This paper presents a novel approach towards the development of a generic model of an autonomous sensor (independent, self-reliant, self-sufficient). A case is made for autonomy as the ability to operate optimally using information available to the sensor without the intervention of other systems. The model's taxonomy accommodates behaviors common to most sensors. A taxonomy that includes knowledge and data bases is defined such as to equip the sensor model with decision making and learning capabilities based not only on analytic, or statistical methods, but also based on heuristic knowledge, and on the ability to represent behaviors that are difficult or impossible to represent with analytical or deterministic methods. The objective of the model is to allow the sensor to operate in an "intelligent" manner regardless of the control system in which it may be used. A simulated autonomous thermistor has been instantiated as an example. As defined, autonomous sensors would simplify implementation and operation of complex systems with large numbers of sensors, would improve systems that use multisensor integration and fusion, would maintain sensors at optimum operating conditions, and would aid in automatic task error detection and recovery in complex systems. © 1994.