KNOWLEDGE CACHING FOR SENSOR-BASED SYSTEMS

Sensor-based systems must interact with their environments while extra cting crucial information, necessary for their performance, from the s ensors. In most cases, the projection from the environment to the sign al is many-to-one, resulting in irrecoverable information about the en vironment. To recover this information assumptions must be made. Consi dering the complexity of the world, we posit that intricate assumption s are necessary for recovering the information. More assumptions requi re larger knowledge bases, making the performance of the system slower than acceptable. To avoid the crippling effects of large knowledge ba ses, we accept additional assumptions about the structure of the worki ng environments and the interaction of systems with their environments along different dimensions. These assumptions allow systems to dynami cally hide large portions of knowledge that are irrelevant at a given time. We call this approach knowledge caching. We introduce an impleme ntation of this approach in the context-based caching (CbC) technique in which knowledge items are swapped based on precompiled relations be tween knowledge items. This technique enhances system performance prov iding it with the right information at the right time.