The CKS, a knowledge repository for a vision system, is designed to serve a community of independent, interacting processes that cooperate in achieving the goal of scene modeling. These processes may represent sensors, interpreters, controllers, user interface drivers, or any other information processor. Each process can be both a producer and a consumer of information. Each has access to, and control over, a certain limited portion of the knowledge/database resources. The CKS architecture permits access to stored knowledge both by geographic location and by semantic content. One of its principal innovations is to provide a way of dealing with inconsistent data resulting from unreliable sensors and their associated interpretation processes.