Service dogs that open doors, switch on lights, and perform other useful tasks offer a much needed lifeline to people with disabilities. Now researchers at the Georgia Institute of Technology are developing robots that mimic the relationship between humans and their canine helpers.
Robotics researchers have long sought to create robots that can help out around the home. But while robots are good at carrying out preprogrammed tasks and following a clear trajectory, navigating a complex home environment and interacting with real people remains a formidable challenge.
Charles Kemp, a professor at Georgia Tech, believes that animal helpers may offer the ideal model for robotic assistants. He began by studying the way that helper monkeys–capuchins trained to perform useful tasks for disabled people–fetch an object or operate a device when it is highlighted with a laser pointer. “That got us excited about what we can learn from state-of-the-art biological systems,” says Kemp. It also inspired him and his colleagues to develop El-E, a robot that they trained to respond to commands given via a laser pen earlier this year.
More recently, Kemp and his student Hai Nguyen realized the potential of canine helpers after seeing a demonstration given by a charity called Georgia Canines for Independence. These dogs are trained to open doors, drawers, and cupboards and to fetch objects or operate lights when given a command. “We were amazed at what the dog could do [and] found out there’s a list of commands service dogs obey,” Kemp says. “That seemed like a great model to go by. If we could make a robot that obeys all those commands, we knew that we would have something valuable.”
The latest version of El-E has been upgraded so that, in addition to responding to a laser pointer, it understands voice commands and can perform a wider range of tasks. The robot can be commanded the same way as a service dog–to grab hold of a towel attached to a door, drawer, or cupboard when given the right vocal command. As with service dogs, towels help the robot with both perception and physical interaction. “[El-E] doesn’t know anything about the specific drawer or doors: it’s able to generalize with these commands,” says Kemp. “A towel is actually easy to grasp because you can be at many locations on it and still get a good grip.”
The researchers trained El-E to recognize a towel visually. The robot identifies the knot and the hanging end of the towel, grips the bottom of the towel, and wheels itself backward, forward, or pulls down. It uses voice-recognition software to follow the same basic commands as a service dog, such as “tug it,” “tug it down,” “push,” “bring it here,” and others.
Kemp presented his work on Tuesday at the IEEE International Conference on Biomedical Robotics and Biomechatronics, which took place in Scottsdale, AZ. Out of 40 trials, the robot was able to open and close drawers with a 90 percent success rate and open doors with an 80 percent success rate. El-E was also able to open a microwave door, using a towel attached to the door with a suction cup.
Kemp notes that a robotic service assistant would not require the same training and care as a service dog, potentially offering help to many more people. “A lot of people who would like a service dog are unable to have one because they are costly and there’s a long waiting list,” he says.
Andrew Ng, a professor of computer science at Stanford University, who was not involved in the work, describes it as “very innovative,” and adds that copying service dogs makes it much easier for a robot to operate in an unfamiliar environment. “I can easily imagine robots that use these ideas making it into the homes of disabled individuals in the next few years,” he says.
“If we’re willing to engineer the environment a little bit, [robots] will be in our home a lot sooner,” says Kemp. In addition to those with disabilities, he suggests, the elderly could benefit from having helper robots. “Service dogs give you a level of independence and privacy that you might not have with a human caregiver, and robots have that same potential,” he says.
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
Data analytics reveal real business value
Sophisticated analytics tools mine insights from data, optimizing operational processes across the enterprise.
Driving companywide efficiencies with AI
Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.