Adaptive sampling algorithms for multiple autonomous underwater vehicles

Abstract

-Sampling is a critical problem in the observation of underwater phenomena using single or multiple AUV platforms. The determination of optimal paths and sampling strategies that effectively utilize available resources is critical to these missions. Recent work performed jointly at RPI and AUSI on the development of Adaptive Sampling Algorithms (ASA) utilizes information measures, estimation theory, and potential fields to direct the robots to the locations in space most likely to yield information about the sensed field variable of interest. Typical sensory information can consist of spatial distribution of one or more field variables, such as salinity, dissolved oxygen, temperature, current, etc. In order to test our algorithms we have created the MATCON simulation environment, an underwater experimental platform using solar AUVs, and a land-based experimental testbed using inexpensive wheeled robots.

A Mechanism to Structure Mission-Aware Interaction in Mobile Sensor Networks

“…5) is an example of a security or law-enforcement AUV mission where a line of AUVs monitor a passageway for unusual activity [15]. Another example of a mission is inspired by the RiverNet project [16] where AUVs are used to monitor pollutants in the Hudson River. Fig.…”

Information-theoretic integration of sensing and communication for active robot networks

“…The main drawback of this category of algorithms is failure to consider uncertainty in sensor information. The second category of algorithms, often referred to as adaptive sampling, focus on the directed sampling of a scalar field in order to obtain certain statistical properties or to locate sources, contours, and boundaries [5], [16]- [18]. The main drawback of these adaptive sampling approaches is their reliance on local information only and failure to develop and consider models of the underlying phenomena being measured.…”

“…The concept of using system inputs to improve the quality of information sensed by the system has been applied to a variety of problems. Example applications include: target tracking [2], [20], [21]; collaborative perception by unmanned ground and aerial vehicle teams [15]; vision-based navigation and control [22]; hierarchical reconnaissance by teams of unmanned aircraft [23] ; plume tracking by multiple autonomous underwater vehicles [5]; and cooperative radio source localization [24].…”

“…Future robotic sensor networks are expected to provide ubiquitous perception for a wide variety of applications. These applications include in-situ volumetric sensing [1], multi-target surveillance [2], chemical plume detection and tracking [3], autonomous underwater networks [4], [5], and networked communication relaying [6].…”

“…One approach was developed for adaptive sampling by multiple autonomous underwater vehicles [5]. Algorithms developed for this application combine one-step look-ahead optimization of the Kullblack-Liebler divergence of estimates of scalar fields with network flow control algorithms that weight link utility based on the Shannon-Hartley Capacity [25].…”