. The robot can also behave similarly to the hummingbird. Control algorithm has to be robust and adaptive, say, if you have wind gust and external disturbances, or bumping into the wall, you can still deal with that. So the control algorithm itself should have some robustness built into that. We are not really shooting for long-range higher speed, like a conventional aircraft or conventional

    00:01:27 bigger vehicles. So for these bio-inspired devices, it's best suited for confined space, for maneuverability, turning and maneuvering some space which is inaccessible for bigger vehicles or humans. After an earthquake, you want to go into a collapsed building to find people or something. Search-and-rescue, surveillance. Because the wings are compliant, and also flexible, even if they flap at very high frequency, they don't get easily broke. And also because of the noise, hummingbird is pretty quiet. My students are also quite passionate and interested with this research. I mean, this research itself is naturally intriguing.