======
README
======
The files in this archive are MatLab implementations for the algorithms presented in the
IROS 2013 submission, Heterogeneous Map Merging Using WiFi Signals, by Gorkem Erinc, 
Benjamin Balaguer, and Stefano Carpin. We strongly encourage potential users to read the paper 
to get a better understanding of the implemented algorithms, file structure, and provided data. 
Please send an email in case you have any comments or questions: gerinc[at]ucmerced.edu

File Structure:
---------------

	Maps:
	-----
		Inside the maps.zip file there are 3 occupancy grid maps saved as "gridmapX.mat" and 
		3 appearance-based map saved as "vismapX.mat" where X is the run number used to build 
		the map (either 49, 50, or 51). Each map is stored as a struct with several fields 
		primarily used for testing purposes. The fields that are used by the merging algorithms 
		described in the paper are the following.
		
		gridmap:
		--------
			gridmap.wifi.pos		: Ground truth Cartesian coordinates of cluster centers
			gridmap.wifi.posAll		: Ground truth Cartesian coordinates of locations at which a WiFi reading is acquired
			gridmap.wifi.APs		: Access points from which signal strength values are acquired
			gridmap.wifi.w			: Signal stregth values
			gridmap.wifi.labels		: Indices of clusters each WiFi reading belongs to
		
		vismap:
		--------
			vismap.wifi.pos			: Ground truth Cartesian coordinates of cluster centers (Cartesian positions of images)
			vismap.wifi.APs			: Access points from which signal strength values are acquired
			vismap.wifi.wc1			: Signal stregth values
			vismap.wifi.labelsc1	: Indices of clusters (images) each WiFi reading belongs to
			vismap.dict				: Bag of Words dictionary of appearance-based map
			vismap.graph.V			: Vertices of appearance-based map
			vismap.graph.E			: Edges of appearance-based map
	
	Code:
	-----
		main.m			: Loads maps to be merged, train them if not trained before, and plot the merged map
		mergeMaps.m		: Calls mapOverlap and localize functions to merge two maps
		mapOverlap.m	: Applies OC SVM to compute map overlap between two maps (corresponds to the first step of the merging algorithm)
		localize.m		: Applies the merging algorithm described in the paper
		plotMergedMap.m	: Given the estimated and actual positions, plots them on the occupancy grid map