AUV-Based Chemical Plume Tracing

By Pang, Shuo Farrell, Jay A; Li, Wei

Development and Demonstration in Near-Shore Ocean Conditions Many animals (e.g., lobsters, blue crabs, ants and moths) are capable of using olfaction for foraging or reproductive activities in a turbulent fluid-flow environment. This ability may seem fairly trivial, as most people have responded to the scent of barbecue or baking bread by turning upwind. However, in a turbulent fluid flow, the turbulence of the fluid continuously stretches and twists the filaments of the chemical plume, and the changing flow vector causes the plume centerline to meander. The resulting fine-scale structure of the sensed odor, even at a fixed location, is highly intermittent. Due to turbulence and flow variation, chemical plume tracing (CPT) over significant distances, and source declaration based on olfaction alone, is not trivial.

For many military and civilian applications in a turbulent fluid- flow environment, it would be useful to detect and track a chemical plume to its source. The U.S. Office of Naval Research (ONR) Chemical Sensing in the Marine Environment and ONR/Defense Advanced Research Projects Agency (DARPA) CPT programs were intended to develop and demonstrate such capabilities using autonomous underwater vehicles (AUVs). These AUV capabilities would be of great significance for many applications, such as the detection of chemical leaks, locating unexplored ordinances and locating biologically interesting phenomenon like thermal vents.

This article describes the development and field test of a planning and guidance system for an AUV to find a chemical plume, trace the chemical plume to its source, reliably declare the source location and map the plume source area after source declaration. The AUV CPT system includes an adaptive mission planner (AMP) that automatically responds to the sensor inputs to generate a trajectory for the AUV. Hydroid Inc.’s (Pocasset, Massachusetts) REMUS navigation system performs dead-reckoning based on acoustic Doppler data with periodic navigation fixes based on data from a long- baseline acoustic transponder system. A guidance system transforms the AMP trajectory commands into depth, speed and heading control system commands that are within the maneuverability constraints of the AUV (i.e., speed less than two meters per second and a heading rate of less than 10[degrees] per second).

Chemical Plume Tracing

The basic idea of AUV CPT is that an AUV is constrained to maneuver within a region, referred to as the OpArea. Within the OpArea, the vehicle should search for a specified chemical for which a chemical sensor is available. If the above threshold chemical is detected, the vehicle should perform maneuvers designed to maintain intermittent contact with the plume, trace the chemical plume to its source and declare the source location accurately. Following the source declaration, additional vehicle maneuvers might be performed to acquire additional data, possibly using auxiliary sensors. Example post-declaration maneuvers include driving in the direction of the computed flow (past the declared source location to capture the AUV near the source on video imagery), performing maneuvers in the vicinity of the declared source location (to acquire side scan sonar data useful for ground truth evaluation of source declaration accuracy) and performing maneuvers perpendicular to the computed flow in the region downstream from the declared source location (to acquire data indicative of the chemical concentration as a function of source relative location).

Initial approaches to the CPT problem often focus on extremum- seeking approaches, such as gradient following. While such approaches may have applicability in laminar-flow environments, they have limited utility in turbulent-flow situations. The result of the turbulent diffusion process is a highly discontinuous and intermittent distribution of the chemical. This results in two difficulties. First, if the AUV did have a local map of the chemical distribution in its vicinity, that map would contain many local maxima and have a gradient vector that varies rapidly with position. Second, the AUV does not have access to such a map. The onboard chemical sensor samples the concentration at a single point that moves with the vehicle, and the chemical distribution changes rapidly relative to the rate at which the AUV can maneuver. The above issues could be addressed by various methods, including additional vehicles, additional sensors and sensor averaging. However, biological entities are proofs of concepts capable of performing CPT with a single searcher, very similar sensor suites and maneuvering capabilities.

The architecture for the AUV-based CPT includes a control system, a guidance system, a mapping system and an AMP. The task of the mapping system is to accumulate flow and chemical detection information over the duration of the mission and to form an estimate of a source likelihood map. The AMP incrementally determines a desired vehicle trajectory based on the history of flow and detection information. The guidance system outputs heading, speed and depth commands to the controller to achieve the trajectory determined by the AMP without violating the heading rate and velocity constraints. The AUV was commanded at a constant altitude (one to two meters) to allow the CPT strategy to work in two dimensions. A main motivation for implementing the algorithms in two dimensions is the computational simplification achieved. However, neutral buoyancy of the chemical and stratification of the flow often yields a plume which may be approximated as 2D.

The CPT experiments were performed using the REMUS AUV, owned by the U.S. Navy’s Space and Naval Warfare Systems Command in San Diego, California. The REMUS was modified to contain a PC 104 computer to run the AMP and guidance algorithms. The AMP computer received sensor data from the REMUS main computer via a serial connection, processed the sensor data and supplied heading, speed and depth/altitude commands to the main computer via the same serial connection. Up and down-looking acoustic Doppler current profilers were onboard the REMUS. Finally, a fluorometer capable of detecting Rhodamine dye was mounted near the nose of the AUV. Rhodamine dye was used to create the plume for the experiments. The maximum speed of the REMUS AUV is 2.6 meters per second. The missions were normally performed with the vehicle traveling at a constant speed of 1 .5 to two meters per second. The maximum turning rate of the vehicle is 15[degrees] per second. The guidance system limits the turning rate to 10[degrees] per second.

CPT Behavior-Based Design

The AMP operated by switching between six biomimetic behaviors. The go-to behavior was implemented to allow the vehicle to be driven to a desired location. Once the go-to behavior was completed, the find behavior searched the OpArea trying to detect the presence of a chemical. Whenever a chemical was detected, AMP switched to track- in behavior, which tried to steer the vehicle toward the source while maintaining intermittent contact with the plume. As the time since the last chemical detection increased, AMP switched to track- out behavior, then re-acquire behavior and, finally, back to find behavior. The track-in, track-out and re-acquire behaviors were all motivated by behaviors observed in trajectories of moths tracking pheromone plumes.3,4

Field Testing

Different variations of CPT algorithms were tested in three different sets of experiments. The first CPT algorithm was tested off San Clemente Island (SCI), California, in November 2002.5 This was the first in-water CPT evaluation. Based on these experimental results, several behaviors were revised, the AMP post-declaration maneuvers were added and the guidance system was designed. The AMP described herein was demonstrated at SCI in April 2003 and off Duck, North Carolina, in June 2003.3

In the SCI field test during the period of April 29 to May 2, 2003, eight vehicle runs successfully performed CPT and declared the source location with eight to 1 7 meters’ accuracy relative to ground truth. The experiments included ground truth confirmation of declared source locations via side scan sonar. Several post- declaration flyby maneuvers were captured on video. In fact, one flyby maneuver resulted in a collision between REMUS and the source that was caught on video.

The June 2003 test off Duck included seven runs. The mission area was 367 by 1,094 meters-significantly larger than the mission area of previous in-water tests. In this field test, two types of missions were of interest. The first mission type, labeled ST, contained a single chemical source in the OpArea. The ST mission was intended to find the plume, trace it over a long distance and declare the source location. In run MSN003 of this test, me AUV traced the chemical plume and declared a source location within 13 meters of the true source location, as confirmed by side scan sonar data. The distance between the first point of chemical detection and the declared source location was 975 meters. The second mission type, labeled MT, might contain a few chemical sources in the OpArea. In an MT mission, the OpArea is divided into subregions. The AUV searches each subregion for chemicals until one of three events occurs. First, me search within a subregion may timeout. In this case, the subregion is declared source-free and the AUV moves on to the next subregion. Second, the AUV may detect a chemical and declare a source location within the region. It will then move on to the next subregion. Third, the AUV may trace a chemical to the upflow edge of the region. In this case, a source will be declared at the intersection of the plume. The upper edge of the subregion and the AUV would then move on to the next subregion. When the declared source locations are analyzed at the end of an experiment, it is up to the test director to decide whether source locations at the edge of a subregion are due to sources near that location or the result of plumes generated by sources in the adjacent region. Acknowledgements

Ring Carde and John Murlis were key collaborators while developing the theory for CPT under the ONR/DARPA CPT program. Rich Arrieta, Jeff Deschamps, Vladimir Djapic, Andy Dreiling, Brian Granger, Paul Holland, Gerry Hong, Bill Morris, Phil Selwyn, Ken Vierra and U.S. Navy divers were key collaborators during the experimental phase of this effort.

Roger Stokey and Greg Packard at Woods Hole Oceanographie Institution were instrumental in getting the AMP on, and communicating with, the REMUS, /st/

REMUS vehicle swimming in the water.

“For many military and civilian applications in a turbulent fluid- flow environment, it would be useful to detect and track a chemical plume to its source.”

Visit our Web site at www.sea-technology.com and click on this article in the Table of Contents to be linked to the respective company’s Web site.

References

1. Pang, S. and J. Farrell, “Chemical Plume Source Localization,” Institute of Electrical and Electronics Engineers Transactions on Systems, Man and Cybernetics, Part B: Cybernetics, vol. 36, pp. 1,068-1,080, 2006.

2. Stacey, M., E. Cowen, T. Powell, E. Dobbins, S. Monismith and J. Koseff, “Plume Dispersion in a Stratified, Near-Coastal Flow: Measurements and Modeling,” Continental Shelf Research, pp. 637- 663, 2000.

3. Farrell, J., S. Pang and W. Li, “Chemical Plume Tracing Via an Autonomous Underwater Vehicle,” Institute of Electrical and Electronics Engineers Journal of Ocean Engineering, vol. 30, pp. 428- 442, 2005.

4. Li, W, J. Farrell, S. Pang and R. Arrieta, “Moth-Inspired Chemical Plume Tracing on an Autonomous Underwater Vehicle,” Institute of Electrical and Electronics Engineers Transactions on Robotics, vol. 22, pp. 292-307, 2006.

5. Farrell, J., W. Li, S. Pang and R. Arrieta, “Chemical Plume Tracing Experimental Results with REMUS AUV,” Proceedings of the Marine Technology Society/Institute of Electrical and Electronics Engineers Oceans 2003 Conference, pp. 962-968, 2003.

By Shuo Pang

Assistant Professor

Department of Computer and

Software Engineering

Embry-Riddle Aeronautical University

Daytona Beach, Florida

Jay A. Farrell

Professor

Department of Electrical Engineering

University of California

Riverside, California

and

Wei Li

Professor

Department of Computer Science

California State University

Bakersfield, California

Shuo Pang received his M.S. and Ph.D. in electrical engineering from the University of California, Riverside, in 2001 and 2004, respectively. Currently, he is an assistant professor at Embry- Riddle Aeronautical University.

Jay A. Farrell received his B.S. (1986) in physics and electrical engineering from Iowa State University, and his M.S. (i988) and Ph. D. (1989) in electrical engineering from the University of Notre Dame. He is a professor of electrical engineering at the University of California, Riverside.

Wei Li received a Ph.D. in electrical and computer engineering from the University of Saarland in 1991. Currently, Li is a professor at California State University, Bakersfield.

Copyright Compass Publications, Inc. Jul 2007

(c) 2007 Sea Technology. Provided by ProQuest Information and Learning. All rights Reserved.