Figure 4
From: Learning search polices from humans in a partially observable context
Resulting search GMM and information gain map. Left: Resulting search GMM. A total of 67 Gaussian mixture components are present. We note the many overlapping Gaussians: this results from the level of uncertainty over the different choices taken. For example, humans follow along the edge of the table in different directions and might leave the edge once they are confident with respect to their location. Right: Information gain map of the table environment. Dark regions indicate high information gain as oppose to lighter ones. Not surprisingly, the highest are the corners, followed by the edges.