Transcript FAF’05 Mission A1 Objective: Test Acoustic Rapid Environmental Assessment mechanisms. • Construct an adaptive AUV path control. • Predict ocean in real-time. • Optimize control.

FAF’05 Mission A1
Objective: Test Acoustic Rapid Environmental
Assessment mechanisms.
• Construct an adaptive AUV path control.
• Predict ocean in real-time.
• Optimize control parameters in real-time, s.t.
minimize TL uncertainty.
Principle & Method
Depth (m)
• Adaptive AUV path control --- yoyo control
Range (km)
Forward
Backward
(m/s)
(m/s)
Principle & Method
• Adaptive AUV path control --- yoyo control
c
with a threshold d
• Compare
z
Depth (m)
• Relative position to thermocline.
• Relative position to upper bound
and lower bound and bottom.
Control parameters:
• n: number of sampling points to calculate
• d: threshold of c
z
Range (km)
c
z
Virtual Experiment
Principal Estimate
MIT
Ensemble of
HOPS/ESSE forecasts
Nowcasts at future time
Data Assimilation
Smaller
Statistics &
Acoustic
model
Sample
variance of TL
Principle & Method
CTD
noise
Yoyo 2
Yoyo 1
R1
P.E.
P.E. new
OA
Err. new
SVP Generator
TL 1,1
..…
……
Yoyo 7
Rm
TL uncertainty
associated with
TL 1,m
Yoyo 1
Implementation
• Plan 7/13/2005~7/16/2005
Bravo
Charlie
Alpha
NC
Echo
2.5 km
10 6’ E
2 km
ACOMM Bouy
LBL transponder
POOL
Delta
42 35’ N
Implementation & Results
• Plan for 7/17/2005~7/26/2005
Optimal: n=30, d=1000 for afternoon of Jul 26
Max range=2.1km, frequency=100Hz
Implementation & Results
Optimal: points=30, threshold=1000 for morning 7/21/05
Max range=4.3km, frequency=100Hz
Implementation & Results
Optimal: points=30, threshold=0.1 for afternoon 7/21/05
Max range=2.1km, frequency=500Hz
Summary
Major Accomplishment:
• Constructed an AUV yoyo control.
• Coupled HOPS outputs, AREA simulator and optimization codes
together in a simple version.
• Implemented ocean prediction and control parameters optimization
in real-time.
Future work:
• Speed up cost function computation.
• Improve optimization for nonlinear, nonseparable cost
function.
• Stochastic optimization.
Forcast
Ensemble
Principal Estimate
CTD
noise
P.E. new
P.E.
OA
Err. new
SVP Generator
TL 1,1
TL uncertainty
associated with
Rm
TL 1,m
R1
TL 1,1
Yoyo 1
+ ……. +
Yoyo 2
Yoyo 1
R1
..…
……
Yoyo 7
CTD
noise
Yoyo 2
Yoyo 1
cost 
P.E. new
Rn
OA
Err. new
SVP Generator
..…
……
Yoyo 7
Rm
E

 
 
Esum varsum
TLOAvar TLOA
CDT noise,
noise R1 , R 2 R n

TL uncertainty
associated with
TL 1,m
Yoyo 1
Correlation
Lengths
A priori
error field
New P.E. Error field
CTD noise
ith Yoyo
pattern
Objective
Analysis
Principle
Estimate
RAM
TL Uncertainty
for ith yoyo
Sound Speed
Generator