A few of the many robots utilizing the capabilities provided by the Idaho National Laboratory Intelligence Kernel.
Robots as Team Members?
June / July 2007
Volume 5 Number 3
Since the early 20th century, science-fiction films and stories have amused and terrified us with robot characters, human-like machines programmed to serve.
These are often depicted as super-human, performing dangerous tasks in impossible conditions. However, a recurring theme in robot stories is lack of a trustworthy control system, allowing them to fail to do their jobs or become destructive. You may remember silky-voiced, out-of-control HAL in Stanley Kubrick's film, 2001: A Space Odyssey.
The real-life robotics industry has struggled with the same problem: how to engineer a control system that enables a robot to perform complex and significant operations independently, such as finding victims, locating dangerous materials, mapping the best routes in and out of a disaster site or uncovering hidden explosives. The adaptive robotics project at Idaho National Laboratory has overcome substantial obstacles to such a control system with its award-winning Robot Intelligence Kernel (RIK).
INL, operated by Battelle Energy Alliance, performs research and development for the Department of Energy and other agencies. Robotics is an obvious technology for DOE to pursue, since it has responsibility to monitor the nation's nuclear sites, including clean up and disposal of nuclear waste. Robots, of course, can be designed to detect contamination, collect samples or move containers in hazardous environments. Developments in robotics at INL have proven to be useful not only for DOE needs, but far beyond.
Urban search and rescue is one application in which the adaptive robotics program has shown great promise. A robot equipped with INL's RIK earned first place in the Urban Search and Rescue competition at the International AAAI (American Association for Artificial Intelligence) 2003 robot competition and exhibition in Acapulco. Researchers from other institutions, whose robots required hours of effort to calibrate them to the conditions at the competition, expressed surprise when INL's robot was set up in a matter of minutes then went on to win the competition by a wide margin. A competitor from a university robotics program commented admiringly, "You guys have stuff that actually works!"
The lab designers, led by David Bruemmer, used highly trained police dogs as one of the models for their robots. Says Bruemmer, "The model of the police dog is so valuable because dogs are good at the same sorts of things that our Intelligence Kernel is good at: sniffing bombs, locating landmines, finding plumes of various chemicals. They can detect things pretty well. They can also find humans, track them down. [They] get through buildings really well. They can kind of roughly keep track of where they are; in fact, they have a better sense of direction than most humans have."
To resolve control issues and design a robot that could actually perform similar tasks as a police dog, the team knew they had to focus on the relationship between the operator and the robot and create a system that would make best use of the strengths of each while compensating for one another's weaknesses.
Bruemmer says, "We had the goal, very early on, of dynamic autonomy, sharing control, better task allocation, predicting the human's needs, and shared understanding between the systems." So successful were the INL designers that their technology won an R&D 100 award for 2006. RIK can be used in several different robot applications and includes innovations both in the robot architecture—€”the design of the hardware and software that makes the robot work—€”and the methods used to communicate with the robot as it performs tasks.
The INL team equipped the robot with a five-layer "brain" that provides increasingly complex abilities: (1) a base layer generic robot architecture, (2) generic robot abstractions, (3) robot behaviors, (4) "cognitive glue" and (5) dynamic autonomy. The design combines both new and existing hardware, software and sensors, allowing robots carrying the RIK to be fully functional within themselves. All of the computing, sensing, and integration capabilities that make the robot work are onboard the robot.
The human operator receives the robot's perceptions on a tablet-style computer from a remote location and may direct all or part of the robot's activities. One key innovation is that the robot's perceptions are translated and simplified so that a small-bandwidth link, such as a cell phone modem, can carry them to the operator. This elegantly simple system allows successful communication between a robot inside a concrete containment building and an operator outside. In some very common applications, such as search and rescue, typical robot-to-human communication systems, requiring direct feeds from a video camera on the robot, may fail because common structural materials such as heavy concrete walls can interfere with wireless transmission.
The operator can receive information from the robot and communicate instructions in a variety of formats, including the video game-style interface shown below. This link is full of useful information gleaned from the robot's sensors and represented in a simple three-dimensional view.
RIK can operate at four levels of control with increasing autonomy: (1) full autonomy mode, where the robot functions independently to perform a specific mission, such as radiation survey, land mine identification and marking, or perimeter surveillance; (2) standard shared control mode, where the robot can search an area, plan a path or go to a place with intermittent input from the operator; (3) safe mode, which gives most of the control to the operator, but still allows the robot some autonomy to slow down on turns to avoid falling over, or to stop before it bumps into an obstacle; and (4) tele-operation mode, which gives all control to the operator and is the conventional commercial robotic control approach.
Colleen Thompson is a member of the technical writing staff at Idaho National Laboratory.
Copyright © 2012 | Innovation America