Researchers at Georgia Tech are creating a next-generation humanoid robot—a self-described “robot that behaves like MacGyver”—for the military that can create simple tools out of any nearby objects. The robot acts autonomously and can rescue troops with impromptu tools, as well as block suspects' range of motion by jamming doors and windows shut.
The Office of Naval Research (ONR) recently gave Georgia Tech $900,000 for the project. A team led by Mike Stilman, a scientist who specializes in humanoid robots designed to interact with humans and autonomously move objects, is in the process of developing a robot the size of an adult human which can perform high-level tasks. The development of algorithms that allow robots to identify random physical objects, determine their potential functions, and turn them into simple machines for performing actions is at the core of Stilman's project.
If you're wondering why one wouldn't bring along, say, a travel size set of tools instead of a robot that can fashion many of those same tools from a wire coat hanger and some old dental floss, then let Stilman explain the big idea.
“Our goal is to develop a robot that behaves like MacGyver, the television character from the 1980s who solved complex problems and escaped dangerous situations by using everyday objects and materials he found at hand […] We want to understand the basic cognitive processes that allow humans to take advantage of arbitrary objects in their environments as tools. We will achieve this by designing algorithms for robots that make tasks that are impossible for a robot alone possible for a robot with tools,” Stilman says.
ONR researchers have been interested in human-robotic interaction for quite some time. The institution is conducting active research into human-centric autonomous robots designed to work as “peers” in the field with flesh-and-blood warfighters. Georgia Tech has been involved in robotics projects for the ONR for quite some time—in 2010 another professor at the university, Ronald Arkin, conducted a study on the possibility of teaching robots deceptive and self-benefiting behavior with a Navy grant.
Machine learning technologies will be applied to one of Stilman's robots, Golem Krang, in order to teach it to autonomously build tools. A separate team at the university is developing a specialized tool-building cognitive architecture for the robot based primarily around perception and inference.
It is important to note that there are no guarantees for a working tool-building robot coming out of the project, and that Georgia Tech's work will more likely be a groundbreaking proof of concept rather than an intelligent, autonomous robot for mass use by the military. Researchers will be working on the project until late 2015.
[Image: Georgia Tech/Josh Meister]