ABSTRACT
This chapter describes techniques for extracting semantic information from sensor networks and applying them to recognizing the behaviors autonomous vehicles based on their trajectories, and predicts anomalies in mechanical systems based on a network of embedded sensors. Sensor networks generally observe systems that are too complex to be simulated by computer models based directly on their Physics. The techniques are based on integrating and converting sensor measurements into formal languages and using a formal language measure to compare the language of the observations to the languages associated with known behaviors stored in a database. One method for generating a stream of symbols from the resampled sensor network data divides the phase-space volume of the network into hyper-cube shaped regions and assigns a symbol to each region. The sensor data are sampled at regular intervals, the behavior for each interval is determined and changes in the corresponding languages can be analyzed.
