By: Ryan N. Smith,

1
AUV Trajectory Design Based on Ocean Model
Predictions

• By Ryan N. Smith,
• Yi Chao, Burton H. Jones, David A. Caron, Peggy
  P. Li and Gaurav S. Sukhatme

7th International Conference on Field and Service
Robots Cambridge, MA July 15, 2009
2
The USC Center for Integrated Networked Aquatic
PlatformS

• CINAPS sin-aps
• Bridging the gap between technology,
  communication, and the scientific exploration of
  aquatic ecosystems.
• Collaborative research effort between RESL,
  usCLAB and Caron Lab at USC
• A main area of CINAPS research is addressing
  scientific questions regarding the formation,
  propagation and prediction of Harmful Algal
  Blooms (HABs).

3
Robotic Embedded Systems Laboratory (RESL)?

• Design, implement and understand large-scale,
  distributed, robotic systems.
• Aggregate mobile robots and unattended sensors
  into a sensor network
• Applications in urban security, military
  reconnaissance, and environmental monitoring.

4
Introduction

• Develop an innovative sampling method
• Utilize ocean model predictions
• Generate trajectories for AUVs that track an
  ocean feature
• Allow for the collection of data that is both
  meaningful to the oceanographic community as well
  as increases the skill of the predictive model.

5
Goals

• Design and implement an innovative technology
  chain for AUV trajectory design
• Multi-agency collaboration
• USC -gt JPL -gt Field Robot -gt JPL -gt USC
• Develop basic algorithms for feature tracking by
  an AUV utilizing ocean model data
• Demonstrate the ability to task, assimilate data
  and retask AUVs currently operating in the field
  for the purpose of feature tracking

6
Oceanography Motivation

• Harmful Algal Blooms (HABs)?
• Rapid growth of algae and phytoplankton
• Produce potent neurotoxins that can be
  transferred through the food web
• Toxins affect zooplankton, shellfish, fish,
  birds, marine mammals, and even humans
• Blooms are most likely to occur in an areas of
  cooler, nutrient-rich waters
• Both occur in southern California due to
  upwelling and anthropogenic inputs

7
Features of Interest

• Fresh water plume
• River discharge
• Anthropogenic input
• Nutrient-rich, low density water
• Potential for productivity
• Interested in the centroid (eye of the storm) and
  the extent of dispersion (boundary)?
• Eddy
• Cold-core eddies raise the thermocline and bring
  cooler, nutrient-rich water toward the surface
• Promotes productivity
• Interested in a cross-section view

Images courtesy of Ocean Imaging, Inc. (top) and
Defant, 1927 (bottom).
8
Ocean Model

• Regional Ocean Model System
• ROMS - Split-explicit, free-surface,
  topography-following-coordinate oceanic model
• Daily provides a 12-hour hindcast and 36 hour
  forecast
• Open-source and widely accepted in many
  communities
• Carried out at JPL, California Institute of
  Technology under a contract with NASA
• Initially developed to model the ocean in
  southern California

1 km
9 km
3 km
12 km
9
Mobile Sensor Platform

• Webb Slocum Autonomous Glider
• Passive actuation
• Long-term deployments (1 month)?
• Slow moving vehicle (1km/hr)?
• Waypoint-based trajectory plan
• Robust
• Depth-rated to 200m

10
General Concept

• Track and collect daily information about an
  ocean process or feature
• Feature has a lifespan on the order of weeks
• Tracking trajectory duration of approximately 12
  hours
• Assimilate data into ROMS for an updated
  prediction
• Generate a new tracking trajectory
• Repeat until feature is out of range or no longer
  of interest

11
Trajectory Design Algorithm

• 1. Feature Observation and Delineation
• Feature bounded by a set of points referred to as
  drifters
• 2. ROMS Prediction
• Hourly prediction for given time period
• Web-based GUI developed for this research
• 3. Waypoint Generation Algorithm
• 4. Trajectory QA/QC
• 5. Mission Upload
• 6. Mission Execution
• 7. Data Download and Assimilation

12
http//ourocean.jpl.nasa.gov/SCB
13
Web-based Interface
14
Centroid Tracking Algorithm

• Input Hourly predictions of the locations of the
  drifters
• Trajectory waypoints are the predicted locations
  of the centroid at 4 hour time intervals.
• Minimize surfacings (safety)?
• Three possibilities
• dlltd(Ci,Ci4)ltdu
• d(Ci,Ci4)ltdl
• d(Ci,Ci4)gtdu

Ci
d(Ci,Ci4)?
dllt
lt du
d(Ci,Ci4)ltdl
d(Ci,Ci4)gtdu
Ci4
Ci6
15
Boundary Tracking

• Input Hourly predictions of the locations of the
  drifters
• Trajectory waypoints computed for 4 hour time
  intervals.
• Circle of radius rv4 is drawn around the
  vehicle or its predicted location (distance
  reachable in 4 hours)?
• The intersection of this circle with the boundary
  leads to three possibilities
• ?2 intersection points
• 1 intersection point
• Empty intersection

Average azimuth of all drifters
16
Data Assimilation
17
Observational Area
Nevada
San Francisco
Las Vegas
California
Los Angeles
600 km
18
Southern California Bight
100 km
19
Testing Region
10 km
20
Implementation Results

• Two Webb gliders
• On deployment for communications testing
• Task the gliders to track a feature of interest
• Proof of concept mission
• Test the technology chain
• Tracked the centroid and boundary
• Two separate tracking trajectories
• Data upload and assimilation between missions
• Feature re-delineated for day two

21
May 11, 2009 1100a
5 km
22
May 11, 2009 300p
5 km
23
May 11, 2009 1100p
5 km
24
May 12, 2009 300a
5 km
25
Intermission

• 16 hour mission
• Data sent back from glider for assimilation
• Re-delineate feature of interest
• Updated ROMS prediction

26
May 12, 2009 300p
5 km
27
May 12, 2009 700p
5 km
28
May 12, 2009 1100p
5 km
29
May 13, 2009 300a
5 km
30
May 13, 2009 700a
5 km
31
Conclusions

• Established and demonstrated successful
  implementation of a new technology chain for AUV
  trajectory design
• Effectively tasked, assimilated data and retasked
  AUVs in the field for feature tracking
• Working on incorporating 3D currents into the
  trajectory design
• Preparing for Bight 2010 comprehensive survey
  of SCB.
• 4 gliders, 2 months
• Planning to track and monitor a REAL HAB event.