This can be fleshed out later. This is a very clearly written paper that posits that a robot doesn't need to have any representation of its world in order to make intelligent decisions. It uses independent and parallel activity producers that all interface directly to the world through perception and action, and have minimal interfacing with each other.
He proposes that tests of artificial intelligence must use robots embodied in the real world, rather than simulations. Anything less provides a candidate with which we can delude ourselves.
Unexpected conclusion: When we examine very simple level intelligence we find that explicit representations and models of the world simply get in the way. It turns out to be better to use the world as its own model.
Radical hypothesis: Representation is the wrong unit of abstraction in building the bulkiest
parts of intelligent systems.
This is a schematic of their robot control system. They call this subsumption architecture, which embodies the fundamental ideas of decomposition into layers of task achieving behaviors, and incremental composition through debugging in the real world.
A more detailed description of the control system is found in another one of Brooks' papers.
He posits that as of the late 1980s, this is the best mechanism for programming intelligent robots.
No comments:
Post a Comment