Soon, robots will be able to help the elderly and disabled answer the door, prepare a meal, take their medicine, and even get from a bed to a wheelchair (Fig. 1). As the number of retirees in the U.S., Japan, and other parts of the world continues to rise at an alarming rate, the need for robotic assistants becomes more prevalent.
A number of factors play into this need. More and more people are waiting until later in life to have children, making it difficult for adult children to care for elderly parents. And with the rising cost of healthcare, it's becoming increasingly costly for patients and their families to afford home health aides, personal nurses, and the like. So if robots can become affordable (and insurance companies would be willing to contribute to the cost), the non-human caregivers could revolutionize the way healthcare is provided.
The changing face of healthcare is prompting companies like Honda to kick their research into high gear. Yet researchers disagree about how these robots should appear and operate.
TWO RESEARCHERS, TWO OPINIONS
Hiroshi Ishiguro, a professor in Osaka University's Department of Adaptive Machine Systems, believes robots should look, act, and even feel like humans. (See figure). Yet Stephen Keeney, the project leader for Honda America's North American Advanced Step in Innovative Mobility (ASIMO), says they will have to look more artificial and like something out of a sci-fi movie before humans will accept them.
"The Japanese want humanoid robots, especially the elderly," says Ishiguro. "Japanese culture is very robot-centric, as they are used as toys from early childhood, and children in Japan love to play with real androids. There are even special places for androids in Japan, and they are constantly shown in cartoons and movies. Therefore, Japanese people are much more accepting of androids."
There's a different perspective in the West, where Keeney says robots aren't meant to replace human workers. "Rather," he says, "they are meant to be a machine to help make our lives better. We should always be cognizant that ASIMO is a machine and should be approachable and not be scary to children. It is a comfortable middle ground between machine and humanoid androids that others are working on."
ASIMO ON THE GO
Five unique processing systems control the ASIMO, and in terms of compute power, each system would rival today's top-end personal computers. Honda's engineers also had to devise complex algorithms for predictive step taking, including walking, running, running in circles, and climbing steps. This was particularly challenging because no study has ever really revealed the logic behind how humans walk, shift their weight, and avoid objects.
ASIMO is the culmination of 20 years of research and development. It uses sensors and algorithms to process moving objects and access distance and direction. For example, ASIMO stops when someone walks in front of it. Much like bats as they navigate and find prey, it also uses supersonic waves to sense movement and map its environment.
Other sensors enable ASIMO to better interact with humans. It can recognize postures and gestures, distinguish voices and sounds, identify the source of sounds, and distinguish between up to 10 different human faces. It can even address people by name.
Working in conjunction with an infrared laser beam, two visual cameras in the eyes and two infrared cameras help ASIMO find objects on the floor up to six feet away. The infrared cameras use landmarks placed on the floor so ASIMO can reposition and recalibrate itself within its environment. The markers consist of two triangles with different markings, so ASIMO can differentiate between them. It then uses itself and the two markers to triangulate its position.
Network integration lets ASIMO access the Internet and provide news, weather, and other info. It can integrate with a user's network system to access information about people at an event, show a picture of the visitor's face, and then guide visitors to their destination.
A series of actuators in the robot's joints provides 34 degrees of freedom, including three for head rotation and up/down movement, seven for each arm, two for each hand (not including joints for five bending fingers), six in each leg, and one for the torso.
ASIMO FOR THE ELDERLY AND
Robots must take on some human characteristics to deliver the kind of care a human counterpart provides.
"We are making a humanoid robot instead of one that runs on treads or wheels to deal with steps, doorknobs, and so on, so it can operate effectively in our world. Therefore, it needs to reflect a human form," says Keeney.
ASIMO was created with human-like arms, legs, feet, hands, and digits so it could do things like turn a doorknob to answer the door, bend its legs to lift a patient from his wheelchair, and climb stairs to find a medicine bottle—feats that robots of old couldn't accomplish.
"We are not trying to limit ourselves in the number of ways ASIMO can help," Keeney says. "ASIMO is not supposed to be a replacement for human caregivers. If you want to reach the mass market, it must be affordable. Our objective is to help people that are disabled or those who may need extra help."
While ASIMO's capabilities increase daily, such a robot is not yet within financial reach. "A lot will depend on the momentum with which technology continues to improve," says Keeney.
"As memory becomes cheaper and technology gets scaled down, ASIMO will be cheaper to manufacture. ASIMO must be smart enough to carry out commands, and we have a lot of work on artificial intelligence (AI) to make ASIMO useful in a home or hospice environment. We are probably more like 40 to 50 years from the ASIMO that can take over the household," he notes.
For now, though, Honda expects a useful version of ASIMO to be made available in about 10 years.
Honda's goal is to establish a database that contains all of the necessary information for ASIMO to function as an assistant. Most likely, such a database would include terabytes of information that would enable ASIMO to recognize voice commands and rationalize thought processes.
"The more we study AI, the more we learn what we don't know about the human brain. Every day, new challenges are emerging. We are so much in the infancy that I don't even know where to begin," says Keeney. "The more that you think about even the simplest of processes, the more you realize we rely on past experiences and common sense, and how do you write code for that?"
Key challenges lie in improving and even perfecting interpersonal and social relationships between androids and humans. While Ishiguro and his colleagues have arguably created the world's most human-looking androids, their behaviors, facial gestures, and other body movements still need improvement.
Ishiguro also is attempting to better understand the human brain and apply cognitive science to his robots' programmed behaviors. For example, if two young women are walking next to each other and chatting, we assume they're probably friends. If two people are holding hands, we assume they're in some kind of close relationship.
Yet what is the android to conclude if a short mother and her tall son are walking together and holding hands? As humans, we would likely be able to surmise the relationship. But for androids, this task could be challenging.
Also, for instance, clearing our throats in a certain way may indicate discomfort or the need for attention. How is an android to determine when we are simply clearing our throats and when there may be some other meaning?
While there are social and moral implications to consider, Ishiguro and his colleagues first need to tackle some practical issues. For example, robots need to differentiate between individuals in large crowds of people. In one test, individual children in a large group received their own RFID tags, and the android had no problems identifying them.
Perhaps the most interesting challenge is a phenomenon known as the Uncanny Valley, theorized by Masahiro Mori in 1970. As an android's appearance and motion become more human, a human's emotional response to it becomes more positive. But this positive emotional response only increases to a point where the appearance and motions are almost too perfect and become eerie.
Then, human beings become strongly repulsed by the nearly human android. The positive response only returns when androids and humans are indistinguishable. At that point, human beings may empathize with androids as if they also were human.
Humans may be repulsed if the rate at which an android can blink its eyelids is too fast or too slow, or if it isn't smooth.
Yet if the blinking pattern matches that of an average human, we would empathize with the android. Not surprisingly, children are the first to pick up on and be repulsed by non-humanlike motions and appearances.
Just last year, Ishiguro's android's were still in the Uncanny Valley, but recent progress has changed that. "Now, with the correct number of actuators, the android has come out of the Uncanny Valley. The movements still need improvements, but the current generation is much improved," Ishiguro says.
"The young infants are now not afraid. We are still very far from replicating the exact movement of humans, and there is no way to have a perfect copy of humans within 50 years. However, with very short interactions of a minute or two, most would not know they are interacting with an android."
According to Ishiguro, human-like movement is the most important characteristic. Androids must also be able to understand answers and surmise information based on conversations. And, there's a need to research androids in real-life situations. For instance, take an android to a shopping mall and observe its behavior. Then it's back to the lab to make improvements and continue the cycle.
To better understand human behaviors and apply cognitive science, Ishiguro's team is about 50% psychologists and 50% engineers of varying fields. To improve social behaviors, the psychologists are working closely with the engineers to create algorithms that attempt to mimic the cognitive aspects.
"We are interested in making a human robot so we better understand humans," says Ishiguro. "In our search, we ask ourselves what it means to be human. This is the psychological aspect of android research. Yet there are obviously several hard sciences involved in robotics as well. So it is a great combination to learn about both humans and engineering."
MAKING ANDROIDS HUMANOID
Now take ASIMO's underlying technologies and apply human-looking skin made of silicon with integrated piezoelectric touch sensors and plenty of actuators for controlling facial features. You'd get Hiroshi Ishiguro's Androids. Ishiguro and his fellow researchers at Osaka University want to create androids that, at first glance, are indistinguishable from the humans they resemble (Fig. 2).
"The android is \[a\] communication tool. In Japan, we have a serious problem: too many old people without enough young people to entertain them. Many elderly can walk themselves, so assistance with a mechanical system can be provided with communication support. Physical support can also be provided," says Ishiguro.
"And, androids are excellent for general-purpose use with children. Androids can also provide entertainment in the form of tour guides and can be used as receptionists to explain a company and greet people as they entered the building. Androids would make good companions, and this is the most important market for the android technologies."