The industry's recognized authority for design, engineering and application of equipment and services in the global ocean community
Issue link: http://sea-technology.epubxp.com/i/477967
30 st / March 2015 www.sea-technology.com be mentioned that the result does not represent the "abso- lute truth." Changing the coeffcients, the algorithm can be adjusted, thus, like in any other statistical method, there is some uncertainty remaining in the result. Additionally, physical parameters that have an infuence on sound prop- agation in water (density, temperature, salinity) are highly variable in the Arctic marginal ice zone. These conditions restrict the performance of every acoustic tracking system. Taking this into account and considering that the vehicle operated close to an ice cover, GAPS performed very well. Considering the navigation error that occurred during this exemplary dive (PS80/187-2), correction methods are necessary to gather resilient scientifc data during these kinds of AUV missions. It is clear that the algorithm tremen- dously improves the quality of the navigation data. Mean- while, the navigation errors due to the foat maneuver have been minimized. However, as the vehicle still conducts mis- sions without bottom lock, deviations in navigation data did not vanish completely. For this reason, the herein described algorithm is used by default at AWI. Acknowledgments The authors would like to thank Ulrich Hoge (AWI) for installing and operating GAPS on RV Polarstern and Nor- men Lochthofen (AWI) for his support in handling GAPS raw data. Additionally, the authors would like to thank James Kinsey from Woods Hole Oceanographic Institution (WHOI) for a very interesting meeting in Bremen in April 2013, which ultimately resulted in a big improvement of the algorithm. The investigations described in this article were accomplished within the framwork of the Helmholtz Alli- ance "Robotic Exploration of Extreme Environments—RO- BEX." n Uwe Wulff is a retired software engineer who has been involved in AWI´s AUV operations since 2010. After he received a diploma in mathematics from the University of Hamburg, Germany, he specialized in system analysis and was employed at Asea Brown Boveri (ABB) in network management until 2004. Thorben Wulff is a Ph.D. student at the Alfred Wegener Institute in Bremer- haven, Germany. He holds a master's degree in mechanical engineering from the University of Applied Sciences in Mannheim, Germany. Using an AUV, he is investigating the three-dimensional distribution of biogeochemical parameters in the Arctic marginal ice zone. a certain minimal number of consecutive tracking results (elements) from the set of valid GAPS data. The minimal number of elements within these sequences is given by the variable n S . The last elements of each of these sequences are used to calculate a weighted average—pushing the calculated po- sition towards the more reliable end of the sequence. The number of elements included in this weighted average is specifed by variable n A . In order to defne a time stamp for the knots, they are assigned to the recording time of the last element of the re- spective sequence. A depth is defned by selecting the re- spective depth value of the AUV (AUV depth ). For a time T x that represents the recording time of the last element of a sequence, the knot (K P3 ) can now be described in all four dimensions. For this example, 15 elements were applied to identify sequences. The weighted average was calculated using the last fve elements as follows: n S =15 n A =5 For this dive, the number of knots was only 4 percent of the valid GAPS data. These knots are considered to be loca- tions at which the AUV was de facto. In the fourth and fnal phase of the correction process, the preliminarily corrected AUV data of phase one are merged with the phase three knots derived from GAPS. As the number of knots is relatively small compared to the number of AUV navigation data points, the track of the vehi- cle between the knots has to be approximated. As in phase one, a linear increase of the navigation error is assumed and a similar approach can be applied. By means of the time stamp, particular positions of the AUV P1 data set can be assigned to knots and can be relo- cated onto them. These particular positions in the P1 data set are further referred to as "fxpoints." This cuts the P1 data set into sections with a fxpoint representing the beginning of each section. Maintaining their original orientation and length, every section and the associated fxpoint are moved via parallel translation until the fxpoint is located on its re- spective knot. The algorithm calculates the error (distance and orientation) between a knot and the last position of the previous section. According to the assumed linear increase of the navigation error, this measured error is then distrib- uted on the elements of the section—rotating the section around its fxpoint and stretching it until it bridges the space between two consecutive knots. Using this approach from knot to knot, the entire track of the vehicle can be reconstructed piecemeal. Discussion The correction algorithm described in this article is an attempt to merge the advantages of two error-containing sources for navigation data, so that the result is a reliable reconstruction of the actual AUV track. However, it is to "It is clear that the algorithm tremendously improves the quality of the navigation data. Meanwhile, the navigation errors due to the foat maneuver have been minimized."