This paper offers an approach for defining and gauging the intelligence of an unmanned vehicle system. The meaning of "intelligence," as in natural intelligence and machine intelligence, varies greatly depending on the context in which the word is used. The paper describes certain basic expectations associated with being intelligent in general, and proceeds to present three possible means of gauging the intelligent behavior of an unmanned vehicle system (UVS). The first is a qualitative perception that is based on an extension of the Turin's test for perceiving an unsuspecting UVS behavior as that of a person. The second is a quantitative measure where task-specific intelligence performance of a smart system is evaluated. Examples of task-specific intelligence and how they are gauged are presented in the context of the Intelligent Ground Vehicle Competition. The third is the comparative scale that gauges the difficulty of challenges against the intelligent skills of humans. For example, an experiment revealed that it requires the mind of at least a four-year-old child to successfully navigate an autonomous navigation course. Alternative architectures for intelligent unmanned vehicle systems were also presented, including complementing machine intelligence with human supervision via telematics and information technology.
展开▼
