132

IEEE JOURNAL OF ROBOTICS AND AUTOMATION, VOL. R A - I , NO. 3, SEPTEMBER 1985

Automatic Body Regulation for Maintaining

Stability of a Legged Vehicle
Rough-Terrain Locomotion

Abstract-The evolution of legged vehicles has progressed significantly

in recent years. These vehicles
offer the potentialof increased mobility for
traversing rough terrain. The ability to maintain stability is an important
consideration in the development of any control algorithm for a legged
vehicle. Previous work on legged vehicle control generally assumes that
the terrain is regular enough that only minimal operator interaction is
necessary. However, for very irregular terrain the operatormay require a
guidance mode that gives maximum resolution and flexibility in controlling body, leg, position,andorientation.
Several automaticbody
regulationschemes that aid theoperator in thisimportant task are
described. A major development is the use of an improvedstability
measure which can be automatically optimized. This measure, together
with a consideration of constraints on the kinematic limits of individual
legs, leads to thedevelopment of schemes for automatic body regulation.
The automatic bodyregulation schemes are incorporated into thevehicle
control algorithm to provide a highdegree of vehicle maneuverability
while reducing the operators burden.

I. INTRODUCTION
LEGGED VEHICLE possesses a tremendous potential
for maneuverability overrough terrain, particularly in
comparison to conventional wheeled or tracked vehicles [ 11.
In general, a legged vehicle can offer more degrees-offreedom for movementthan conventional vehicles. Legged
vehicles can provide the capabilities of stepping over obstacles
or ditches, climbing over obstacles, or maneuveringwithin
confined areas of space [ 2 ] , [3].
However, the coordination of the movements of the various
leg joints in such a way as to produce the desired locomotion
of the vehicle is an extremely complex task. Previous studies
have shown that if the leg coordination is left entirely to the
human operator, evena relatively simple walkingmachine
presents such a highly complex task that the operator becomes
exhausted after only a short period of operation [4]. Therefore, it is essential to relieve the operator of as much of this
complex task as possible.
Fig. 1. Hexapodvehicles at the Ohio State University. (a) The OSU
At The Ohio State University, research is being conducted
Hexapod, walkingin
dual tripod mode. (b)Model
of the adaptive
suspension vehicle (ASV).
in several areas of legged locomotioh. A major development
of this research is the OSU Hexapod vehicle (Fig. l(a)). This
six-legged vehicle is an experimental prototypewhich is being evaluation of newsensors and sensing systems. Eachof the six
used to develop various control schemes and leg placement legs of this vehicle is comprised of three independent rotary
algorithms and serves as a test-bed for the development and
joints arranged in an arthropod configuration. The vehicle is
interfaced to a PDP-11/70 computer via an optically isolated
ManuscriptreceivedMarch
7, 1985. This workwassupported
by the
digital-data
link [5].
Defense Advanced Research Projects Agency under contract DAAE07-84-KPresently under construction is a new vehicle referred to as
ROO1.
D. A. Messuri iswithPackard
Electric Division of General Motors the adaptive suspension vehicle (ASV). A preliminary model
Corporation, P.O. Box 431, Warren, OH 44486, USA.
C. A. Klein is with the Department
of Electrical Engineering, The Ohio of this vehicle is shown inFig. l(b). This vehicle will be afullscale self-contained walking machine. Each of the six legs of
State University, 2015 Neil Avenue, Columbus, OH 43210, USA.

0882-4967/85/0900-0132$01.00 0 1985IEEE

133

MESSURI AND KLEIN: AUTOMATIC BODY REGULATION OF A LEGGED VEHICLE

the ASV will have a planar pantograph geometryin a rotatable

plane, and will be controlled by three independent actuators
[6]. This vehicle will provide a test-bed for further development and evaluation of control algorithms and sensor systems.
The control schemes and algorithms discussed in this paper
have been implemented onthe OSU Hexapod vehicle and thru
the use of computer graphic simulations on the ASV.
Followingthe
concept of supervisory control [7] the
operation ofthe vehicle has beenpartitionedinto
a setof
operational modes, which will allow the vehicle to function in
a variety of terrain conditions andwhich require varying
degrees of operator control. When the terrain is relatively
smooth, the vehicle should be able to operate with minimal
input from the operator. As the terrain conditions become
morecomplex, it becomesnecessary
for the operator to
provide additional input. Following is a list of the major
operational modes,arrangedfrom useonrelatively
simple
terrain to more complex terrain [SI, [9].
Cruise: This mode is intended for locomotion over reasonably smooth terrain, and the minimum turning radius and the
deviation between walking direction and body heading may be
limited. The control algorithm will probably not require use of
vision sensors.
Previous researchon walking algorithms hasbeen primarily
related to the cruise mode, having been restricted to relatively
smooth terrain, although speedwaslimited
due to vehicle
constraints [IO]. These algorithms were extended for use on
uneven terrain by the addition of force sensors and attitude
sensors [ l l ] , [12].
TerrainFollowing: A terrain scanner willbeused
to
provide terrain preview data which can then be used by the
control computer to predict foothold locations and to determine average slope and elevation of the terrain for use in
adjusting body attitude and altitude. The terrain-following
mode mayutilize free-gait algorithms [131, [141, which to date
have been implemented in computer simulations. Work is also
being done on development of a terrain scanning system [SI,
[15] needed for this mode of operation. Incorporation of some
terrain preview information has been demonstrated using the
OSU Hexapod vehicle [161.
Close Maneuvering: The operator uses a hand controller,
like a joystick,to command combinationsof forward velocity,
lateral velocity, and body rotation rates [17]. One approach,
which has been developed for the close maneuvering mode, is
referred to as the dual tripod algorithm [181. In this algorithm
the sixlegs of the vehicle are treated as two independentsets of
tripods. At all times, at least one of these tripods supports the
vehicle. Leg motion is limited only by geometrical considerations, and there are no time-sequencing constraints as in the
typical wave-gait formulation. A predominant problem in the
close maneuveringmode has been gait transitions as the
direction ofbodymotionis
changed.Previousalgorithms
required a trade-off of speed versus agility; the velocity input
commandsneeded a longtime-constant filter in order to
maintain smooth motion. The dualtripod algorithm has no gait
transition problem, so it does not require filtering of the input
commands.The result is anextremely agile walking algorithm, well suited for the close-maneuvering mode. Unlike

a general wave gait, the dual tripod scheme does not use the
maximum possible number of supporting legs for agiven
speed, but since the vehicle must already be designedso that it
can be supported by three legs, this does not pose a problem
for close maneuvering mode. The dual tripod algorithm has
been implemented on the OSU Hexapod and simulated on the
ASV.Other recent workdealing with this modehas been
performed by Lee [191.
Precision Footing: In situations involving very irregular
terrain, the operator may want to control individual legs and
body motion with a joystick, keyboard, or other means. The
precision footing mode is, by definition, veryoperator
intensive andmaneuvering the vehicle with this type of control
modecouldbe
anextremelycomplex
task. To make this
control mode useful, it is essential that the precision-footing
computer control algorithm includefeatures to aid the operator
as much as possible without greatly restricting the freedom of
movement inherent to this mode.
This paperis primarily concernedwith the precision-footing
mode of operationand, in particular, the development of
automaticbody
regulation schemes that allowautomatic
movement of the vehicle body in order to aid the operator in
maneuvering
the
vehicle. Section I1 describes how the
operator would use the precision footing mode to control the
vehicle. Since this mode would be used on irregular terrain
where the operator i s concerned with the vehicle tipping over,
a measure of stability is very important and will be discussed
in Section 111. This measure, together with a consideration of
constraints on kinematic limits of individual legs, leads to two
new control schemes described in Section IV.
11. PRECISION-FOOTING
MODE
The precision-footing operational mode can provide maximum maneuverability, particularly for complex tasks such as
climbing over large obstacles or crossing ditches. However,
the control algorithm should provide
the operator asmuch help
as possible in order to alleviate some of the burden of
manipulating the body and limbs.
A specific computer control algorithm has been developed
to implement the precision-footing operational mode [ 181. A
variety of features have been incorporated into this algorithm
to helpsimplify
the operators control task,to
provide
necessary information to the operator, and to assure safety.
For theOSUHexapod the vehicle operator can issue commands to the algorithm by using a three-axis joystick and a
selected set of keys on the computer terminal keyboard.When
construction of the ASV is completed, the operator control
mechanism will consist of a custom-designed arm controller
and a set of function-select switches [9]. Regardless of the
hardware interface, the algorithm functions are the same.
By choosing one of a set of six switches, the operator can
select the desired foot to be moved. The operator canthen use
the joystick to command the desired velocities of the foot in the
longitudinal, lateral, and vertical directions. Foot movement
can be simplified for the operator by theuseofJacobian
control and resolved motion rate control [20]. This allows the
operator to specify rectilinear velocities of the foot rather than
specifying actuator velocities.

134

IEEE JOURNAL OF ROBOTICS AND AUTOMATION, VOL. RA-I, NO. 3, SEPTEMBER 1985

exceed their kinematic limits. The operator is inhibited from

lifting a critical foot, which would cause the body to be
statically unstable. Also, the position of the body center of
gravity is monitored to insure that the body is not moved to a
statically unstable position. These automatic monitoring features provide safeguards in case the operator does not heed the
information provided via the graphics display, or if the
operator decides to continue a certain leg or body movement
until a limit is reached.

111. ENERGY
STABILITY
MARGIN

Fig. 2.

Graphics displayprovidesessentialguidanceinformation
operator of a walking machine vehicle.

to the

Using a function-select switch, the operator can choose to

control the movement of the vehicle body, using the joystick to
command the three linear body velocities of longitudinal body
motion, lateral body motion, and body altitude. Similarly, the
operator can choose to control the body attitude, using the
joystick to command the three angular body velocities of pitch,
roll, and yaw.
The computer algorithm provides feedback information to
the operator via a graphics display terminal (Fig. 2 ) . Information contained on the display consists of
a) the location of each supporting foot, denoted by an X;
b) the polygon whose vertices consist of the vertical
projection of the supporting feet;
c) the predicted polygon of b) that would result if a foot
being controlled individually in the air were immediately
lowered;
d) the vertical projection of the center of gravity (denoted
by the small square, near the center of the polygon);
e) the reachable area of each f w t (denoted as circles
around each foot);
f ) any critical feet (denoted by a box around that foot);
g) the Energy Stability Levels (a measure of vehicle
stability discussed in Section 111); and
h) the pitch and roll attitude of the vehicle body.
This graphics display provides essential information to
simplify the operators control task and, because of the
graphical format, the operator canquickly assimilate the
needed information. The operator can readily discern the
location of the support feet, which feet can or cannot be
moved, and within what area each foot can be moved.
Furthermore, the operator receives continuous feedback on the
stability of the vehicle, so the effects of bodyand limb
movements can be quickly evaluated, and the operator can
avoid placing the vehicle in an unstable configuration.
The algorithm includes various automatic monitoring features to assure the safety of the operator and vehicle. The
positions of all legs are monitored to insure that they do not

A critical foot is definedas a foot which, if lifted, would cause the body
to
be statically unstable.

An important consideration in the development ofany

control algorithm for a legged vehicle is the ability to maintain
stability. If atany point during the locomotion the vehicle
becomes unstable, there is the possibility that the vehicle will
overturn, unless the vehicle can dynamically compensate in
such a way as to remain upright [21]. This paper will only
consider the situation of static stability.

A . Previous Measures of Stability

Previous formalizations of the criteria for determining the
stability of a legged vehicle have been based on the assumption
of constant speed, straight line locomotion over flat terrain.
Basedupon
these assumptions, McGhee and Frank [22]
developed a series of definitions and theorems concerning the
static stability of a legged machine. These criteria were later
generalized to the situation for rough terrain [ 131. The
following definitions are the basis for determining if a vehicle
is statically stable.
Definition I: The supportpattern associated with a given
support state is the convex polygon, in a horizontal plane,
which contains the vertical projection of all of the supporting
feet [ 131.
Definition 2: The magnitude of the static stability
margin for an arbitrary support pattern is equal to the shortest
distance from the vertical projection of the center of gravity to
any point on the boundary of the support pattern. If the pattern
is statically stable, the stability margin is positive. Otherwise it
is negative [ 2 2 ] .
Until recently, the majority of research activity has dealt
with locomotion over relatively level terrain, and the previous
definition of stability has been extremely useful. However, the
static stability margin is independent of height, since it is based
solely upon the vertical projections onto a horizontal plane.
Because the precision footing operational mode is intended for
use on very rough terrain, it is necessary to have a measure of
stability which takes into account the effects of uneven terrain.

B. Calculation of the Energy Stability Margin

Consider, as an example, the situation depicted in Fig. 3.
The vehicle has four supporting legs and is standing on an
inclined plane, with the body horizontal. According to the
previous definitions, the support pattern, in this case, is a
rectangle formed by the vertical projection of the four
supporting feet onto the horizontal plane. Assuming that the
center of gravity of the vehicle is located at the center of the
vehicle body, then the position shown represents the maximum
static stability margin for this type of situation since the

MESSURI AND KLEIN: AUTOMATIC BODY REGULATION OF A LEGGED VEHICLE

135

Fig. 3 . Vehicle standing on an inclined plane with the body horizontal. The
projection of support feet into a horizontal
planedefines the support
pattern while the curve connecting the tipsof the support feet (shown in
dotted lines) defines the support boundary.

vertical projection of the center of gravity will be at the center

of the rectangular support pattern. However, intuition seems
to indicate that the vehicle is more likely to tip downhill
rather than uphill. This suggests thatmaximum static
stability would be achieved for the given situation if the body
were shifted some distance in the uphill direction, to the point
where there would bean equallikelihood of a downwardtip or
an upward tip.
Thisobservation leads to the realizationthatthe
static
stability margin does not provide a sufficient measure for the
amount of stability when the terrain is not a horizontal plane,
although it does provide a limit which indicates whether the
body is stable or unstable. In order to take into account the
effects of uneven terrain a measure of stability, the energy
stability margin, has been developed. The following definitions form a basis for determining the energy stability margin.
Definition 3: The support boundary associated with a
given support state consists of the line segments which connect
the tips of the support feet that form the support pattern.
Notice that since the support pattern is a convex polygon, its
vertices may not consist of all the support feet. Likewise, the
vertices of the support boundary may not consist of all the
support feet. Also, notice that the supportboundary is a threedimensional curve,as opposed to the two-dimensional support
pattern.
Definition 4: The energy stability level associated with a
particular edge of a support boundary is equal to the work
required to rotate the body center of gravity, about that edge,
to the position where the vertical projection of the body center
of gravity lies along that edge of the support boundary.
Definition 5: The energy stability margin for an arbitrary
support boundary is equal to theminimumof
the energy
stability levels associated with eachedgeof
that support
boundary.
The energy stability margin gives a quantitative measure of
theimpact energy whichcanbesustained
by the vehicle
without overturning. Thismeasure is very similar to the
concept of disturbance capability discussed by Frank [23] for
biped locomotion since both are based on potential energy.
The present definition and the resulting computational formulas, however,are applicable toawiderrange
of terrain
conditions.
For very irregular terrain there is a geometric possibility
that even whenthe supporting feet form aconvex polygon, the
body may notbe able to rotate outward abouta line between an
adjacent pair of feet because another foot braces the vehicle

Fig. 4. Side view of theconfigurationinFig.

3 , showingageometrical
comparison of the energy stability level for the front and rear edges
of the
support boundary.

against turning over. This unlikely possibility does not enter

into the energy stability measure and therefore this measure
provides a conservative estimateof instability danger.
The application of these definitions is demonstrated in Fig.
4. Again, the vehicle has four supporting legs and is standing
on an inclined plane, with the body horizontal. Fig. 3 shows
that the support boundary in this case lies in the plane of the
incline. Fig. 4 shows a geometrical comparison of the energy
stability levels for the front rear edges of the support
boundary.The line segmentfrom point Fl (rearedge of
support boundary) to the point CG (body center of gravity)
represents the radius R 1 of an arc, which the body center of
gravity would trace if the body were rotated about the rear
edge of the support boundary. If the body were rotated to the
position where the body center of gravity is vertically above
the rear edge of the support boundary, thenthe vehicle would
beon the verge of instability correspondingtozero static
stability marginaccordingto
Definition 2. Thechange in
vertical height through which the body center of gravity is
moved from its original position to this position of zero static
stability margin is givenby the distance hl. Therefore, the
amount of potential energy required to rotate the body center
of gravity, about the rear edge of the support boundary, from
its original position to the point of zero static stability is rnghl,
where m represents the mass of the vehiclebody,and
g
represents the acceleration due to gravity.Likewise,the
amount of energy required torotate the body center of gravity
about the front edge of the support boundary, to the point of
zero static stability, is mgh2. Since h2 equals h, + Ah, the
situation depicted in Fig. 4 would require less energy to
overturn the vehicle about the rear support legs as opposed to
the front support legs. Therefore, if it were desired toshift the
body to a position of greater overall stability, the body should
be shifted such that hl equals h2, at which point the energy
stability levels for the front edge andrear edge would be equal.
Such a shift, of course, is implemented by coordinated leg
motion.
The configuration shown in Fig. 4 represents a relatively
simple casein which the location of the body center of gravity
can be solved geometrically such
that hl = h2.It can be shown
that the locus of all points representing the location of the body
center of gravity with hl equalto h2 is described by the

136

IEEE JOURNAL OF ROBOTICS AND AUTOMATION, VOL. RA-1, NO. 3, SEPTEMBER 1985

hyperbola, which has focal points at Fl and Fz. This hyperbola

is indicated by a dashed curve in Fig. 4. Although hl equals h2
when the body center of gravity is located anywhere along this
hyperbola, notice that the magnitude of the energy stability
level increases as the vehicle bodyheight
is decreased.
Becausebody height directly affects the obstacle clearance
capability of the vehicle, the vehicle body is restricted to move
only within the present plane of the body. With this restriction,
the point Q in Fig. 4 represents the desired position of the
body center of gravity so that hl equals h2.
It should be noted that the previous discussion was
concerned only with the front and rear edges of the support
boundary. However, the definition of the energy stability
margin requires the consideration of all edges of the support
boundary. Also in more general situations the position of the
body and legs, especially in the case of very rough terrain,
maynot permit a simple geometric solution. Therefore, a
general equation has been derived which gives a measure of
the energy stability level about any given edge of the support
boundary. Notice that since the potential energy is given by

PE = mgh

(1)

and since the mass rn and acceleration of gravity g are

constant, then for the purpose of finding the energy stability
level it is necessary to find the vertical height h through which
the body center of gravity wouldmove if the body were
rotated, about the given edge of the support boundary, to the
point of zero static stability margin.
Consider the general situation depicted in Fig. 5 , where
points Fl and F2 represent the footholds of two support feet,
and the line segment connecting Fl and F2 represents one edge
of the support boundary. Plane 1 is a vertical plane containing
line F,F2. The point CG represents the location of the body
center of gravity. Vector R is a vector from line F1F2to point
CG, and is orthogonal to line F1F2.Unit vector 2 represents
the upward vertical direction. The vector R ' is obtained by
rotating vector R , about line FlF2, untilit lies in plane 1.
Defining 0 as the angle between R and I? ' , and \k as the angle
between R' and 2, the vertical height h through which the
point CG moves when the vector R is rotated to the vertical
plane is given by

Fig. 5. Derivation of theenergystabilitylevelequation.Theline

F,F,
represents one edge of a support boundary, the point CG represents the
body center of gravity, and the vertical distanceh gives a measure of the
energy stability level.

couldbemoved and still maintain a given energy stability

margin. For any configuration, a family of level energy curves
canbe drawn, each curve representing a different level of
energy stability margin.
Figs. 6 and 7 show some Level Energy Curves for various
configurations. The level energy curves can be thought of as a
contour map of the energy stability margin. In these drawings,
the X ' s represent the support feet which form the support
boundary. The position of these X ' s represents the vertical
projection of the support feet, onto the plane of the body. The
polygon formed by interconnecting these X ' s represents the
curve of zero energy stability margin.

(2) D. Optimally Stable Position

To fully apply the concept of energy stability margin in the
During operation of the walking vehicle, the location of all control of a legged vehicle, it would be desirable to find the
the feet can be found with respect to the body center of position to which the body center of gravity could be moved in
gravity, and the vector formulation provides a simple efficient order to obtain the maximum energy stability margin for a
method of calculating the energy stability margin for any given configuration. In other words we would like to move to
position of the body or legs. It is a general formulation which the highest level on the energy stability margin surface. It
allows for any type of terrain condition.
should be notedfrom the level energy curves of Fig. 6 that the
C. Level Energy Curves
energy stability margin surface does not necessarily have a
Having developed a general equation that allows the unique peak point but can instead have a ridge as its highest
calculation of the energy stability margin, it is possible to level. This leads to the following definition.
Definition 6: An optimally stable position is any position
analyze the energy stability margin for various configurations
of body and leg positions. By considering the energy stability in the plane of the body at which the energy stability margin
margin as a function of the position of the projection of the would be maximal if the center of gravity were moved to that
center of gravity in the present plane of the body, one can draw position.
A study of the level energy curves discussed in Section 111-C
level energy curves, which are the locus of all points, in the
present plane of the body to which the body center of gravity indicates that the three-dimensional energy surface is monoh = 1R/(1-cos 0)

COS

!I?.

137

MESSURI AND KLEIN: AUTOMATIC BODY REGULATION OF A LEGGED VEHICLE

'

I'

#
1-36

IO
FORE-AFT

36 I
(IN)

(0

1-36

IO
FORE-AFT

(IN)

m
rn
1-36

FORE-AFT

(IN)

I
36 I

IO
FORE-AFT

A-

31

(IN)

(C)
Fig. 6 . Optimal paths of the vehicle centerof gravity, for variousvehicle
configurationsforseveraldifferentstartingpoints.The
dotted lines
represent level energy curves. (a) Vehicle standing on level terrain, body
horizontal. (b) Vehicle standing on a20" inclined plane, body horizontal.
(c) Vehicle standingon a 20" inclined plane,body horizontal, left front leg
off the ground.

tonically increasing up to the maximal level. Because of this

characteristic, it is possible to find the optimal path that the
body center of gravity should follow in order to shift from its
present location to the optimally stable position by utilizing the
gradient of the energy stability margin. Beginning at the given
present location ofthebody center of gravity, an iterative
technique can be used to find the optimal path by following the
direction of the energy stability margin gradient. This optimal
path traces the steepest slope from the given body center of
gravity location to the maximal level of the energy stability
margin, and this maximal point isan optimally stable position.
Recall from Definition 5 that the energy stability margin is
the minimumofall
the energy stability levels for a given
support boundary.As can be seenin Figs. 6 and 7, the trace of
a level energycurve involves sharp changesin direction.
These direction changes indicate that the minimum energy

Fig. 7. Optimal pathsof the vehicle centerof gravity, for various vehicle
on
configurations for several different starting points. (a) Vehicle standing
a 20" inclined plane,body horizontal, with right front legand left rear leg
on rocks. (b) Vehicle standing on a 20" inclined plane, body pitched at
20". (c) Vehicle standing on level terrain, body horizontal, with a tripod
support phase.

stability level hasswitchedfromoneedge

of the support
boundary to a different edge. These direction changes on the
level energycurvescorrespondto
ridges on the threedimensional energy surface.
The optimal path leading to the optimally stable position
traces the maximum increase of the minimum energy stability
level. Since the algorithmtotrace
the optimalpath
is
implemented on a digital computer, the gradient of the energy
stability margin is calculated at discrete intervals. Therefore,
as the optimal path approaches a ridge on the energy surface
there will be sharp direction changes in the optimal path,
because the direction of the optimal path is normal to the level
energy curves. The effect is that the optimal path oscillates
back and forth across the ridge; an example is shown in Fig. 8.
Although this optimal path does lead to the Optimally Stable

138

IEEE JOURNAL OF ROBOTICS AND AUTOMATION, VOL. R A - I , NO. 3, SEPTEMBER 1985

:
FORE-AFT

( I N)

Fig. 8. Tracing the optimal paths of the vehicle center of gravity from four different starting points, when a blending function is not
included. Note the oscillations, which indicate sharp direction changes in the optimal path, due to the discrete time calculation of
the energy stability margin gradient.

Position, the oscillation is undesirable for use in controlling a

physical system such as a legged vehicle.
To reduce oscillation when the optimal path follows aridge
on the energy surface, a blending function is introduced.
The objective of the blending function is to allow the optimal
path to follow the center of the ridge without oscillating across
the ridge. The blending function isonly active whenthe
optimal path enters within a narrow band on either side of a
ridge. Outside this band, the optimal path traces the gradient
of theminimum energy stability level, as usual. However,
once inside the band the optimal path traces a path determined
by the weighted combination of the gradients of the two
smallest energy stability levels, rather than just the smallest.
Figs. 6 and 7 show the traces of some optimal paths
or various
starting vehicle configurations with the blendingfunction
included. These curvesshow some of the optimal paths for the
body center of gravity to move from variousinitial positions to
an optimally stable position. Notice, as mentioned previously,
that there is not always a unique optimally stable position.
Examination of Figs. 6 and 7 also provides a comparison
between the Static and Energy Stability Margins. For example, for Figs. 6(a) and (b) and 7(a) and (b) the static stability
marginwouldbeoptimized
at the origin, whilethe level
curves of energy stability margin show the center of gravity
should be moved uphill. Several of these figures also show
cases in which the level curves are significantly different from
the shape of the support polygon. Thus,because the concept of
energy stability margin takes into account such factors as the
vertical height of the body center of gravity, the pitch and roll
of the vehicle body, and the location of the support feet in
three-dimensional space, it providesa more accurate and
quantitative measure of stability thantheconcept
of static
stability margin, although both of these concepts provide a
qualitative measure to determine whether or not a vehicle is
statically stable.

IV. AUTOMATIC
BODYREGULATION
Maneuvering a vehicle overrough
terrain canbean
extremely complextask. The features in the precision footing
control algorithm which have been discussed thus far simplify
the operators control task, provide feedback information,and
assure safety. To enhance the capabilities of the precision
footing operational mode, the concept of automatic body
regulation was developed whereby the operator can allow the
computer control algorithm to automatically adjust the position of the vehicle body in accordance with some predefined
criteria. Two automatic body regulation schemes were developed and have been
incorporated into the computer control
algorithm. The two schemes are
referred to as body accommodation and body stabilization.

A . Body Accommodation
As explainedin Section 11, the precision-footing control
algorithm enablesthe vehicle operator toselect and control the
motion of individual legs of the vehicle. The algorithm also
allows the operator to directly control the body motion in its
six degrees-of-freedom.These features make the vehicle
highly maneuverable for extremely rough terrain situations.
However, eachvehicle leg has a limited reach, due to the legs
kinematic limits. This limited reach may sometimes require
the operator to perform increased maneuvering in order to
place a foot at a desired foothold. In order to alleviate the
operator of some ofthis maneuvering task, a body accommodation feature was incorporated into the control algorithm.
Inthe precision footing control algorithm,whenever the
operator selects a vehicle leg for individual leg control, the
position of that leg is monitored to insure it is never extended
beyond the kinematic limits. With the body accommodation
scheme, if this individually controlled leg reaches the kinematic limits, thenthe vehicle body is automatically commanded to move in such a direction as to accommodate the
operators desired motion of that individual leg. This accom-

139

MESSURI AND KLEIN: AUTOMATIC BODY REGULATION OF A LEGGED VEHICLE

modation increases the ability of the vehicle to reach a desired

foothold. Of course, the position of all ofthe support legs must
be monitored during this accommodation movement to insure,
that the body movement does not extend a support leg beyond
its kinematic limits. Also, vehicle stability must be monitored
to insure that the body is not shifted to a position where the
vehicle is unstable. This concept of body accommodation is
analogous to the situation where a human may lean his body in
such a manner as to helphimreachhishand
or foot to a
desired position.
The body accommodationscheme greatlyenhances the
maneuverabilityof the vehicle during theprecision-footing
operational mode. Since the movement of the vehicle body is
automaticallycommanded
to accommodate the operators
desired movement of an individual leg, this feature reduces the
complexity of the operators control task. Body accommodation allows the operator to manipulate a vehicle leg within a
much larger volume than the typical reachable volume of each
leg, asdeterminedby the kinematic limits. Therefore, the
vehicle can be maneuvered over a much larger range, while
simplifying the operators control task.

B. Body Stabilization
While traversing a region of extremely rough terrain, the
vehicle operator may find that, due to the terrain conditions,
the vehicle body and legs have become oriented into a rather
precarious configuration. The feedback information provided
via the cockpits graphics display (as discussed in Section 11)
indicates such things as the critical support feet which cannot
be lifted, the support polygon, and the energy stability level
for eachedge ofthe supportboundary.Furthermore,
the
control algorithm includes safeguards to keep all legs within
kinematic limits and to maintain static stability. If the display
indicates that the present vehicle situation has a low energy
stability margin, the operator may desire to shift the body to a
positionof
greater stability beforeproceeding
with leg
maneuvers. This repositioning task is simplified by incorporating a body stabilization feature into the control algorithm.
The body stabilization feature is activated when theoperator
chooses the appropriate function-select switch. The body
stabilization scheme
determines
the optimal
path
to an
optimally stable position, based
upon
the current body
orientation and leg positions. The vehicle bodyisthen
automatically shifted, with the bodymovement
restricted
within the present plane of the body, to the point where the
body center of gravity coincides with the optimallystable
position. After body stabilization is completed, the operator
can proceed with whatever maneuvers are desired. The body
stabilization feature can also beusefulwhen
the operator
desires to position the vehicle at an optimally stable position
before attempting to maneuver across some obstacles. Since
the body stabilization routine is completely automatic when
activated, it permits the vehicle stability to be optimized
quickly and easily, for any body orientation and leg positions.
It should be noted that there may be occasions where the
body orientation and leg positions are such that the leg
kinematic limits prohibit thebodymovementnecessary
to

Fig. 9. Demonstration of the body stabilization mode. (a) Photo of OSU

Hexapod on irregular terrainand (b) corresponding graphics display. The x
symbol, located inside the support polygon, indicates the optimally stable
position to which the center of gravity should move. (c) Graphics display
showing completion of body stabilization.

have the body center of gravity coincide with the optimally

stable position. In these instances, the body stabilization
scheme wouldmovethebody
along the optimal path until
further movement is prohibited by the kinematic limits. The
result would be that the body is positioned at the point of
maximum stability allowable with the present body orientation, leg positions, and kinematic limits.

140

IEEE JOURNAL OF ROBOTICS AND AUTOMATION, VOL. RA-I, NO. 3, SEPTEMBER 1985

The operation of the body stabilization routine is demonstrated in Fig. 9. Fig. 9(a) shows the OSU Hexapod in a rather
precarious configuration. where the vehicle has climbed over
an obstacle and
has
been
maneuvered intoan
awkward
position. The graphics display information corresponding to
this situation is shown in Fig. 9(b). The small square inside the
support polygon represents the location of the body center of
gravity. The X symbol, which appears when the body
stabilization routine is activated, indicates the optimally stable
position of the center of gravity. Fig. 9(c) shows the results
after invoking the body stabilization routine. The graphics
display information indicates that the body center of gravity is
nowat the optimally stable position as evidenced by the
location of the present center of gravity and the magnitudes of
the energy stability levels.
The general formulation of the equation for calculating the
energy stability margin allows an interesting extension. Since
the OSU Hexapod vehicle is equipped with force sensors on
each foot, this force information can beused to actively
compute the location of the vehicle center of gravity. This
active center of gravity would take into account any effects of
vehicle cargo loading, changes in fuel level, etc. The energy
stability margin could be calculated using the active center of
gravity. The result is improved vehicle stability since any
variation in the vehicle center of gravity could now be
compensated.
V. CONCLUSION
A computer control algorithm has been developed for the
precision-footing operational mode. This algorithm includes
the incorporation of a body accommodation feature and a body
stabilization feature. These automatic body regulation schemes
allow greater vehicle maneuverability, particularly during
rough-terrain locomotion.
One of the major developments presented in this paper is the
concept of energy stability margin. A general equation was
introduced which allows the calculation of the energy stability
margin for any given position of the bodyand legs. By
utilizing the gradient of this function, an optimal path can be
found leading from a given initial location of the body center
of gravity to an optimally stable position. A blending function
was introduced to reduce the oscillation which can occur when
the optimal path approaches a ridge on the energy surface.
The energy stability margin provides an accurate quantitative measure of vehicle stability, particularly for rough-terrain
conditions. Previously implemented stability criteria provided
a qualitative measure of stability, but did not fully account for
rough-terrain conditions. It is this quantitative measure,
provided by the energy stability margin, which allowed the
development of a computer algorithm for determining an
optimally stable position. This capability was incorporated
into the precision footing algorithm to achieve the body
stabilization feature.
One of the particularly interesting future applications for
these concepts and the automatic body regulation schemes
presented here wouldbein
the development of free-gait
algorithms [ 131. For example, the body accommodation
feature would provide a wider selection of allowable foot-

holds, and the body stabilization feature mightbeused

to
maintain maximal stability.
The concepts of energy stability margin and an optimally
stable position can be used in a wide variety of situations and
should lead to the development of more sophisticated control
algorithms for legged vehicles. These new algorithms can
further the realization of a legged vehicles potential for
maneuverability over rough terrain.
REFERENCES
M. G. Bekker, Introduction to Terrain-Vehicle Systems. Ann
Arbor, MI: The University of MI, 1969.
R. B. McGhee, Vehicular legged locomotion, Advances in Roboticsand Automation, (vol. I),G. N. Saridis,Ed.
Greenwich,
CT: JAI, 1984.
Int. J. Robotics Res., (Special issue on legged locomotion), vol. 3,
no. 2, Summer 1984.
R. S. Mosher, Exploring the potential of a quadruped, presented at
the Int. Automotive Engineering Conf., SAE paper no. 690191,
1969.
R. L. Briggs, A real-timedigital system for control of ahexapod
vehicle utilizing force feedback, Ph.D. dissertation, The Ohio State
University, Columbus, OH, 1979.
K. J. Waldron, V. J. Vohnout, A . Pery, S. M. Song, and S. L. Wang,
Mechanicaland geometric design of the adaptive suspensionvehicle, CISM-IFToMM Symp. Theory and Practice of Robots and
Manipulators, June 1984.
W. R. Ferrell and T. B. Sheridan, Supervisory control of remote
manipulation, IEEESpectrum, vol. 4, no. 10, pp. 81-88, Oct. 1967.
R. B. McGhee, D.E. Orin, D.R.Pugh, and M. R. Patterson, A
hierarchically-structuredsystem for computer control of ahexapod
walking machine, in Proc. 5th IFToMM Symp. Robots and
Manipulator Syst., 1984.
D. B. Beringer, The design of manually operated controls for a sixdegree-of-freedom groundborne walkingvehicle: control strategies
and stereotypes, Proc. Ninth Symp. Psychology inthe 000,1984,
pp.188-192.
D. E. Orin, Interactive control of a six-legged vehicle with optimization of both stability and energy, Ph.D. dissertation, The Ohio State
University, Columbus, OH, 1976.
C. A. Klein and R. L. Briggs, Use of active compliance in the control
of legged vehicles, IEEE Trans. Syst., Man, Cybern., vol. SMC10, no. 7, pp. 393-400, 1980.
C. A. Klein, K. W. Olson, and D. R. h g h , Use of force and attitude
sensors for locomotion of a legged vehicle over irregular terrain, Int.
J. Robotics Res., vol. 2, no. 2, pp. 3-17, 1983.
R. B. McGhee and G.I. Iswandhi, Adaptive locomotionofa
multileggedrobot over rough terrain, IEEE Trans. Syst., Man,
Cybern., vol. SMC-9, no. 4, pp.176-182, Apr. 1979.
S. H. Kwak, A simulation study of free-gaitalgorithms for omnidirectional control of hexapod walking machines, M.S. thesis, The Ohio
State University, Columbus, OH, 1984.
M.R. Patterson, J. J. Reidy, andB. B. Brownstein, Guidance and
actuation techniques for an adaptively controlled vehicle, Tech. Rep.,
contract MDA903-82-C-0149,Battelle
Columbus Laboratories, Columbus, OH, 1983.
F. Ozguner, S . J. Tsai,and R. B. McGhee, An approach to the use of
terrain preview information in rough terrain locomotion by a hexapod
walking vehicle, Int. J. Robotics Res., vol. 2, no. 2, pp. 3-17,
1984.
D. E. Orin, Supervisory control of a multilegged robot, The Int. J.
Robotics Res., vol. 1, no. 1 , pp. 79-91, Spring 1982.
D.A. Messuri, Optimization of the locomotion of a legged vehicle
with respect to maneuverability, Ph.D. dissertation, The Ohio State
University, Columbus, OH, 1985.
W. J. Lee, A computer simulation study of omnidirectional supervisory control for rough-terrainlocomotionbyamultileggedrobot
vehicle, Ph.D. dissertation, The Ohio State University, Columbus,
OH, 1984.
D. E. Whitney, Resolved motion rate control of manipulators and
human prostheses, IEEE Trans. Man-Mach. Syst., vol. MMS-10,
no. 2 , pp. 47-53, 1969.
M. H . Raibert and I. E. Sutherland, Machines that walk, Scientific
American, vol. 248, no. 2, pp. 44-53, Jan. 1983.

MESSURI AND KLEIN: AUTOMATIC BODY REGULATION

LEGGEDOF A

141

VEHICLE

[22] R. B. McGheeandA.
A.Frank, On the stability of quadruped
creeping gaits, Mathematical Biosciences, vol. 3, no. 3, pp.331351, Oct. 1968.
[23] A. A. Frank, On the stability of an algorithmic biped locomotion
machine, J. Terramechanics, vol. 8, no. 1, pp. 41-50, 1971.

the OhioStateUniversity.Heiscurrentlyemployed
Engineer in the Advanced EngineiringDepartment
Division.
Dr. Messuri isa member of Tau Beta Pi.

Charles A. Klein (83) was

Dominic A. Messuri was born on October 26,

1953; in Youngstown, Ohio. He received the B.E.
and M.S. degrees in electrical engineering from
Youngstown State University, Youngstown, Ohio,
in 1975.and 1978, respectively. He received the
Ph.D. degree in elqctrical engineering from the
1985.
State
Engineering
Ohio
the at
Ohio State University, Columbus, in University,
From 1976 to 1980 he was employed as a Design
Engineer in the Advanced Engineering Department,
Packard Electric Division, General Motors Corporation, Warren, Ohio. During his graduate studies,
he was a Graduate Rest:arch Associate in the Digital Systems Laboratory of

as a SeniorProject
of PackardElectric

born in Aurora, IL,

on February 5, 1949.
He
received B.S.
the
degree
in electrical
engineering
and computer
science,
and
the M.S. and Ph.D.
degrees
in electrical
engineeringUniversity
from
the
of Illinois,
Urbana-Champaign, in 1971,
1972,
and 1975,
respectively.
In 1977 he joined
Department
the Electrical
of
Columbus,where
he currentlyholdsthe
position of
AssociateProfessor.Professor Klein teaches and
performs research in the fields of robotics, digital
systems, and computer graphics.
Dr. Kleinis a member of the National Computer Graphics Association.