You are on page 1of 9

INTEG. AND COMP. BIOL.

, 42:118–126 (2002)

Maneuvering and Stability Performance of a Robotic Tuna1

JAMIE M. ANDERSON2 AND NARENDER K. CHHABRA


Mechanical and Instruments Division, Charles Stark Draper Laboratory, 555 Technology Square, Cambridge, Massachusetts 02139

SYNOPSIS. The Draper Laboratory Vorticity Control Unmanned Undersea Vehicle (VCUUV) is the first
mission-scale, autonomous underwater vehicle that uses vorticity control propulsion and maneuvering. Built
as a research platform with which to study the energetics and maneuvering performance of fish-swimming
propulsion, the VCUUV is a self-contained free swimming research vehicle which follows the morphology
and kinematics of a yellowfin tuna. The forward half of the vehicle is comprised of a rigid hull which houses
batteries, electronics, ballast and hydraulic power unit. The aft section is a freely flooded articulated robot
tail which is terminated with a lunate caudal fin. Utilizing experimentally optimized body and tail kinematics
from the MIT RoboTuna, the VCUUV has demonstrated stable steady swimming speeds up to 1.2 m/sec and
aggressive maneuvering trajectories with turning rates up to 75 degrees per second. This paper summarizes
the vehicle maneuvering and stability performance observed in field trials and compares the results to
predicted performance using theoretical and empirical techniques.

INTRODUCTION ploy a long slender hull with a propeller as the main


Unmanned Undersea Vehicle (UUV) technologies propulsor and moveable lifting surfaces which provide
have evolved in recent years to produce highly func- maneuvering control. Although several advanced dem-
onstrations have been made in recent years with con-

Downloaded from icb.oxfordjournals.org by guest on September 30, 2011


tional and capable platforms for a wide variety of un-
dersea missions. As supporting technologies have pro- ventional designs, these types of vehicles are funda-
gressed, so have the mission requirements. Today’s mentally limited in their maneuvering performance.
UUV missions require a variety of capabilities which Typically requiring several body lengths turning di-
in some cases can be mutually exclusive: high transit ameter, these vehicles can have fatally poor perfor-
speed, long range and duration, maneuverability and mance at low speeds. Attempts to improve low speed
station keeping ability. Fish and marine mammals have performance by using cross-axis thrusters have been
captured the interest of vehicle designers as they are effective but the net result is loss of useful hull volume
able to cruise great distances at significant speed, ma- and degraded performance at higher speeds.
neuver in tight spaces and accelerate and decelerate In this paper we present results and analysis of the
quickly from rest or low speed with the same inte- first engineering demonstrations of the Draper Labo-
grated propulsion and steering system. Although trade- ratory Vorticity Control UUV (VCUUV), a prototype
offs exist between efficient transit and maneuverability flexible-hull UUV which propels and maneuvers like
in marine animals (Webb, 1994), engineers have the a tuna. Named after the vorticity control flow control
luxury to select the most desirable attributes to mimic mechanisms employed by fishes to propel and maneu-
in their designs. ver, the VCUUV is an engineering approximation of
In recent years, research in the fluid flow mecha- the form and movement of a large yellowfin tuna
nisms used by fish and marine mammals for propul- (Thunnus albacares). Across the broad spectrum of
sion and maneuvering has demonstrated the utility of fish form and movement, the tuna was selected as our
bio-propulsion for undersea vehicles. Barrett and his natural model to emulate as they are streamlined, pos-
collegues (Barrett, 1996; Barrett et al., 1999) illustrat- sess a rigid forebody and propel with low amplitude
ed with his RoboTuna apparatus that manipulation of movements in conjunction with a high performance
the body of an undersea vehicle in a fish-like manner hydrofoil (caudal fin) (Dewar and Graham, 1994). Re-
could enhance energetic performance significantly. nowned as a highly efficient cruising fish, the tuna is
Wolfgang et al. (1999) showed that unsteady fish-like not considered a highly maneuverable fish compared
movements produce large fluid dynamic effects which to other fishes (Blake et al., 1995). Despite their ma-
can dramatically affect propulsion and maneuvering neuvering limitations, tunas outperform conventional
performance. rigid undersea vehicles by a large margin, even at low
Conventional methods for improving unmanned un- speeds where maneuverability generally suffers.
dersea vehicle (UUV) performance (low drag hull pro- Fish-like propulsion and maneuvering is suited to
files, propeller design, energy storage technology) meet the challenges of today’s undersea missions.
have made little progress in improving propulsive ef- These missions often require long transit, long dura-
ficiency and maneuverability. Conventional UUVs em- tion on site, loitering without loss of power, operation
in close proximity to objects for docking, tagging, etc.,
1 From the Symposium Stability and Maneuverability presented at
and in dynamic environments such as in shallow water
the Annual Meeting of the Society for Integrative and Comparative near the beach zone. Fish-like propulsion and maneu-
Biology, 3–7 January 2001, at Chicago, Illinois. vering may prove to be essential in realizing these mis-
2 E-mail: Jamie@draper.com sions in that fish-like maneuverability may not be as

118
MANEUVERING AND STABILITY IN ROBOTIC TUNA 119

FIG. 1. The Draper Laboratory Vorticity Control UUV (VCUUV) shown assembled and during autonomous operation in a test tank. The
vehicle is shown executing a turning maneuver with full body deflection.

energetically taxing as conventional means of gener- the components required to execute autonomous mis-
ating large side forces (such as thrusters) in varying sions: onboard energy, actuation and control. Figure 2
conditions. Fish are not poor performers at low veloc- illustrates how the vehicle length is divided equally
ities like conventional vehicles. In fact, speed is not a between a rigid forebody which houses the electronics,
requirement for maneuverability as illustrated with the batteries, ballast and hydraulic unit, and a flexible ar-
rigid-bodied boxfish (Walker, 2000) and the fast start ticulated tail-caudal fin assembly.

Downloaded from icb.oxfordjournals.org by guest on September 30, 2011


acceleration of fish from rest (Domenici and Blake, The VCUUV is an engineering approximation of the
1997). Fish-like propulsion and maneuvering marries yellowfin tuna. The body shape was directly scaled
the benefits of range and speed of conventional low from a casting of a 1 m long animal obtained specially
drag hulls driven by propellers, with the ability to ma- for this purpose. No corrections were made to adjust
neuver precisely when necessary. Although design for changes in body shape between the 1 m fish and
trade-offs between transit and maneuverability perfor- the 2.4 m vehicle. To replicate the fin geometries, pho-
mance must be made, flexible fish-like hulls provide tographs of recently captured specimens were scaled
more engineering options for vehicle design. and major fin geometries were averaged. Standard
symmetric airfoil sections (NACA 0015) were utilized
VCUUV SYSTEM DESCRIPTION for all the fin profiles based on chord and thickness
Recent work at MIT with the RoboTuna apparatus measurements. Although an engineered system such as
(Barrett, 1996; Barrett et al., 1999) inspired construc- the VCUUV can never be completely true to the nat-
tion of the 2.4 m, 173 kg VCUUV prototype (Fig. 1). ural model, the VCUUV is an excellent replica. Small
The VCUUV is fully autonomous and is a major de- deviations from the natural model were made in the
parture from the externally manipulated and powered pectoral fins, the size of the dorsal and anal fins, and
MIT RoboTuna apparatus. Although not intended as the omission of the first dorsal fin, pelvic fins and fin-
an ocean-going vehicle, the VCUUV contains all of lets. We believe these deviations have a small effect

FIG. 2. VCUUV system layout illustrating the major subsystems of the vehicle. The cutaway view shows the arrangements of components
internal to the pressure hull and the tail drive structure underlying the tail surface skin and scales.
120 J. M. ANDERSON AND N. K. CHHABRA

on overall performance due to the small projected ar- The maximum speed attained in the field trials was
eas involved compared to the tail and caudal fin. Bar- 1.25 m/sec (2.4 knots) at 1 Hz tail oscillation while
rett (1996) also produced good results with a tuna plat- under heading control (closed loop control using com-
form which also ignored these elements. pass feedback). Without heading control, the maxi-
The vehicle has been successfully operated in shal- mum speed attained was 1.19 m/sec (2.3 knots). We
low, controlled environments to evaluate its propul- do not report statistical variances on our results as
sion/maneuvering system. Vehicle design and fabri- speed measurements were only made during a few se-
cation issues including sizing of actuators, design of lect runs by the addition of a small external speed
articulated body, pressure hull design, and on-board sensor mounted to the top of the pressure hull, pro-
intelligence, sensors and power are described in detail truding beyond the estimated boundary layer thick-
by Anderson and Kerrebrock (1997, 1999). The ness. The speed sensor was a custom designed cali-
VCUUV was designed to operate in freshwater envi- brated free-wheeling propeller which was instrumented
ronments at depths less than 10 m at typical live ani- with four magnets attached to its shaft. A Hall-effect
mal cruise speeds near one body length per second. sensor was used to count the rotations of the propeller
shaft to obtain speed. Calibration by timed distance
FIELD TRIALS confirmed that the velocity error was less than 5%.
Early field trials (April–June, 1998, at University of The measured vehicle speed was 16% less than that
New Hampshire test tank) focused on engineering di- predicted by Strouhal scaling of the MIT RoboTuna
agnostics and tuning of the hydraulic system and con- results which may be due to differences in the kine-
trol system performance. Hydraulic pressure, control matics between the VCUUV tail motion and that of
gains, and desired tail kinematics were adjusted to give the RoboTuna. Although morphologically similar, the
good tracking performance for frequencies up to 1 Hz. VCUUV and RoboTuna are not identical platforms.

Downloaded from icb.oxfordjournals.org by guest on September 30, 2011


Tail oscillations above 1 Hz were not pursued due to The VCUUV utilizes four active links to achieve the
larger than expected forces which led to poor tail con- desired body undulation whereas the RoboTuna has
trol performance. Simple maneuvers were explored in- eight, six of which are independently controlled. Thus,
cluding coasting turns and biased swimming move- exact duplication of the kinematics was not possible.
ment resulting in circular trajectories. Because of the Using the same length reference (L, length from nose
short distances available in the UNH tank, steady state to caudal fin pivot), the VCUUV achieved 0.61 lengths
swimming at terminal velocity could not be achieved. per second (L/sec) whereas the RoboTuna achieved
Thus, field trials were moved to local freshwater lakes 0.65 L/sec. Unfortunately, degraded performance of
for longer duration and higher speed tests. the hydraulic system above 1 Hz did not allow the
During the summer and fall of 1998, the VCUUV VCUUV to achieve the design speed of 1.0 L/sec. Ac-
was tested in freshwater lakes in Hopkinton State Park ceptable linkage trajectory tracking performance
and Nickerson State Park. These sites were chosen for above 1 Hz could not be achieved because of phase
their proximity to Draper Laboratory and their clean, lag caused by the high loads at those frequencies.
undisturbed fresh water. A systematic study of straight The VCUUV was very stable in straight, forward
swimming parameter variations was completed as well motion. A small amount of yaw oscillation was present
as several basic open and closed loop horizontal plane during swimming (2–38 amplitude) and it decreased
maneuvers. The swimming parameter study involved with increased velocity. This amount of yaw is con-
variations around a baseline parameter set based on the sistent with that observed with live animals by Dewar
best results reported by Barrett (1996) for self-pro- and Graham (1994). Roll perturbations were imper-
pelled optimal motion. Approximately 20 straight ceptible, presumably due to the large metacentric
swimming cases were studied including variations of height and the significant roll damping from the large
frequency of oscillation, amplitude of caudal lateral caudal and pectoral fins. Significant pitch oscillation
movement, caudal fin angle of attack, and caudal fin (5–108 amplitude) during swimming was problematic
phase with respect to the body wave. Maneuvering ex- in certain cases, largely due to the vertical separation
periments were done for straight swimming with vary- of the center of thrust from the caudal fin-tail assembly
ing linkage bias angles and full deflection coasting and the center of gravity. The thrust oscillation cycle
turns. coupled into the pendulous pitch mode. The effect of
A typical experimental run began with the vehicle specific parameters on pitch oscillations was not ex-
at rest and submerged approximately 1.5 m while plored due to limited test time. However, it was noted
baseline inertial data were collected. The vehicle then that some cases consistently produced pitching and
began the swimming tail motion that accelerated the that the maximum pitch was somewhat dependent on
vehicle to steady state velocity. The typical accelera- initial conditions.
tion period was 20–30 sec for 1 Hz tail oscillation.
After swimming for a total of one minute, the vehicle MANEUVERING RESULTS
coasted to a stop while the tail was held in a straight The VCUUV possesses excellent maneuverability
position. The tail movement for swimming adhered as that approaches the performance of live animals. Blake
well as possible to the kinematics reported by Barrett et al. (1995) report average turning diameters for small
(1996) for optimal performance of the RoboTuna. Thunnus albacares (14 cm length) to be 0.94 body
MANEUVERING AND STABILITY IN ROBOTIC TUNA 121

FIG. 3. VCUUV heading angle during a full deflection coasting FIG. 4. VCUUV heading angle during repeated turning (zigzag ma-
turn. The vehicle swims from rest up to full speed. At point A, the neuver). The vehicle accelerates from rest to full speed and begins
vehicle stops swimming and assumes full body curvature for a coast- a repeated series of 908 heading changes every 10 sec. As in hard
ing turn which ends at point B in a full stop. Two curves are shown, turn results in Figure 3, the compass tends to overshoot during high
one corresponding to the gyro estimate (the integral of the gyro rate rate events due to gimbal movement.
measurement), and the compass measurement. The compass oscil-
lates freely in a gimbal mechanism during high rate events as illus-
trated at the beginning (A) and end of the turn (B).
operation in highly perturbed environments and launch

Downloaded from icb.oxfordjournals.org by guest on September 30, 2011


and recovery to a moving submarine.
lengths. Although we did not vary the kinematics in Figure 4 illustrates the heading angle achieved in an
search of the minimum turning diameter, we found that aggressive course-changing trajectory. The vehicle
simple biases on the straight swimming kinematics swam straight for 20 sec to attain steady state velocity
produced turning diameters on the order of a body and then swam zigzag courses at 645 degrees from
length (2.4–3.0 m) consistent with Blake et al. (1995). the initial heading with 10 sec spent on each segment.
Several maneuvering trajectories were explored in- The vehicle maintained forward velocity during the
cluding abrupt discrete turns and continuous turning maneuver at 1.2 m/sec. The maximum yaw rate for
which yielded circular and 2-dimensional spiral trajec- this maneuver was approximately 308/sec which also
tories. Turning diameters were estimated visually by outperforms conventional systems by several factors.
comparing the turn diameter to the vehicle length. The zigzag maneuver indicates the level of perfor-
Figure 3 illustrates the heading angle measured dur- mance that can be achieved in simple search patterns
ing a coasting turn. The vehicle attained maximum or in close inspection missions. The vehicle can make
speed (1.2 m/sec) after 60 sec of straight swimming, aggressive course changes in-stride while effectively
and then coasted with full body deflection to a stop. preserving forward speed. Conventional UUVs slow
Two measurements are indicated in the Figure 3: the down considerably during a turn maneuver due to in-
integral of the heading gyro yaw rate and the absolute creased drag forces due to angle of attack on the hull
heading (yaw) angle measured by the compass. The and increased drag of deflected control surfaces (So-
compass is mounted in a pendulous gimbal such that ciety of Naval Architects and Marine Engineers,
when subjected to high angular rates, the compass 1989).
swings in its mount. Thus, the compass slightly over
predicts turn rates. The gyro has much better precision HYDRODYNAMIC MODEL
in rapid maneuvers but is subject to drift over time. A preliminary hydrodynamic model for the
The combination of the two sensors gives a good ap- VCUUV was constructed using a corrected and up-
proximation of the actual turn dynamics. The data dated version of the Submarine Hydrodynamic Anal-
from both sensors are in agreement and indicate that ysis Tool (SUBHAT) (Boeing Computer Services
the vehicle turned through approximately 160 degrees Company, 1987) that is used at Draper Laboratory for
in 10 sec. conceptual designs of conventional UUVs. This code
The maximum measured yaw rate during this ma- originated in the Hydrodynamic Analysis Techniques
neuver was 758/sec. As a point of comparison, our own (HYDAT) system of computer programs developed by
calculations (presented in Fig. 9) have shown that a the Naval Coastal Systems Center (1983). These tools
conventional UUV with control surfaces turns at ap- estimate relevant hydrodynamic coefficients (force and
proximately 3 to 58/sec at that speed, often requiring moment derivatives) using semi-empirical methods,
several body lengths and as long as a minute to com- which utilize both theoretical analysis and experimen-
plete a reversing turn. This radical improvement in tal data. They include added mass, drag, and lift co-
maneuverability may enable mission elements previ- efficients and take into account all of the important
ously considered beyond the capabilities of UUVs in- effects: potential or inviscid effects, fluid acceleration
cluding close inspection and rapid course change tasks, effects, and viscous effects. Much of the background
122 J. M. ANDERSON AND N. K. CHHABRA

FIG. 5. Body centered coordinate system indicating surge, sway


and heave velocities (u, v, w) and roll, pitch and yaw velocities (p,
q, r). Distances are given in meters from the origin at the center of
buoyancy.

needed to understand these tools can be found in the FIG. 6. Plan (view from above) and elevation (view from side)
books by Hoerner (1965, 1975) and Newman (1977). views of the VCUUV geometry modeled for hydrodynamic coeffi-
These references have gathered theory and experimen- cient derivation. The body and fin shapes were simplified slightly
as shown to conform to the inputs of the hydrodynamic analysis
tal data from various sources and presented them in a tools.
concise manner.

Downloaded from icb.oxfordjournals.org by guest on September 30, 2011


We adhere to standard naval architecture conven-
tions (Society of Naval Architects and Marine Engi-
the hydrodynamic forces and moments that are used
neers, 1952) and assign a body centric coordinate sys-
in the 6-DOF equations of motion for the vehicle such
tem as illustrated in Figure 5. The x-axis points for-
as presented by Feldman (1979). Most of the hydro-
ward and the surge velocity in this direction is u. The
dynamic forces and moments are represented in terms
y-axis points starboard and the sway velocity in this
of hydrodynamic coefficients and combinations of the
direction is v. The z-axis points downward and the
vehicle rates (u, v, w, p, q, r) and control surface (fin)
heave velocity is w. As indicated in Figure 5, positive
deflections (dr for rudder, ds for stern plane, and da
rotational velocities about these axes are p (roll), q for aileron). For example, a major contribution to the
(pitch) and r (yaw), respectively. pitch moment is
The Draper version of SUBHAT was exercised for
the VCUUV hull and fin geometries illustrated in Fig- Mwuw
ure 6 in plan and elevation views. For the simplified where Mw is a hydrodynamic coefficient. The non-di-
analysis presented here, we assume that the body re- mensional coefficient is obtained by dividing the hy-
mains fixed to the centerline and completely rigid and drodynamic coefficient with a dimensionally appropri-
that the caudal fin rotates rigidly about its pivot. In the ate factor, e.g.,
actual system, the aft half of the body undulates with
a traveling wave of increasing amplitude envelope as Mw
described in Anderson and Kerrebrock (1997). Al- M9w 5
1 3
though this analysis approach is simplistic, it fits read- rL
2
ily into the framework with which we characterize the
stability and maneuvering performance of convention- where r is the fluid density, and L is the vehicle char-
al undersea vehicles. This analysis produces conser- acteristic length.
vative maneuvering performance estimates since body A major contribution to Mw is known as the Munk’s
articulation increases the effective rudder control au- moment and is the difference between the z- and x-
thority. direction added masses (m33 2 m11). Here, m11 is the
The VCUUV caudal fin model is a simplified rep- added mass in the x-direction (forward) and m33 is the
resentation of the actual fin comprising two rudder sur- added mass in the z-direction (downward). Munk’s
faces as illustrated in Figure 6. The actual caudal fin moment destabilizes the vehicle in that turning tends
of the vehicle is wholly moveable and more accurately to increase turn rate, causing the vehicle to turn further.
represents the actual animal’s lunate caudal profile. Similarly, there is a Munk’s moment in the yaw direc-
The pectoral fins are similarly modeled as canards tion included in the term
(bow planes) as illustrated. Dorsal and anal fins are
Nvuv
assumed as fixed surfaces although on the actual ve-
hicle they are deflected with the tail segment to which where (m11 2 m22) is the Munk’s moment contribution
they are attached. to Nv. m22 is the added mass in the y-direction.
Using SUBHAT, we produced a complete set of hy- As with Munk’s moments, there are other contri-
drodynamic coefficients. These coefficients provide butions to hydrodynamic forces and moments from
MANEUVERING AND STABILITY IN ROBOTIC TUNA 123

TABLE 1. Comparison of stability coefficients.

VCUUV Generic UUV

Mass (kg) 173.1 m9 3.09E202 Mass (kg) 1327.1 m9 1.141E202


L (m) 2.22 r 1.027E103 L (m) 6.1 r 1.027E103
Vertical Plane Horizontal Plane Vertical Plane Horizontal Plane
Z9w 21.90E201 Y9v 26.90E202 Z9w 21.31E202 Y9v 21.20E202
Z9q 9.08E203 Y9r 3.15E202 Z9q 26.16E203 Y9r 5.59E203
M9w 3.28E202 N9v 22.11E203 M9w 3.54E203 N9v 24.11E203
M9q 21.27E203 N9r 21.82E202 M9q 22.98E203 N9r 22.70E203
Gv 24.44 Gh 1.00 Gv 0.52 Gh 0.26
UNSTABLE STABLE STABLE STABLE

added mass. The added-mass tensor mij is constant for horizontal plane. These are computed as (Humphreys
a given vehicle geometry and can be computed from and Watkinson, 1992):
standard formulas based upon shape of the body, e.g.,
M9w [Z9q 1 m9]
ellipsoidal (Kochin et al., 1964), spheroidal, cylindri- Gv 5 1 2 (2)
cal, or strip theory (Newman, 1977). The added mass Z9w M9q
tensor reflects the potential flow contribution to both
r 2 m9]
N9[Y9
Gh 5 1 2 v
the acceleration and the velocity coefficients. The fol- . (3)
lowing equations give the added mass contributions to Y9N9
v r

Downloaded from icb.oxfordjournals.org by guest on September 30, 2011


the hydrodynamic forces (F) and moments (M): Here m9 is the non-dimensional vehicle mass and
Fx 5 2m11 u̇ 1 m22 vr 2 m33 wq 2 m35 qq 1 m26 rr the other non-dimensional hydrodynamic coefficients
have similar meaning to M9w and N9v (Feldman, 1979).
Fy 5 2m22 v̇ 2 m26 ṙ 2 m11 ur 1 m33 wp 1 m35 pq A vehicle is unstable in the respective planes if the
coefficients Gv or Gh are negative. It is statically stable
Fz 5 2m33 ẇ 2 m35 q̇ 1 m11 uq 2 m22 vp 2 m26 pr if coefficients are greater than 1 and dynamically sta-
Mx 5 2m44 ṗ ble for values between 0 and 1. As explained by Hum-
phreys and Watkinson (1992), equations (2) and (3)
My 5 2m55 q̇ 2 m35 ẇ 1 (m33 2 m11 )uw 1 m35 uq are derived by a series of approximations making an
‘‘infinite speed’’ approximation, which neglects any
1 m26 vp 1 (m66 2 m44 )pr buoyancy restorative effects due to the separation of
centers of gravity and buoyancy. If a vehicle is dy-
Mz 5 2m66 ṙ 2 m26 v̇ 1 (m11 2 m22 )uv 2 m26 ur
namically stable as indicated by equations (2) and (3)
2 m35 wp 1 (m44 2 m55 )pq. (1) it will be dynamically stable at all ahead speeds as the
neglected terms increase dynamic stability.
A dot over a variable represent the rate of change with Table 1 compares pertinent hydrodynamic and dy-
respect to time; thus u̇ is the vehicle’s acceleration namic stability coefficients computed with equations
along its x-axis. (2) and (3), for the VCUUV and for comparison, a
generic highly-maneuverable axis-symmetric UUV.
STABILITY
The selected generic UUV has properties that resemble
A trade-off relationship exists for sizing maneuver- several research and development UUVs and is sub-
ing surfaces for an undersea vehicle. For given ex- stantially larger than the VCUUV. As all results are
pended power, larger control surfaces produce greater normalized for the vehicle length, comparison can be
drag in headway motion, leading to smaller attainable made without concern for the actual size. The generic
maximum speeds. Larger control surfaces make the UUV is 6.1 m long with a maximum diameter of 53
vehicle more stable and also reduce its maneuverabil- cm and has a mass of 1327 kg and a maximum thrust
ity. In contrast, overly small control surfaces can de- of around 334 N. Four symmetrical moveable fins at
stabilize the vehicle. Maneuvering coefficients for the its stern provide maneuvering forces and moments.
VCUUV provide drag, stability coefficients and turn- The stability coefficients in Table 1 indicate that the
ing diameters. Simple measures for static stability are VCUUV is dynamically stable in the horizontal plane
the values for M9w and N9v , the static pitch and yaw and dynamically unstable in the vertical plane. The
moment coefficients. A negative value for M9w and a vertical plane instability of the VCUUV is mitigated
positive value for N9v indicate static stability. High as- mechanically by separating the centers of gravity and
pect ratio vehicles are generally statically unstable be- buoyancy. This large metacentric height acts as a
cause of the Munk’s moment effect. spring against pitch and roll motions. The generic
Another more quantitative measure of vehicle sta- UUV is slightly stable in both planes as is typical of
bility (or maneuverability) is given by dynamic sta- most UUVs. It is designed to be highly maneuverable
bility coefficients Gv in the vertical plane and Gh in the and has an autopilot to control its maneuvers.
124 J. M. ANDERSON AND N. K. CHHABRA

TABLE 2. Component contributions to stability contributions.

Coefficient Hull alone Pectoral Dorsal Anal Caudal

Y9v*103 25.59 210.35 28.7 244.4


Y9r*103 0.32 2.01 1.95 27.22
N9v*103 232.56 1.9 1.85 26.72
N9r*103 21.06 20.36 20.41 216.36
Z9w*103 25.97 2183.94
Z9q*103 20.47 9.54
M9w*103 19.99 12.8
M9q*103 20.72 20.54
FIG. 7. Rigid VCUUV predicted turning diameter as a function of
caudal fin (rudder) deflection at 1.2 m/sec.

To illustrate the contributions of fins to the stability


of the complete VCUUV vehicle, contributions to the are adjusted to produce zero total yaw moment. Under
pertinent hydrodynamic coefficients of various vehicle these conditions, the yaw rate and forward speed are
elements (hull alone, pectoral fins, dorsal fin, anal fin, used to compute the turning diameter. We assume for
and caudal fin) are listed in Table 2. Stability coeffi- simplicity that the vehicle follows a circular trajectory
cients for different combinations of components are with no loss of speed in the turn (i.e., constant yaw
also computed and listed in Table 3. Results show that rate).
hull alone is highly unstable (Gv 5 2140 and Gh 5 Figures 7 and 8 present horizontal turning diameters
2167) because of the large Munk’s moment in both and yaw rates of the VCUUV vehicle for different cau-
the vertical and the horizontal planes. Pectoral fins in- dal fin (rudder) deflections at 1.2 m/s. Turning diam-

Downloaded from icb.oxfordjournals.org by guest on September 30, 2011


crease Mw9 by 64% (add 0.0128 to 0.01999) as these eter decreases from 70 lengths at 1 degree of fin de-
fins are ahead (towards bow) of the hull center of flection to 3.2 lengths at 30-degrees fin deflection. A
buoyancy and thus decrease stability in the vertical nonlinear cross-flow drag model is used in these cal-
plane. Conversely the vertical lift force coefficient Z9w culations (Chhabra and Scholten, 1997).
of pectoral fins is 31 times the lift force coefficient of To turn on smaller turning circles, the yaw rate must
bare hull and this makes the vehicle more stable in the correspondingly increase as illustrated in Figure 8. To
vertical plane. Gv still has unstable values, equal to compare with experimental data, we use a 258 caudal
24.46. Dorsal and anal fins add stability in the hori- fin deflection as a representative value. This is the
zontal plane, as they are aft of the center of buoyancy maximum value that the fin was allowed for turn bias
of the hull. They reduce the magnitude of N9v by 12%, angle. Some caudal deflection range was reserved for
adding 0.00375 to 20.03256 to obtain 20.02881. the swimming motion upon which the turn was super-
The horizontal stability coefficient is still computed posed. The actual angles achieved on a given turn were
to be unstable at 215.87 as the horizontal force co- selected by a proportional control algorithm which act-
efficient Y9v due to these fins is only 3.4 times the hor- ed on the heading error. At this caudal fin deflection,
izontal force coefficient due to hull alone. Addition of the theoretical model predicts the rigid VCUUV’s turn-
caudal fins decreases the magnitude of Nv9 by an ad- ing diameter at 1.2 m/sec is 3.7 body lengths and the
ditional 93% adding 0.02672 to 20.02881 to obtain vehicle yaw rate is 16.98/sec. Sideslip angle (hull angle
20.00209. The horizontal stability coefficient is now of attack in the yaw plane) for this maneuver is 168,
stable and is calculated to be 1.00. Horizontal dynamic independent of speed.
stability of the VCUUV is primarily due to the large
caudal fin. DISCUSSION
Yaw rates for the actual VCUUV during autono-
MANEUVERING mous maneuvers were significantly higher than those
Hydrodynamic coefficients computed above were predicted from the above rigid body model. Although
used to compute vehicle turning diameters. Nonlinear the theoretical model assumes constant yaw rate, the
computations of turning diameters include cross-flow measured rates are highly dynamic. When entering a
drag moments along with uv, ur, and u2dr terms in the turn, the yaw rate peaks, and then quickly decays. For
summation for total yaw moment. For a given rudder the full stop bias turn illustrated in Figure 3, the peak
(caudal fin) deflection, dr, sideslip angle and yaw rate yaw rate entering the turn was 758/sec and by the end

TABLE 3. Stability coefficient summary.

Configuration Gv Gh

Hull alone 2140.19 2166.80


Hull 1 pectoral 24.46 2165.19
Hull 1 pectoral 1 anal 1 dorsal 24.44 215.87
Hull 1 pectoral 1 anal 1 dorsal 1 caudal 24.44 1.00
MANEUVERING AND STABILITY IN ROBOTIC TUNA 125

FIG. 8. Rigid VCUUV predicted yaw rate as a function of caudal


fin (rudder) deflection at 1.2 m/sec.

FIG. 9. Yaw rate as a function of speed for rigid VCUUV and


of the maneuver 10 sec later, the yaw rate was essen- generic UUV.
tially zero. For repeated zigzag maneuvering at 1.2 m/
sec (Fig. 4), the VCUUV turned at approximately 308/
sec entering the turn, with decay to zero yaw rate ap- tional vehicle hydrodynamic analysis. Based on the as-
proximately 5 sec into each trajectory segment. The sumption that the VCUUV body remains rigid and
turning diameter for this case was approximately one only the caudal fin deflects, hydrodynamic and stabil-
body length (by visual observation). ity coefficients were calculated. As expected, the cau-
Although our measured peak yaw rates do not cor- dal fin stabilizes the VCUUV in straight-ahead motion.
respond directly with the constant yaw rates we predict The VCUUV is hydrodynamically unstable in the ver-
with our theoretical model, some general observations tical plane. The rigid VCUUV model does not capture
can be made. The theoretical model of the rigid the added effectiveness of the body undulation as a
VCUUV with 258 fin deflection produces turning rates rudder and under predicts the actual maneuverability
by roughly a factor of two.

Downloaded from icb.oxfordjournals.org by guest on September 30, 2011


of only 16.98/sec which as expected, under predicts the
measured peak rates of 30–758/sec. This indicates that
ACKNOWLEDGMENTS
the unmodeled body deflection may significantly in-
crease the fin effectiveness. In order to predict the turn The research described in this paper was funded in-
rates observed in the field, the required rudder moment ternally by the Charles Stark Draper Laboratory.
is roughly double that produced by the caudal fin
REFERENCES
alone. With this model correction, the turning diameter
is two body lengths, which is larger than our rough Anderson, J. M. and P. A. Kerrebrock. 1997. The Vorticity Control
Unmanned Undersea Vehicle—An autonomous vehicle employ-
visual estimate. A possible explanation is that the as- ing fish swimming propulsion and maneuvering. Proc. 10th Int.
sumption of constant yaw rate does not apply exactly Symp on Unmanned Untethered Submersible Technology, Dur-
for this trajectory. ham, New Hampshire, Sept, 1997, 189–195.
Hydrodynamic coefficients used to predict turning Anderson, J. M. and P. A. Kerrebrock. 1999. The Vorticity Control
Unmanned Undersea Vehicle (VCUUV) performance results.
diameters are generally independent of speed, except Proc. 11th Int. Symp on Unmanned Untethered Submersible,
cross-flow drag coefficients. Thus, yaw rate is nearly Technology, Durham, New Hampshire, August, 1999, pp. 360–
linear with speed. Figure 9 illustrates this trend for the 369.
corrected rigid VCUUV model and for the generic Barrett, D. 1996. Propulsive efficiency of a flexible hull underwater
UUV presented previously. At 1.2 m/sec, the VCUUV vehicle, Ph.D. Diss., Massachusetts Institute of Technology,
Cambridge, Massachusetts.
has 7.5 times more yaw rate than the generic UUV. Barrett, D., M. S. Triantafyllou, D. K. P. Yue, M. A. Grosenbaugh,
For this case at the 258 rudder deflection, the generic and M. J. Wolfgang. 1999. Drag reduction in fish-like loco-
UUV has a sideslip angle of 8.2 degrees and a turning motion. J. Fluid Mech. 392:183–212.
diameter of 5.4 body lengths. Blake, R. W., L. M. Chatters, and P. Domenici. 1995. Turning radius
of yellowfin tuna (Thunnus albacares) in unsteady swimming
SUMMARY manoeuvres. J. Fish Biol. 46:536–538.
Boeing Computer Services Company. 1987. SUBHAT—Submarine
The VCUUV has demonstrated the utility of fish- Hydrodynamic Analysis Tool, System User Manual, Vol. I.
like propulsion and maneuvering for improving un- Chhabra, N. K. and J. R. Scholten. 1997. All-angle-of-attack hydro-
dersea vehicle performance. The VCUUV propels and dynamic models for underwater vehicles, validated by tank test
data, ASME Int. Mechanical Engineering Congress and Expo-
maneuvers stably in the horizontal plane, with perfor- sition, Dallas, Texas, 16–21 November 1997.
mance surpassing conventional axis-symmetric de- Dewar, H. and J. B. Graham. 1994. Studies of tropical tuna swim-
signs. In field trials, the VCUUV achieved 1.2 m/sec ming performance in a large water tunnel, Part III: Kinematics.
and turn rates up to 758/sec. On aggressive repetitive J. Exp. Biol. 192:45–59.
Domenici, P. and R. W. Blake. 1997. The kinematics and perfor-
maneuvers, the VCUUV turned at 308/sec without loss mance of fish fast-start swimming. J. Exp. Biol. 200:1,165–
of forward speed. Conventional vehicles cannot 1,178.
achieve this trajectory due to their low yaw rate of 48/ Feldman, J. 1979. DTNSRDC Revised standard submarine equations
sec. For this maneuver, a generic UUV has turning of motion, David W. Taylor Naval Ship Research and Devel-
diameter of 5.4 body lengths whereas the VCUUV has opment Center, DTNSRDC/SPD-0393–09, June, 1979.
Hoerner, S. F. 1965. Fluid—dynamic drag. Published by the author.
turning diameter of approximately two body lengths. Hoerner, S. F. and H. V. Borst. 1975. Fluid—dynamic lift. Published
A simplified hydrodynamic model was developed by the authors.
within the framework customarily used for conven- Humphreys, D. E. and K. W. Watkinson. 1992. The impact of open-
126 J. M. ANDERSON AND N. K. CHHABRA

loop vehicle stability on closed-loop control system perfor- Society of Naval Architects and Marine Engineers. 1952. Nomen-
mance: An analysis of four AUV concepts. A.R.A.P. Group, clature for treating the motion of a submerged body through
Proc. Oceans ’92 Conference, October, 1992, 533–538. fluid, SNAME Technical and Research Bulletin No. 1–5, April,
Kochin, N. E., I. A. Kibel, and N. V. Roze. 1964. Theoretical hy- 1952.
drodynamics. Interscience Publishers, New York. Walker, J. A. 2000. Does a rigid body limit maneuverability? J. Exp.
Naval Coastal Systems Center. 1983. Hydrodynamic Analysis Tech- Biol. 203:3,391–3,396.
niques (HYDAT). Panama City, Florida, March 1983. Webb, P. W. 1994. The biology of fish swimming. In L. Maddock,
Newman, J. N. 1977. Marine hydrodynamics. MIT Press, Cam-
Q. Bone, and J. M. V. Rayner (eds.), Mechanics and physiology
bridge, Massachusetts.
Society of Naval Architects and Marine Engineers. 1989. Analysis of animal swimming, pp. 45–62. Cambridge University Press,
of Turning Ability. In E. V. Lewis (ed.), Principles of naval Cambridge, UK.
architecture, Second revision, Vol. III, motions in waves and Wolfgang, M. J., J. M. Anderson, M. A. Grosenbaugh, D. K. Yue,
controllability, pp. 209–215. The Society of Naval Architects and M. S. Triantafyllou. 1999. Near-body flow dynamics in
and Marine Engineers, New Jersey. swimming fish. J. Exp. Biol. 202:2,303–2,327.

Downloaded from icb.oxfordjournals.org by guest on September 30, 2011

You might also like