Our firefront tracking algorithm estimates the position of firefronts by analyzing crowdsourced incident reports. The algorithm has been validated in an offline environment using historical data about emergency calls and fires (below). This week, Peter started migrating the algorithm into an online (streaming) environment.
The online approach uses Kepler, a free open source platform designed to help scientists create, execute, and share analysis processes. Kepler is a scientific workflow environment capable of integrating disparate software components, such as merging Matlab scripts with C and Java code. At its simplest, Kepler’s graphical user interface enables users simply to select and then connect pertinent analytical components and data sources (known as actors) to create a scientific workflow—an executable representation of the steps required to generate results.
Peter implemented the first basic workflow in Kepler for the firefront tracking algorithm (see above). The workflow uses a sliding window to aggregate input tuples. Tuples in the sliding window are processed online in the “Algorithm” actor, where the Matlab code of the algorithm is executed every time the sliding window fires. Peter is currently improving this basic workflow in two main directions:
- Developing more efficient, native actors for the algorithm.
Setting up a web-GIS to enable online visualization of the estimations.