Artificial intelligence research has foundered on the issue of representation. When intelligence is approached in an incremental manner, with strict reliance on interfacing to the real world through perception and action, reliance on representation disappears. In this paper we outline our approach to incrementally building complete intelligent Creatures. The fundamental decomposition of the intelligent system is not into independent information processing units which must interface with each other via representations. Instead, the intelligent system is decomposed into independent and parallel activity producers which all interface directly to the world through perception and action, rather than interface to each other particularly much. The notions of central and peripheral systems evaporateeverything is both central and peripheral. Based on these principles we have built a very successful series of mobile robots which operate without supervision as Creatures in standard office environ...

this paper for subsumption programs have physical embodiments as copper wires which provide the medium to support the serial sensing of messages. Herbert has 30 infrared proximity sensors for local obstacle avoidance, an onboard manipulator with a number of simple sensors attached to the hand, and a laser light stripping system to collect three dimensional depth data in 60 degree wide swath in front of the robot out to a range of about 12 feet. A 256 pixel wide by 32 pixel high depth image is collected every second. Through a special purpose distributed serpentine memory, some number of the onboard 8 bit processors are able to expend about 30 instructions to each data pixel. By linking the processors in a chain we are able to implement quite good performance vision algorithms. Connell (1988b) is programming Herbert to wander around office areas, go into peoples offices and steal empty soda cans from their desks. He has demonstrated obstacle avoidance and wall following, real-time recognition of soda can like objects and desk like objects, and a set of 15 behaviors (Connell, 1988b) which drive the arm to physically search for a soda can in front of the robot, locate it and pick it up. These fifteen behaviors are shown as fifteen separate finite state machines in Figure 8.7. Herbert shows many instances of using the world as its own best model and as a communication medium. The laser-based table-like-object finder initiates a behavior which drives the robot closer to a table. It doesn't communicate with any other subsumption layers. However when the robot is close to a table there is a better chance that the laser-based

How can we build artificial creatures which inhabit the same world as us? To me this has always been the romance of Artificial Intelligence. Artificial Intelligence as a discipline has gotten bogged down

Before artificial intelligence set its mind on developing abstract intelligent agents which can think, Alan Turing suggested training embodied machines equipped with sensors and actuators to accomplish intelligent tasks like understanding spoken English (Turing, 1950).