Modern Day Da Vinci Designs Smart Spider

It’s “A Bug’s Life” meets “WALL-E,” except this other-worldly creation is no product of Pixar Animation. It’s the real deal. This six-legged robot is fully equipped with Artificial Intelligence, it can crawl creepily like a spider on its own, or bust into syncopated flamenco dance moves.

modern-davinci_thoven_1.jpg

This hexapod replica, built by Matt Bunting for Intel, uses computer vision and machine learning to figure out its next moves.

It even has a tantalized Twitter following @hexapod, where fans can track its whereabouts.

With 21 motors and a Web camera for an eye, this skateboard-sized robot is fast moving and fully aware of its surroundings. Which gives many people the heebie-jeebies at first sight, according to creator Matt Bunting. Especially for anyone who is afraid of spiders. Aside from arachnophobes, Bunting said the hexapod is a real attention-grabber that often peaks people’s curiosity.

“When people first see the hexapod, their instinct is to just wave their hand at it, but it can’t recognize waving,” said Bunting, adding that it can quickly find and track faces.

“It’s very cool to see a little kid walk up and stick his face into the hexapod, said Bunting, raising his eyebrows. “The hexapod looks right back and follows along as the kid moves a little bit to the right or left. It’s a very chilling, unearthly feeling to have this creepy looking robot follow your own movements.”

“Creepy and cute and the same time,” he added.

Bunting tinkered around using very inexpensive parts to build his first hexapod when he was in high school. A few years later, in 2009, he evolved his earlier ideas into a fully functioning hexapod while in a cognitive robotics class during his undergraduate studies at the University of Arizona.

“I liked the idea of having a machine being able to learn,” said Bunting. “So I built a hexapod that was much more capable than the first one I built in high school. I didn’t want to pre-program it with algorithms, so I didn’t use inverse kinematics and force all of the legs to move in the way that I wanted them to, but instead I used this reinforcement learning technique called cue-learning. With cue learning the hexapod was able to experiment in the world and figure out an optimal way to walk.”

After the hexapod could walk, Bunting used a Playstation 3 game controller that he hacked and made the robot dance for a video he posted on YouTube.

Since then, Bunting’s work has captured the attention of technology companies, including Intel and HRL Laboratories, his current employer and graduate school sponsor. His hexapod currently graces the cover of the June edition of Linux Journal, and the electronics and engineering industry publication, EETimes, just honored Bunting with a prestigious ACE Award, naming him this year’s Top Student “whose discipline, hard work and academic success are considered hallmarks for other engineering or science students.”

modern-davinci_thoven_2.jpg

After building his first hexapod in high school, Matt Bunting reinvented his robot design as an undergraduate at the University of Arizona, where he’s now pursuing a PhD. in Electrical and Computer Engineering and working on DARPA’s “Cheetah Project.”

Two days after Bunting posted his first video on YouTube, back in 2009, he was contacted by Stewart Christie from Intel’s embedded and communications group. Christie’s job is to find the next generation of developers, and sometimes that means meeting robot builders in their gear-cluttered garages or taking requests for equipment or project sponsorship on his team’s website.

Christie calls Bunting “a real modern-day Da Vinci,” pointing out that Bunting even created the Spanish guitar music that compels his spider robot to dance, as seen in the YouTube video.

“My job is to look for hobbyists, hackers and future customers who are working on cool projects and who might not be aware that they can build on Intel architecture,” he said.

“When Intel first contacted me I was a little bit shocked,” said Bunting. “I didn’t believe it at first … because you don’t normally think this kind of thing is going to happen to you … that somebody would recognize you for the work that you do.”

Soon after, the hexapod began to multiply as Bunting worked to meet Intel’s request for two replicas.

“With the help of Intel, I really worked hard to make it look the way I always wanted it to look, said Bunting. “I created an even better machine-learning technique, which involved an artificial neural network, and I used a generic algorithm to tune all of the weights.”

Bunting refers to his works as “biologically influenced robotics,” in which he combines the ability to see, learn and respond with movements.

Early on, Bunting said he wanted all of the processing to be done onboard, controlling all of the 21 motors, and processing vision with advanced machine learning techniques. Each of the six legs would be powered by three servos, or smart motors, and three more would control the camera.

“I needed something that was very powerful,” he said. “I also wanted something that would eventually run on batteries, so it needed to be powerful and computationally efficient at the same time. Fortunately, CompuLab came out with the Fit PC right around that time, so I ordered one of those and it has an onboard Intel Atom processor (running at 1.6 GHz). It could handle vision processing seamlessly along with all of the kinematic equations to operate all of the motors.”

Bunting used OpenCV, or Open Computer Vision, which is a library of free software for real-time computer vision.

modern-davinci_thoven_3.jpg

Matt Bunting has been described as a “real modern day Da Vinci”

“I’ve implemented a few vision computing techniques such as object recognition,”” said Bunting. “There’s face tracking onboard, so the hexapod can find your face and actually follow it around. For the machine learning, like the reinforcement techniques that I implemented, vision was used to measure the optic flow. Using the optic flow, it’s able to take two successive images and figure out how the objects are moving around in the world. The hexapod could tell if a person’s head has tilted, moved forward, up or down and that was all critical for machine learning.”

Currently a graduate student at the University of Arizona, Bunting spends much of his time coding and tweaking his creations inside the windowless Robotics and Neural Systems’ lab led by UofA Professor Tony Lewis. Inspired at a young age, and even today, by building Lego models and mathematics, Bunting has evolved into a valuable asset for the Tucson school’s Electrical and Computer Engineering Department, even helping to attract new students and big-time research projects like DARPA’s “Cheetah.”

According to Bunting, his team will collaborate with other researchers in a competition to create a four-legged robot that could eventually lead to a two-legged robot able to run faster than any robot today. Proof of his team’s prototype rests in his iPhone in the form of a photo showing robotic hind legs of a small cheetah.

“If we scale it up, I think we can actually achieve 70 mph,” Bunting recently told Arizona Engineer, a publication of the university’s College of Engineering.

Bunting said his hexapods are coming in handy as he focuses more time on DARPA’s Cheetah project. He said he’s extending the “machine learning” that he built into the six-legged robot, and refining the 3-D printer techniques he used to fabricate the light, hard plastic skeleton and legs of the hexapod.

In addition to the hexapod he has inside his lab, Bunting keeps in touch with his other hexapods through Twitter, where Intel’s Christie shares the robots’ whereabouts like elementary school science class visits, technology tradeshows, media events such as the recent Tech Heaven in New York, and competitions like Robocup in Rome.

Christie said that someday a robot like this could be programmed to interpret facial gestures and human expressions, which could make it useful in such places as school and hospitals.

“This hexapod is surprisingly easy to command using a Bluetooth game controller,” said Christie. “I can hand the controller to anyone, and they’re off and running, especially kids who are used to playing with the Playstation.”

The collaboration continues. “Now I’m encouraging Matt to train the hexapod to post photos to its Twitter account so people can see what the robot sees when it’s in action at an event,” said Christie.

Christie also pointed out that several other teams have developed ancillary projects, using the same Atom-based hardware that Bunting used, including a 10-foot wingspan autonomous aircraft and an underwater robot.

“They all cited Matt’s success as the reason for choosing the Atom processor,” Christie said.

Modern Day Da Vinci Designs Smart Spider

It’s “A Bug’s Life” meets “WALL-E,” except this other-worldly creation is no product of Pixar Animation. It’s the real deal. This six-legged robot is fully equipped with Artificial Intelligence, it can crawl creepily like a spider on its own, or bust into syncopated flamenco dance moves.

modern-davinci_thoven_1.jpg

This hexapod replica, built by Matt Bunting for Intel, uses computer vision and machine learning to figure out its next moves.

It even has a tantalized Twitter following @hexapod, where fans can track its whereabouts.

With 21 motors and a Web camera for an eye, this skateboard-sized robot is fast moving and fully aware of its surroundings. Which gives many people the heebie-jeebies at first sight, according to creator Matt Bunting. Especially for anyone who is afraid of spiders. Aside from arachnophobes, Bunting said the hexapod is a real attention-grabber that often peaks people’s curiosity.

“When people first see the hexapod, their instinct is to just wave their hand at it, but it can’t recognize waving,” said Bunting, adding that it can quickly find and track faces.

“It’s very cool to see a little kid walk up and stick his face into the hexapod, said Bunting, raising his eyebrows. “The hexapod looks right back and follows along as the kid moves a little bit to the right or left. It’s a very chilling, unearthly feeling to have this creepy looking robot follow your own movements.”

“Creepy and cute and the same time,” he added.

Bunting tinkered around using very inexpensive parts to build his first hexapod when he was in high school. A few years later, in 2009, he evolved his earlier ideas into a fully functioning hexapod while in a cognitive robotics class during his undergraduate studies at the University of Arizona.

“I liked the idea of having a machine being able to learn,” said Bunting. “So I built a hexapod that was much more capable than the first one I built in high school. I didn’t want to pre-program it with algorithms, so I didn’t use inverse kinematics and force all of the legs to move in the way that I wanted them to, but instead I used this reinforcement learning technique called cue-learning. With cue learning the hexapod was able to experiment in the world and figure out an optimal way to walk.”

After the hexapod could walk, Bunting used a Playstation 3 game controller that he hacked and made the robot dance for a video he posted on YouTube.

Since then, Bunting’s work has captured the attention of technology companies, including Intel and HRL Laboratories, his current employer and graduate school sponsor. His hexapod currently graces the cover of the June edition of Linux Journal, and the electronics and engineering industry publication, EETimes, just honored Bunting with a prestigious ACE Award, naming him this year’s Top Student “whose discipline, hard work and academic success are considered hallmarks for other engineering or science students.”

modern-davinci_thoven_2.jpg

After building his first hexapod in high school, Matt Bunting reinvented his robot design as an undergraduate at the University of Arizona, where he’s now pursuing a PhD. in Electrical and Computer Engineering and working on DARPA’s “Cheetah Project.”

Two days after Bunting posted his first video on YouTube, back in 2009, he was contacted by Stewart Christie from Intel’s embedded and communications group. Christie’s job is to find the next generation of developers, and sometimes that means meeting robot builders in their gear-cluttered garages or taking requests for equipment or project sponsorship on his team’s website.

Christie calls Bunting “a real modern-day Da Vinci,” pointing out that Bunting even created the Spanish guitar music that compels his spider robot to dance, as seen in the YouTube video.

“My job is to look for hobbyists, hackers and future customers who are working on cool projects and who might not be aware that they can build on Intel architecture,” he said.

“When Intel first contacted me I was a little bit shocked,” said Bunting. “I didn’t believe it at first … because you don’t normally think this kind of thing is going to happen to you … that somebody would recognize you for the work that you do.”

Soon after, the hexapod began to multiply as Bunting worked to meet Intel’s request for two replicas.

“With the help of Intel, I really worked hard to make it look the way I always wanted it to look, said Bunting. “I created an even better machine-learning technique, which involved an artificial neural network, and I used a generic algorithm to tune all of the weights.”

Bunting refers to his works as “biologically influenced robotics,” in which he combines the ability to see, learn and respond with movements.

Early on, Bunting said he wanted all of the processing to be done onboard, controlling all of the 21 motors, and processing vision with advanced machine learning techniques. Each of the six legs would be powered by three servos, or smart motors, and three more would control the camera.

“I needed something that was very powerful,” he said. “I also wanted something that would eventually run on batteries, so it needed to be powerful and computationally efficient at the same time. Fortunately, CompuLab came out with the Fit PC right around that time, so I ordered one of those and it has an onboard Intel Atom processor (running at 1.6 GHz). It could handle vision processing seamlessly along with all of the kinematic equations to operate all of the motors.”

Bunting used OpenCV, or Open Computer Vision, which is a library of free software for real-time computer vision.

modern-davinci_thoven_3.jpg

Matt Bunting has been described as a “real modern day Da Vinci”

“I’ve implemented a few vision computing techniques such as object recognition,”” said Bunting. “There’s face tracking onboard, so the hexapod can find your face and actually follow it around. For the machine learning, like the reinforcement techniques that I implemented, vision was used to measure the optic flow. Using the optic flow, it’s able to take two successive images and figure out how the objects are moving around in the world. The hexapod could tell if a person’s head has tilted, moved forward, up or down and that was all critical for machine learning.”

Currently a graduate student at the University of Arizona, Bunting spends much of his time coding and tweaking his creations inside the windowless Robotics and Neural Systems’ lab led by UofA Professor Tony Lewis. Inspired at a young age, and even today, by building Lego models and mathematics, Bunting has evolved into a valuable asset for the Tucson school’s Electrical and Computer Engineering Department, even helping to attract new students and big-time research projects like DARPA’s “Cheetah.”

According to Bunting, his team will collaborate with other researchers in a competition to create a four-legged robot that could eventually lead to a two-legged robot able to run faster than any robot today. Proof of his team’s prototype rests in his iPhone in the form of a photo showing robotic hind legs of a small cheetah.

“If we scale it up, I think we can actually achieve 70 mph,” Bunting recently told Arizona Engineer, a publication of the university’s College of Engineering.

Bunting said his hexapods are coming in handy as he focuses more time on DARPA’s Cheetah project. He said he’s extending the “machine learning” that he built into the six-legged robot, and refining the 3-D printer techniques he used to fabricate the light, hard plastic skeleton and legs of the hexapod.

In addition to the hexapod he has inside his lab, Bunting keeps in touch with his other hexapods through Twitter, where Intel’s Christie shares the robots’ whereabouts like elementary school science class visits, technology tradeshows, media events such as the recent Tech Heaven in New York, and competitions like Robocup in Rome.

Christie said that someday a robot like this could be programmed to interpret facial gestures and human expressions, which could make it useful in such places as school and hospitals.

“This hexapod is surprisingly easy to command using a Bluetooth game controller,” said Christie. “I can hand the controller to anyone, and they’re off and running, especially kids who are used to playing with the Playstation.”

The collaboration continues. “Now I’m encouraging Matt to train the hexapod to post photos to its Twitter account so people can see what the robot sees when it’s in action at an event,” said Christie.

Christie also pointed out that several other teams have developed ancillary projects, using the same Atom-based hardware that Bunting used, including a 10-foot wingspan autonomous aircraft and an underwater robot.

“They all cited Matt’s success as the reason for choosing the Atom processor,” Christie said.