UDT 2020 Extended Abstract Ånonsen Session Autonomy at sea Terrain navigation techniques for AUV MCM operations K. B. Ånonsen 1 , O. K. Hagen 2 and H. S. Telle 3 1 Senior Scientist, Norwegian Defence Research Establishment (FFI), Kjeller, Norway, email:kjetil-bergh.anonsen@ffi.no 2 Principal Scientist, Norwegian Defence Research Establishment (FFI), Kjeller, Norway 3 Commander, Royal Norwegian Navy, Bergen, Norway Abstract — Terrain navigation techniques, which use measurements of the sea floor together with a digital terrain model (DTM) to obtain position updates to the navigation systems, is an attractive technique in mine counter measure (MCM) operations with autonomous underwater vehicles. We show how terrain navigation can be used to facilitate submerged MCM operations without the need for surfacing for GNSS fixes or pre- deployed infrastructure on the sea floor. The concept is demonstrated using test data from one of the Real Norwegian Navy’s newly acquired Kongsberg HUGIN AUVs. scan or SAS system [2], to be able to conduct terrain 1 Introduction navigation. Over the last two decades autonomous underwater An AUV MCM scenario typically consists of three vehicles (AUVs) have proven to be highly efficient tools phases: a survey phase, a detection and classification for conducting underwater mine counter measure (MCM) phase and an identification phase . During the survey operations. One of the fortes of the AUV is that it can enter phase, the operation area is mapped using a side-scan or the operation area covertly, without the need for a surface SAS system. The detection/classification phase was vessel following it into the possibly dangerous area. The traditionally carried out by a human operator, but the success of such operations is dependent on high accuracy development of automatic target recognition (ATR) AUV navigation estimates, i.e. vehicle position and systems has automated this process to a large extent [3]. attitude estimates, to be able to determine the correct The identification phase can be carried out using an optical positions of observed mine-like objects of interest on the camera on the AUV, revisiting the detected targets. This sea floor. phase can either be conducted in a separate run, after the Modern AUVs partially solve the navigation problem survey data have been processed, or be integrated with the by using inertial navigation systems, which are aided by other phases in a single sortie [4]. In either case, a pressure sensors and Doppler velocity logs whenever the previously surveyed area is revisited during the AUV is submerged and GNSS signals are not available. In identification phase, which can be utilized by the terrain extensive submerged operations, the AUV will still need navigation system. external position updates in order to keep the navigation accuracy sufficiently high. As surfacing for GNSS fixes in 3 AUV navigation system many cases is impractical and revealing, terrain navigation, in which measurements and knowledge of the terrain are combined to obtain a position estimate, is an 3.1 Inertial navigation system (INS) attractive alternative in many scenarios. Modern AUVs are equipped with inertial navigation systems (INS), in which measurements from inertial sensors (accelerometers and gyroscopes) are combined 2 AUV MCM Operations with a suite of aiding sensors to counter the inherent INS drift. One example is the Kongsberg HUGIN AUV We here focus on mine hunting operations with AUVs navigation system [5]. equipped with high-resolution side-looking sonar systems (e.g. synthetic aperture sonars (SAS)) capable of locating mine-like objects, in addition to optical cameras for 3.1.1 Core INS identification of the contacts. The AUV can either be run The core INS of a modern AUV consists of an Inertial from a manned surface vessel, or be part of a fully Measurement Unit (IMU), a pressure sensor and a Doppler unmanned MCM system operating with an unmanned velocity log. This system is typically initialized using surface vehicle (USV), in accordance with [1]. In addition GNSS measurements before diving. While the AUV is to the abovementioned sensors, the AUV must be submerged and GNSS is not available, the core INS equipped with a bathymetric sensor, preferably a accuracy will degrade with time due to integration of multibeam echo sounder (MBE) or interferometric side- measurement noise. In [6], a submerged drift rate of 0.04-
UDT 2020 Extended Abstract Ånonsen Session Autonomy at Sea 0.08 % of traveled distance was reported. In order to maintain sufficient navigation accuracy during longer autonomous operations, the vehicle must either surface for a GPS fix or use alternative aiding methods like underwater transponders [7] or terrain navigation [8-10]. 3.1.2 Terrain navigation Terrain navigation algorithms use measurements of the terrain below or close to the vehicle, together with a digital terrain model (DTM), to find a position estimate, which can be used as a submerged update to the INS. Originally developed for aircraft and cruise missiles, terrain navigation has gained popularity for underwater vehicles over the last two decades and is now an optional add-on to the Kongsberg HUGIN AUV family [10]. This system uses a terrain navigation algorithm based on the point mass filter algorithm [11]. As a rule of thumb, in suited terrain, one can expect a horizontal position accuracy comparable Fig. 1 : Trajectory of the AUV mission overlaid the DTM. to the horizontal resolution of the DTM In their standard form, the terrain navigation The AUV navigation system was initialized using GNSS algorithms are dependent on the existence of a DTM prior before diving and the vehicle was then run submerged for to the operation. However, this requirement can be relaxed approximately 20 hours with terrain navigation as the sole by using SLAM (Simultaneous Localization and Mapping) position update method. After approximately 20 hours, the [12, 13] or SLAM-like techniques. As an example, in an vehicle surfaced for a GNSS fix. The differences between MCM scenario for which no prior DTM exists, the vehicle the real-time estimated position (using terrain navigation) can use a bathymetric sensor to build a DTM during the and the best available post-processed navigation solution survey phase of the mission. When later returning to the are shown in Fig. 2 , together with the estimated real-time same area, e.g. in order to make an optical camera navigation uncertainty. The post-processed navigation identification of detected mine-like objects, the AUV can solution was computed using the navigation processing use this in-situ DTM for terrain navigation [14, 15]. tool NavLab [16], and is a smoothed solution using all the GNSS, IMU, DVL and pressure measurements throughout the mission, but not the terrain navigation updates. Since this solution has a period of more than 20 hours without 4 Results position updates, its uncertainty is quite high and should not be regarded as ground truth. However, as can be seen We here consider a test mission done with one of the Royal from Fig. 2 , the difference between the post-processed and Norwegian Navy’s newly acquired Kongsberg HUGIN real-time navigation estimates stayed within the 1 sigma AUVs. The test was conducted in the Oslo Fjord in band of the real-time navigation system throughout most Norway in January 2020, and was planned as a typical of the run, indicating that the real-time navigation did not MCM mission, in which the vehicle first ran several search suffer from serious wild-points that were falsely accepted patterns with synthetic aperture sonar, before performing by the terrain navigation system. an identification phase in which the optical camera was used for identification of interesting targets. Kongsberg EM2040 multibeam data from an earlier AUV mission in the area was used to build a 5 by 5 meter gridded DTM, for use in the terrain navigation system. This DTM covered parts of the mission plan for our test, allowing the AUV to obtain terrain navigation fixes to the navigation system when passing over the mapped areas. The vehicle trajectory and the DTM are shown in Fig. 1 .
Recommend
More recommend
