Autonomous Underwater Vehicles

Navigation is one of the most critical factors in determining the operational suitability of any unmanned vehicle for its designated environment. In fully autonomous vehicles, due to the lack of a human operator to perform the navigation task, there is a fundamental requirement to incorporate estimation techniques that can provide the desired information necessary for navigation. Such information includes position, attitude, and velocity of the vehicle. In most vehicles, estimating this information is a relatively straight forward process due to the information available from the measurement sensors, particularly position information from Global Positioning System (GPS) sensors. This chapter will focus on the more difficult problem of navigation, particularly positioning, where GPS reception is limited or non-existent. One example where GPS is only partially available is the case of an Autonomous Underwater Vehicle (AUV) where the vehicle is forced to use dead-reckoning in between GPS sightings in order to navigate accurately. Though an underwater environment is used within the context of this example, the techniques examined can be equally applied anywhere where GPS reception is compromised by other electromagnetically opaque mediums, satellite availability or even deliberate denial. When using dead reckoning, due to the integrative nature of the position estimate, the error of the position estimate will grow unbounded. The growth in this error is caused by errors in the velocity and attitude estimates which are in turn affected by the accuracy of the navigational filter and the accuracies of the measurements observing these states. The accuracy of the measurement sensors is typically related to cost, thus some of the effects can be negated through more precise (expensive) equipment. For a given cost the single biggest factor affecting the accuracy of the navigational estimate is the navigation filter itself. There have been numerous papers on the accuracy of various algorithms employed in the generic problem of state estimation (Arulampalam et al., 2001). Research pertaining to navigation algorithms for underwater vehicles have almost exclusively concentrated on ways to improve positioning accuracy below the surface through the use of ranging information other than GPS (Caiti et al., 2004). The methodology most prevalent in the literature is Simultaneous Localization And Mapping (SLAM) (Choset & Nagatani, 2001; Dissanayake et al., 2001)where information used for mapping, such as sonar ‘pings’, is fed 23