Tending to the needs of severely disabled people as a youth inspired Takashi Gomi to ply the tools of his scientific trade in giving them one thing they deeply crave: autonomy.
"To use a motorized wheelchair to go 9 kilometers into town can be a strain [to work the joystick]. Their hands and arms get tired; they can't go far," said Gomi, president and founder of Applied AI Systems.
One way to help, Gomi reasoned, was to create a vehicle that could steer itself under the direction of the passenger but would require little - if any - physical effort. Gomi's ideas were the seeds from which sprouted an intelligent wheelchair, a vehicle that couples Applied AI's own robotics and artificial-intelligence algorithms with off-the-shelf optical and infrared technologies.
The heart of Gomi's intelligent wheelchair is a multitasking operating system that can react in real time to information from various optical inputs, including two CCD cameras, each mounted under an arm of the chair, as well as bump and proximity infrared sensors located under the footrests and on other supporting structures. This software uses a relatively new form of behavior-based AI called subsumption architecture, or biorobotics.
Developed in the mid-1980s by Rodney Brooks, biorobotics is derived from non-engineering disciplines, including biology, physiology, behavioral sciences, genetics, and theories of evolution and psychology. Devices based on biorobotics can "think" more like a human does, adroitly adjusting to sudden changes in an environment. For example, when Gomi's wheelchair comes upon an unexpected obstacle such as a person running toward it, the sensors and cameras send the information to the robotic system processor - a Motorola 68322 - which then signals the chair to slow down and maneuver around it. If subsequent signals indicate that the obstacle has changed positions or is getting larger - as would be the case when someone is walking toward the chair - the system instructs the chair to stop until the person or object has passed.
This "behavior-based" system allows a device such as Gomi's intelligent wheelchair to think on its feet, because the information gathered from the sensors is immediately handed off to the system that sends a signal instructing the chair what to do. There is no hierarchy in the software, so instructions such as "go to kitchen" do not have to jump through countless procedural hoops before they are processed. Gomi's system cuts to the chase, simply reacting to its situation (though the overriding instruction is to avoid obstacles). And all instructions can be overridden at any time by engaging the joystick.
"The program is small, easy to debug, and it's easy to add new capabilities," said Matt Beadon, an intern at Applied AI and an engineering student at the University of Waterloo in Ontario, Canada.
By contrast, traditional knowledge-based AI systems must run through a rigid regime of procedures to execute their instruction sets. A wheelchair based on such a system would have to process what the optics and sensors perceived, map out the entire environment, develop a course of action, break the action down into tasks, and transfer these in steps to the wheelchair. For every new environment such as a room, the system would have to start from scratch in going through this cycle - one that would take longer than a behavior-based model and would not be able to adjust to sudden changes in the landscape, Beadon said. A sudden obstacle would cause the system to hang up; the wheelchair would stop.
Gomi's solemn face and unassuming manner mask the unbridled ambition he has for the intelligent wheelchair. Some of the new capabilities Gomi plans to add to the chair include a level of voice recognition so that users can speak their directions into the robotics system. Eventually, he would like to use GPS technology to give the chair a bigger repertoire of destinations beyond the 15 landmarks that can now be programmed into the system.
But the bright promise that developments like Gomi's offer is viewed through a slightly skeptical lens by those who work with wheelchair users. Betsy Bayha, director of technology policy at the World Institute on Disability, has seen several incarnations of powerchairs - a term coined by wheelchair users - that use a combination of dexterity and breath as propellants.
"There are a lot of technologies built by well-meaning people who see a need," Bayha observed. "But a lot of designs don't serve a particular need because they are built for a person instead of actively involving a person."
Building upon his own experiences as a young volunteer, Gomi has on his staff in Canada a wheelchair user as a consultant. Applied AI is in constant contact with hospitals and organizations that work with disabled people to test the chairs, and some of Applied AI's employees are following Gomi's lead and volunteering to help people with disabilities.
Through her own travels, Bayha has seen a spate of motorized wheelchairs, and noted that people who use them have had little trouble. This would include the founder of the Oakland, California-based World Institute on Disability, the late Edward Roberts. Paralyzed from "his chin to his toes" by polio and unable to lift his arms, Roberts used a powerchair with a sensitive switch that reacted well to the movement in two of his fingers, Bayha recalls. She also noted that a friend of hers uses a wheelchair that has a "sip and puff" switch that allows one to control the chair via lung power.
Bayha is in awe of the technology and hopes that developments such as artificial intelligence and miniaturization of computer components will be of some benefit to disabled people. "It's nice to see this technology that has brought convenience to others making its way into wheelchairs to help people with disabilities," she said.
Nonetheless, technology isn't a panacea for autonomy. Overcoming the true obstacles for the disabled will require some engineering know-how to develop another new type of AI - architectural intelligence, Bayha says.
"There are stairs everywhere, buildings have narrow doorways, and many streets have [steep] pitches that take a lot out of the [wheelchair] batteries," she noted. "The built environment is the obstacle."
