Extends realtime search algorithms for computational systems in which several computational agents interact to perform some set of tasks. Addresses the problem solver's inability to attain its goal without performing superfluous actions, to utilize and improve previous experiments, to be applied directly to a multi-agent world, and to cooperate with other problem solvers. Covers realtime search performance, controlling learning processes, adapting to changing goals, cooperating in uncertain situations, and forming problem- solving organizations. Double spaced. Annotation c. by Book News, Inc., Portland, Or.