AUV Geophysical Navigation using Magnetic Data - the Medusa GN System
João Quintas, Institute for Systems and Robotics, Instituto Superior Técnico, University Lisboa, Portugal; Francisco Curado Teixeira, Institute for Systems and Robotics, Instituto Superior Técnico, University Lisboa, Portugal; António Pascoal, Institute for Systems and Robotics, Instituto Superior Técnico, University Lisboa, Portugal
Location: Big Sur
True autonomous navigation of underwater robotic vehicles is still a challenging problem due mainly to the operational difficulties posed by the underwater environment. Without access to the Global Position System (GPS) and in the absence of acoustic beacons whose deployment is costly and complex, the navigation of autonomous underwater vehicles (AUV) is normally performed by dead-reckoning (DR). Conventional DR based on the integration of data from an attitude and heading reference system (AHRS) and a Doppler velocity logger (DVL) is a convenient technique for relatively short-range and time-limited navigation but, as is well-known, the position error resulting from this solution will grow over time, making it difficult to achieve precise navigation when dead-reckoning is used for a long period of time.
Over the past few years, geophysical navigation (GN) has been proposed as the method of choice to aid AHRS/DVL navigation of underwater vehicles, due the relatively low-cost of GN implementations and its potential to provide accurate estimates of position in the short term combined with bounded localization errors in the long run. In order to estimate the position of robotic vehicles, GN methods exploit geophysical effects observed in the environment, such as terrain topography, gravity, and the geomagnetic field. A conventional implementation of underwater geophysical navigation is the well-studied terrain-aided navigation (TAN) method which relies exclusively on matching a set of range measurements acquired with sonar sensors installed on a vehicle with a previously acquired digital elevation map (DEM) of the terrain to estimate position. The work of Nygren and Jansson , Anonsen and Hallingstad , and Morice et al. , among others, demonstrated experimentally the potential of the TAN solution in different types of terrain and with distinct sensor suites. Despite these successful implementations, the performance of terrain-based navigation systems depends strongly on the availability of topographic information. It is well known that large areas of the ocean floor are characterized by a very smooth, mostly flat topography where GN implementations become ineffective due to the the lack of “exciting” terrain features. To implement geophysical navigation solutions in this type of environments it is necessary to exploit other types of information, such as that available from the Earth´s gravitic and magnetic fields, which may lead to alternative GN solutions capable of yielding better positioning accuracy.
This paper describes the work done towards the implementation of a magnetic-based geophysical navigation module that can be integrated with the dead-reckoning navigation systems of the Medusa-class of hybrid AUV/ASV, developed and operated by the Institute for Systems of Robotics of Instituto Superior Técnico (ISR-IST), Lisbon, PT. The new magnetic navigation module (MAGNAV) constitutes a particular implementation of the GN approach that exploits the information associated to local, time-invariant geomagnetic field anomalies that are measured by a marine robotic vehicle. The MAGNAV approach can be used to replace or to aid classical terrain-based navigation methods in areas where the scarcity of terrain-elevation features does not permit accurate vehicle position estimates using topographic information only. The rationale for this approach arises from the fact that important magnetic anomalies are often found in areas of the seafloor characterized by insufficiently excited terrain topography. As such, terrain topography and geomagnetism can be envisioned as complementary sources of terrain information with high potential to support the implementation of versatile geophysical navigation systems.
The present work proposes economical solutions to the main problems posed by magnetic-based navigation, including the suppression or the compensation of ambient magnetic disturbances and the mitigation of vehicle-induced electromagnetic noise that affect real-time measurements of the geomagnetic field. Relying on prior work by the authors , the paper proposes a Rao-Blackwellized particle filter (RBPF) as the core navigation algorithm. The RBPF mechanizes a data fusion procedure using a marginalized particle filter which estimates the 2D position of the vehicle as well as the 2D components of the velocity bias incurred by dead-reckoning. The RBPF is implemented as a single ROS node in the vehicles software architecture that combines a vehicle motion model fed by data from an AHRS and a DVL and a measurement model associated with a total field magnetometer.
In this work, experimental data were acquired at sea with a Medusa AUV/ASV at a site of approximately 500x500m, offshore S. Pedro do Estoril, near Lisbon. In a preliminary phase of the experiments, the trial area was surveyed with a total field magnetometer in order to obtain a prior total magnetic field map with 1m resolution or better. This map was later used to test the MAGNAV approach hereby proposed. During the sea trials, the Medusa vehicle was operated as an autonomous surface vehicle (ASV) from a support vessel maneuvering in its proximity, connected to the vehicle via a wireless communication link, thus allowing for real-time monitoring of the vehicle´s state and the quality of terrain-related data. Three experiments were conducted at the site: one corresponding to a rectangular trajectory of 200m*150m; another in the form of a square 300m*300m; and, finally a lawnmowing maneuver. All sensor and control messages were recorded in a ROS bagfile, thus ensuring that all messages were associated with their corresponding time-stamps. Based on the information acquired, and using standard ROS tools, the data can be played back and the Medusa navigation experiments can be simulated with the real data as if the vehicle were operating in real-time.
Results and Conclusions
The main purpose of the work here described was twofold: to evaluate the performance of magnetic-based navigation algorithms that are now integrated in the MEDUSA vehicle's software architecture and ii) to assess their performance using real data collected at sea in a scenario characterized by a large variability of magnetic features. To this end, the algorithms developed in our prior work [5, 6] were integrated within the software architecture of a Medusa AUV/ASV and applied to the new magnetic data-sets acquired during oceanic trials. The results now obtained confirm the good performance of the MAGNAV filter in terms of position estimation and demonstrate the potential improvement of classical TAN implementations that can be achieved through the exploitation of magnetic information.
The results obtained will pave the way for the implementation and testing of a GN module in a Medusa AUV in the near future.
 Nygren, I. and Jansson, M. (2004). Terrain navigation for underwater vehicles using the correlator method. IEEE Journal of Oceanic Engineering, 29(3), 906–915.
 Anonsen, K.B. and Hallingstad, O. (2006). Terrain aided underwater navigation using point mass and particle filters. In IEEE/ION Position Location and Navigation Symposium, 1027–1035. San Diego, CA.
 Morice, C., Veres, S., and McPhail, S. (2009). Terrain referencing for autonomous navigation of underwater vehicles. In MTS/IEEE OCEANS’09 Europe. Bremen, GE.
 Teixeira FC, Quintas J, Pascoal A,. AUV Terrain-Aided Navigation using a Doppler Velocity Logger. IFAC Workshop on Navigation, Guidance and Control of Underwater Vehicles (NGCUV’ 2015), Girona, Spain, 2015.
 Quintas, J, Teixeira, F.C., and Pascoal, A. (2016). Magnetic signal processing methods with application to geophysical navigation of marine robotic vehicles. In OCEANS 2016 MTS/IEEE Monterey.
Teixeira, F.C., Quintas, J., and Pascoal, A. (2016b). Experimental validation of magnetic navigation of marine robotic vehicles. In 10th IFAC Conference on Control Applications in Marine Systems - CAMS’2016.