This paper is about the autonomous control of an autonomous underwater vehicle (AUV), and the particular considerations required to allow proper control while towing a 100-meter vector sensor array. Mission related objectives are tempered by the need to consider the effect of a sequence of maneuvers on the motion of the towed array which is thought not to tolerate sharp bends or twists in sensitive material. We describe and motivate an architecture for autonomy structured on the behavior-based control model augmented with a novel approach for performing behavior coordination using multi-objective optimization. We provide detailed in-field experimental results from recent exercises with two 21-inch AUVs in Monterey Bay California.
This paper surveys the developments of last 10 years in the area of vision based target tracking for autonomous vehicles navigation. First, the motivations and applications of using vision based target tracking for autonomous vehicles navigation are presented in the introduction section. It can be concluded that it is very necessary to develop robust visual target tracking based navigation algorithms for the broad applications of autonomous vehicles. Then this paper reviews the recent techniques in three different categories: vision based target tracking for the applications of land, underwater and aerial vehicles navigation. Next, the increasing trends of using data fusion for visual target tracking based autonomous vehicles navigation are discussed. Through data fusion the tracking performance is improved and becomes more robust. Based on the review, the remaining research challenges are summarized and future research directions are investigated.
A multiple feature tracking algorithm for Sector Scan Sonar images is presented. The underlying framework that is employed is the Random Finite Sets (RFS) approach which is the natural representation of the multi-target states and observations. There has been a strong mathematical foundation laid to this approach using Finite Set Statistic(FISST). However, the propagation of the full multi-target posterior distribution using the optimal Bayesian approach is not yet practical due to computational hurdles. A practical alternative to the optimal Bayesian multi-target filter based on RFS is the probability hypothesis density (PHD) filter. PHD is a first order statistical moment of the full multi-target posterior distribution. This paper deals with feature detection and estimation techniques using the particle-PHD filter. The ability to track features in heavy clutter is demonstrated first using the simulated data and then with the real sector scan sonar data.
Underwater surveillance is experiencing a paradigm shift from platform-centric to distributed networks of autonomous underwater vehicles (AUVs). In order to achieve the performance goals with smaller apertures carried by single platforms, new concepts of operations must be developed, which exploit platform mobility to adapt to the changing environmental conditions. Unlike the network of sensor nodes used in air/land, underwater networks cannot rely on the low channel capacity of the acoustic communication available. The operation of such an underwater sensor network requires a much higher level of autonomous, distributed command and control architecture. This paper discusses the proposed nested autonomy architecture developed at MIT for command and control of AUVs.
16 January 2009 in Recent Patents on Computer Science
The objective of SWAMSI is the development and demonstration of robust multi-static detection and classification of proud- and buried seabed objects using cooperative networks of autonomous vehicles with acoustic sources and receiving arrays. The emphasis of the MIT SWAMSI effort has focused on utilizing high fidelity acoustic modeling of both scatterers and shallow-water environments to better understand and bound the limits of detectability for mine-like objects via autonomous networks of sensors, and the assess the performance of time-reversal processing for concurrent detection, classification, localization and Tracking (DCLT) of seabed objects. The analysis s supported by series of experiments using multiple sonar-equipped AUVs in shallow water and then cross-validate the results obtained with high precision modeling and visualization. Another, related objective is to better understand the problems of cooperative autonomous vehicle interaction to define the base-line infrastructure requirements for cooperative detection, classification and navigation, an understanding which may lead to guidelines for optimal collaborative configuration control of the underwater sonar platforms.