Modern robots are built for very specific purposes. Industrial arms can assemble cars, and Roomba can vacuum your home. They are incredibly proficient at tasks that are specifically built and programmed to perform; but if today’s most sophisticated robot was trapped in a burning room by a jammed door, it would probably not know how to locate and use objects in the room to climb over any debris, pry open the door and escape the building.
Researchers at Georgia Tech in the USA want to change that with a robot that behaves like the television character MacGyver, who solved complex problems and escaped dangerous situations by using everyday objects and materials he found at hand. The machine has been dubbed the MacGyverbot and the challenge is to develop software that understands what objects are and then deduce how they can be used.
The team will develop a robot that uses a custom algorithm to scan a room for random objects, sizes up their usefulness and functions and then uses them to accomplish a more complex task. In essence the team will build a highly manoeuvrable robot that has human-like arms and hands and imbue it with an artificial intelligence that can analyse its surroundings to find potential tools. The researchers ultimately envisage a situation in which the machine might be deployed to rescue trapped personnel without needing to risk anyone else’s life.
The project is challenging because there is a critical difference between moving objects out of the way and using objects to make a way. While current robots use computer vision to manoeuvre their way around obstacles (for example Google’s self-driving car), they do not process any information about the obstacle’s function.
The MacGyverbot will look for objects that can be used to remove or otherwise circumvent obstacles. It might scan the environment, see something on the ground and quickly run through a list of possible uses for the thing. If no clear path is available, maybe it sees a chair and realises chairs have height, so it can use the chair to climb up to reach a window to escape a burning room; if a door is locked shut, the robot might look for a key or another implement that would allow it to force the door open; or maybe it sees a metal pipe and realises it can work as a lever to lift something heavy, freeing a human trapped under it.
Starting with a working knowledge of physics, rigid body mechanics and basic engineering, the team is studying the cognitive processes that enable humans to grab arbitrary objects and find creative new uses for them. The key to this approach is to create software that can identify every discrete object surrounding the robot, and then somehow work out their mechanical properties. It needs to be able to look at a chair and know that it is a chair – but it also needs to know what material the chair is made from. Will it hold the robot’s weight? Is it a wooden or steel chair?
The first step is to build a hybrid reasoning system that will use physics-based algorithms and a learning system to teach robots how to recognise and use various objects. This will be achieved by designing algorithms for robots that make tasks that are impossible for a robot alone possible for a robot with tools.
The software will be based on Icarus. This is a unified theory that tries to model human cognitive architecture, in essence how the human brain solves problems. The MacGyverbot is likely to see quite a few burning buildings. After the researchers develop and optimise the hybrid reasoning system using computer simulations, they plan to load the code onto Golem Krang, a headless humanoid robot already designed and built in Stilman’s laboratory to study whole body robotic planning and control, to see if it works in action.
Professor Stilman’s work on the MacGyverbot is the first of its kind and is already beginning to deliver on the promise of having mechanical teammates able to creatively perform in high stake situations.
For more information visit http://tinyurl.com/9cgoqr3
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version