The objectives of this project are as follows: 1) to implement a re-designed version of the structural acoustic (SA) sonar recently demonstrated in Project MR-201714 on a smaller, man-portable autonomous undersea vehicle (AUV) and to do so without causing any significant detection and classification performance degradation relative to that achieved with the larger Bluefin 21 AUV sonar system; (2) to incorporate improved technologies and processing methods successfully demonstrated in the additional tasks added to the final phase of MR-201714; and (3) to demonstrate the detection and classification performance of both the smaller, less-expensive and easily handled system and the larger Bluefin21 system in blind tests carried out at the SERDP/ESTCP Sequim Bay demonstration test site.

Technology Description

A primary innovation in the technology is the exploitation of acoustic color (the echo’s spectral content versus angle) swept out by the AUV. The SA sonar’s low frequencies penetrate the sediment to buried targets whose echoes contain features separating unexploded ordnance (UXO) and non-UXO. Program MR-201714 demonstrated a winged AUV-SA down-looking sonar against buried targets and a complementary interior array AUV-SA sonar looking out sideways to greater distances providing higher coverage rate detection/classification against proud and partially-buried UXO. Prior to a test, the acoustic color spectra and time-angle plots for training the classification algorithms are generated from echo measurements from a number of placed UXO at the actual or similar site or from available numerical data. At the test site, after deploying several Lincoln hat navigation monuments on the sediment surface, echo data is collected over a synthetic/real planar array by flying the AUV-based structural acoustic sonar over the target field in a mow-the-grass fashion. Flight paths include east-west, north-south, and two diagonals 90 degrees apart each separated by between 2m and 20m. The data is then post-processed off-line where detection display maps are formed from volumetric images projected onto a two-dimensional (2D) plane parallel to the sediment-water interface. The echo data associated with above-threshold detections are further processed to produce the classifying acoustic color/time-angle plots. This is implemented on two levels, i.e. with single-path detection surfaces and with multi-path, global object detection surfaces. The latter are constructed from echoes collected over many single-paths having various angular target aspect views in order to obtain full 360 degree echo patterns. To ensure high precision, the detection surface from each line is re-navigated by translating each 2D detection surface so that clearly observed proud Lincoln hat monuments (and/or notable indigenous features) align on a universal grid. The classifying features are then extracted from the multi-static target constructs and inputted into the trained Relevance Vector Machine algorithms which in turn output target calls. In the case of the down-look system, the volumetric images are also projected to form three 2D images which provide information about target size, rough shape, burial angle, and burial depth. In addition, an “expert” using 360 degree acoustic color and time-angle display templates obtained from the training can corroborate calls from comparisons to constructs from the test data together with the observed agreement of the 2D image details with UXO shapes and sizes.


The project team estimates that more than 2 million acres (30%) of the more than 400 underwater sites potentially containing munitions involve water depths applicable to the AUV-based SA technology, and the demonstration in the Boston Harbor’s difficult environment validates its relevance to many Department of Defense sites. The technology’s impact would be considerable given its unique ability to prosecute underwater buried UXO. This project will extend this success to smaller, less-costly man-portable AUVs.