As people live longer, the number of those with disabilities will rise significantly. About 15 percent of the world’s population today lives with some form of disability, up from 10 percent in the 1970s, according to the World Health Organization. Many disabled people are hard-pressed to do such household chores as cleaning and cooking or even simply moving around. But innovative technologies are being developed to help, and IEEE members are at the forefront of these efforts.
In this special issue, The Institute takes a look at the work of several IEEE members concentrating on quality-of-life technologies. Member Kimitoshi Yamazaki and his colleagues have developed a life-size humanoid robot that sorts laundry, sweeps floors, and performs other chores. Member Moritz Tenorth is working on a humanoid for cooking and other household tasks. Member Takanori Shibata has developed a furry, robotic seal [see photo] used therapeutically for patients with cognitive and psychological disorders.
And still more helpful technologies are in the works. Senior Member Hung Nguyen is developing a wheelchair steered by brain waves, while Senior Member Mahesh Krishnamurthy is building a motion-detection system with infrared sensors that makes a wheelchair more aware of its surroundings.
There’s also the work of IEEE Fellow Takeo Kanade in Pittsburgh at the Carnegie Mellon University Quality-of-Life Technology Center. Plus, several IEEE standards and conferences tackle topics related to improving people’s lives.
In Japan, the number of people older than 65 is expected to balloon: from 20 percent of the population in 2005 to 40 percent by 2055. Such statistics inspired Yamazaki to develop a robot to help with household tasks. The assistant professor at the University of Tokyo and his colleagues have built a 1.5-meter-tall, two-armed robotic housekeeper on wheels. Its onboard computer accepts commands from the patient that tell the robot to perform any number of chores, including picking up a tray and taking it to the kitchen, putting dirty clothes in the washing machine, and sweeping the floor with a broom.
To perform its chores, the robot must know about the world around it. “One of the biggest challenges is teaching the robot about real objects and programming it to recognize and manipulate them,” Yamazaki says. His robot uses cameras and sensors to map an environment such as a kitchen. It then produces a database of objects in the house—together with their movable parts, grasping points, and other task-related information—so it will know what to manipulate to perform a chore.
“We use a 3-D geometrical simulator to define a virtual world in which the objects are arranged,” Yamazaki explains. With this information, he and his team programmed the robot to perform each task.
Although such robots have been in development for decades, Yamazaki notes that recent advances have paved the way for them to become a reality.
“A combination of higher CPU power and more reliable hardware has allowed for the development of life-size robots with humanlike bodies that can perform daily tasks,” he and his colleagues wrote in “Home-Assistant Robot for an Aging Society,” which appeared in Proceedings of the IEEE in August. The issue, dedicated to quality-of-life technologies, is available in the IEEE Xplore digital library.
Tenorth, a postdoctoral scholar at the University of Bremen Institute for Artificial Intelligence, in Germany, is working on a similar robot.
“A key motivator for my team was helping elderly and disabled people stay independent,” he says. “While personal and social contact is crucial—and cannot be replaced by robotic assistants—we believe robots can serve as tools that give autonomy to people.” One of the robots he is working with, PR2, came from Willow Garage, a company in Menlo Park, Calif., that develops robots and open-source software.
PR2 has been programmed to perform a number of kitchen tasks, including making pancakes, boiling sausages, and preparing sandwiches. It can unload a shopping basket and fetch items from drawers, too.
“Our two key developments are the cognitive robot abstract machine [CRAM] plan-based control architecture and the KnowRob knowledge base,” Tenorth says. “CRAM is a framework for specifying and executing cognition-enabled robot control programs. KnowRob is a processing system that provides autonomous robots with knowledge they need to competently perform everyday manipulation tasks. Together, they allow programmers to quickly develop cognition-enabled robot control programs for making, for example, a sandwich.”
The robot relies on an inference mechanism to make decisions on the fly.
“A big challenge for robots is to understand vague instructions,” Tenorth says. “When humans explain a task to each other, they omit important information that is obvious. Robots lack this common-sense knowledge, so they have problems understanding vague directions. Giving robots the ability to infer what is meant from what is said is a huge research challenge that we started to address in different projects, for example, in the EU’s RoboHow project.”
Much work lies ahead before PR2 moves out of the laboratory.
“Apart from the daunting price tag [about US $400 000], such robots are not yet robust and useful enough to operate autonomously in real human households,” Tenorth says.
Pets have been used for years to help patients suffering from depression and stress or as a way to increase brain activity and encourage communication. Inspired by such therapy, Takanori Shibata, a chief senior research scientist at Japan’s National Institute of Advanced Industrial Science and Technology, in Tsukuba, built a robot that resembles a cuddly animal. PARO looks like a furry seal, is about the size of a real baby harp seal, and responds as if it were alive, wagging its head, flippers, and tail and making sounds. Selling for $6000, it is now in its eighth generation and recently became certified by the U.S. Food and Drug Administration as a neurological biofeedback therapeutic device.
Shibata, who developed PARO in 1998 as a visiting research scientist at MIT’s Artificial Intelligence Lab, has been refining it ever since. PARO has five types of sensors—tactile, light, audio, temperature, and posture— with which it can perceive people and its surroundings.
For example, with the tiny tactile sensors that cover PARO’s entire body, it feels when it is being stroked or hit or is merely being held, thanks to a posture sensor that determines the orientation of the patient’s body. With its audio sensor, PARO can recognize some words and the direction a person’s voice is coming from, as well as learn its own name (if the patient gives it one.) It can also sense expressions of kindness, such as words of praise, because it can interpret the tone of a person’s voice.
Since 2003, PARO has been used in hospitals in Australia, Europe, Japan, and the United States for patients with cognitive disorders such as dementia, autism, and Down syndrome, as well as brain injuries and psychological problems. The robot seal was used to help victims psychologically traumatized by the 2011 earthquake in Japan.
PARO can learn to behave in a way the user prefers. If gently stroked, for example, PARO will remember what it did just before the stroking began and learn to repeat the action. And if the robot is smacked, PARO takes that as a disciplinary action, remembers what it did just before the reaction, and tries not to do it again.
Shibata says he hopes to create other versions of his seal suited for particular disorders, including the teaching of social skills to children with developmental problems.
“Those who use PARO seem to be very happy with it,” he says. “I hope to see more make their way into medical facilities and even people’s homes.”