Dec 31, 1980
An investigation of potential organizations for automated distributed sensor networks (DSNs), i.e., dispersed nodes that can pool their information to perform accurate situation assessment. Laboratory experiments using a message puzzle task indicate that an "anarchic committee" organization, in which all nodes communicate with one another, consistently outperforms the "dynamic hierarchical cone" organization, in which communication is constrained and information must be obtained only from lower-level nodes. These experiments support the contention that DSNs must emphasize cooperative problem-solving rather than problem-reduction or subgoaling. A computer-based design that minimizes redundant communications in hierarchical organizations by using model-based reasoning to form expectations that guide, limit, and reduce reporting frequency is described. Finally, a method for representing hypotheses to minimize communication requirements--the process assembly network--is suggested. This concept uses active "hypotheses processes" that are responsible for predicting their own evolution over time.