Figure 2.

Algorithm for sensor placement. Initially the bounding box of the surface is subdivided into squares of edge length equal to the desired sampling density (panel 2). These points are then projected onto the surface to initialise the algorithm (panel 3). This is repeated for each face of the bounding box in the optimisation stage, at each loop iteration, a sensor is chosen at random (panel 4) and moved to a neighbouring vertex if the movement brings the observed sampling density closer to the target sampling denisty (panels 5, 6). This process is then repeated.