we were introduced to Bots_Alive, a small company of roboticists who have managed to develop a robotic critter with a carefully thought-out animal-like personality. And by hacking an existing robot toy and using your phone as a brain, they’re ready to sell it to you for 35 bucks.
Bots_Alive is, essentially, selling a replacement brain for Hexbug Spider robots. The Hexbug Spider is a robotic toy that’s been around for, I dunno, forever. It’s made out of plastic and costs about US $25. It comes with a little infrared controller that lets you drive it forwards and backwards and turn left and right, and otherwise it doesn’t really do much. Bots_Alive turns a Hexbug
Spider into a fully autonomous robotic critter by replacing the infrared controller with an IR blaster plugged into your phone, and using computer vision to localize a fiducial sticker placed on the robot’s head. As long as your phone’s camera can see the fiducial, it has full control of the robot’s movements, and gives it a life of its own:
Using a phone as the brains of a robot is an idea at least as old as Romo. The advantage of doing so is obvious: We’re all carrying around extraordinarily powerful little computers with vision systems and all kinds of other stuff packed into them, so why not steal that hardware and use it for robot control? With this in mind, Bots_Alive picked for a platform what has to be one of the cheapest and most common robot toys out there, the $25 Hexbug Spider. If you’re one of the thousands of people who already has one of these robots, all you need from Bots_Alive is an IR blaster and their software, which is why the starter kit for Hexbug Spider owners is just $35.
The kit comes with some blocky things (each with a fiducial on top); the robot will try and snuggle up to the blue ones while avoiding the red ones. It seems straightforward, but the robot will act a bit differently (and occasionally very differently) every time you mess with it. Exactly how those behaviors are programmed, and how they come together to give an otherwise brainless robot a personality, is where Bots_Alive has a whole bunch of their secret sauce.
“[At MIT], sometimes the grad students and the postdocs would get together and we’d think about why we don’t have robots that we ourselves want to interact with. Each of us had our own personal dream of what our robot companion would be like, and so we’d get together and we’d talk about why is that not here.”
The philosophy on robot companions that Knox ended up with was informed by a few different things. First, there’s an awareness that robotics, especially with social interactive robots, is playing (and often losing) an expectations game. And second, Knox really likes dogs. These things fit together because dogs, in a lot of ways, typify what could work with a social interactive robot right now: no speech, not really designed to perform “useful tasks,” but friendly and expressive and can somehow be consistently entertaining. All of this is manageable, and realistic, for an inexpensive robot.
Knox says he wants to build “simple creatures that feel very alive, that can maintain an illusion of life and the magic that comes with that.” To do that, he’s relying on his MIT research on learning from demonstration.
ALSO READ: DON’T BE FOOLED BY THIS GIANT ‘MECH’ ROBOT
“Ten years ago, learning from demonstration was a pretty tiny blip in the overall machine learning and AI sphere. Until we started working with it, nobody had used learning by demonstration for creating personalities or characters. It was for accomplishing various tasks where it’s not about the experience a person gets being around it. It was somewhat of an open question whether it would result in effective behavior and a good character, and we were pretty surprised by how well it worked.”
Knox’s research involved children interacting with MIT’s Dragonbot. First, he’d secretly control the robot when it was interacting with the kids through a Wizard of Oz-type interface, trying to be as natural as possible. Then, machine learning software analyzed what Knox was doing to generate its own decision making policy based on his interactions. Research showed that when the robot used this learned policy to interact with them, the kids couldn’t tell the difference from when Knox was driving the robot himself.
We’ve spent a little bit of time playing with the Bots_Alive Hexbug, and we can promise that it’s one of the most lifelike autonomous robot you’re likely to find for $35. It seems curious. It likes to explore. It makes mistakes and doesn’t always take the most direct route to where it wants to go. All of these behaviors are based off of data collected through Knox’s own manual teleoperation: On some level, it’s his personality embodied in the robot you’re playing with.
ALSO READ: JOB-STEALING ROBOTS: A GROWING CONCERN
The amount of things that the Bots_Alive Hexbug can do isn’t very large. But it’s interesting to speculate about what might be possible with a system like this. For example, there’s no reason that you couldn’t use Bots_Alive with any other remote control robot, whether it’s something cheap that uses IR or something more expensive that uses Bluetooth or Wi-Fi. As long as your phone can track the robot on video and send commands to it, Bots_Alive can make it intelligent and interactive. We also really like the idea of having an overhead camera for control, which hasn’t really been something that home robots have ever had access to before.