FTG ENG | eBooks

Drift Error - Methodology Drift error, referred to in a geospatial context as Translation error, occurs when the model drifts away from the original anchored position over time, thereby shifting the entire model away from its correct alignment with the site. This most notably occurs when a user walks through their physical environment and moves further away from the original anchor point. To record this, key distinguishable features of a model were observed with a sub-mm precision Trimble Robotic Total Station in order to confirm its position over time. The distance by which the model moved is measured as the perceived drift error of the model overlay. Drift error was determined via the following formula: Where, is the d ) . e D = ( A−i * e S − d obs d A−i plan distance from the anchor point to the incremental point, i; and is the approximate scale e S error, as a percentage, found in Scenario 1; and is the observed distance from the anchor d obs point. Scenario 2 - Incremental Distancing A grid of points (see Figure 6 below) spanning up to 50m was laid out using a Total Station. The same grid of points was then modelled and viewed in TCH. The user anchored the model at the start line of the grid (0m) and incrementally walked away from the anchor. At each increment, the position of the model was recorded via Total Station. Upon reaching the end line (50m) the user would return and reset the model. A total of six rounds of one-way observations were taken to gather average positions and to mitigate user errors when aligning the model. A second set of observations were also obtained, this time using the previously released version of TCH (v3.1), which does not utilize the WLT, in order to assess the benefits of the WLT on the model's accuracy. Figure 6: 50m Grid Scenario 3 - Round Trip In Scenario 3, the user anchored a single point model and subsequently traversed a loop away from the point. Upon returning to the original anchor, the user then recorded the position of the model as TCH attempted to recognize its previously known environment. Observations were made via Total Station as in previous scenarios. Figure 7 below shows the described loop; a 6

• Cover