Sensing and Systems in Pervasive Computing: Engineering Context Aware Systems (Undergraduate Topics in Computer Science)
Format: PDF / Kindle (mobi) / ePub
Focus on issues and principles in context awareness, sensor processing and software design (rather than sensor networks or HCI or particular commercial systems).
Designed as a textbook, with readings and lab problems in most chapters.
Focus on concepts, algorithms and ideas rather than particular technologies.
many of the techniques in mainstream use. 7.4.1 R-Trees A data structure which is widely used in processing geographic data is the R-Tree , and its variants, e.g. R*-tree . An R-Tree is a tree data structure which uses location to organise grouping of data within nodes and the relationships between parent and child nodes. Each tree node contains data for a region in space, known as a bounding box (we shall describe a 2D model here with rectangular boxes, but n-dimensions are possible).
process of removing noise often requires multiple readings be taken over time, but this may slow the detection of changes in state. Similarly, transient states (often relating to moving sensors or noise) can be lost in signal processing that favours stability and noise suppression over responsiveness. States can be defined at three levels: the human interpretation, the system interpretation and the sensor input. At both ends of this chain there is variation in perception and so in the system
according to expected noise in the system; may be determined by the definition of states (as discussed above); or may be defined according to experienced deviation (taking care not to interpret real change as deviation where change is discrete). In Chap. 5 we started to discuss the trade-offs in defining event based and polling sensing, and the need to define high-level interpretations and responses to context which may have discrete conditions which are being identified. These requirements often
illustrate an underlying signal being represented by a filter which gives an upper and lower likely bound. This bound is then used to identify which state is to be used. “A” and “B” are more general states which both overlap “ab”; “aa”, “ab” and “bb” are narrower state definitions, which should reflect greater certainty about the real conditions. Note that the more general state “A” is used at time 7 where patterns of noise cause the numeric model to fit into neither “aa” nor “ab” during the
considerable overhead. Gossip routing involves nodes contacting a few others, typically either at random or neighbours assigned at random or through some name space locality or local information about interests, and passing data on—and possibly receiving some back. This gossiping progresses in rounds, allowing the data to spread through the network despite any changes in topology as discovery is typically a part of the partnering process. The protocol is simple and requires no routing tables,