Exploring the sensor requirements for particle filter-based terrain-aided navigation in AUVs

Abstract

One of the main challenges faced by autonomous underwater vehicles (AUVs) is their navigation capability. A commonly used navigation strategy is dead-reckoning inertial navigation integrated with a seafloor-tracking Doppler velocity log (DVL). The disadvantage with this technique is the un-bounded accumulation of errors during operation. Terrain-aided navigation (TAN) has been deemed as a viable approach to compensate for dead-reckoning navigation errors. This type of navigation uses a predefined map of the seafloor and estimates the position of the vehicle on the map by matching the map to bathymetry measurements from vehicle sensors.The physical limitations of an AUV platform places restrictions on what kind of sensors can be used, such as weight, size, resolution, power consumption and accuracy. For this reason, a dedicated TAN-sensor is undesirable. In addition to a DVL, many vehicles make use of a forward-looking sonar (FLS) for obstacle avoidance. In this work, we implement a particle filter-based TAN algorithm using DVL altitude data fused with FLS information. We simulate TAN navigation for several different types of FLS, and compare the navigation performance. In general, we find that navigation performance improves with the number of sonar beams. This is expected, as a more complex sensor means more information is gathered and input to the navigation. However, we also find that a configuration using just three single-beam echo sounders performs as well as a proper multibeam FLS.The results show that the multiple single-beam echo sounder configuration is an interesting solution due to its high accuracy during the navigation while retaining a low impact on the platform dynamics and power consumption.