(How to Be Sure Your Robot Will Turn)
by Chris Hibner, Team 308 – TRW and Walled Lake Schools: The Monsters
contact: hibnerc@monsters308.com
Rev: 10/2/2003
Background
A very common problem experienced by teams is that they put a lot of time into a four
wheeldrive system only to find out that it isn’t able to turn. Every year there are
countless questions on the FIRST forums asking for help with this subject. The purpose
of this paper is to explain the physics behind the fourwheeldrive system, and to provide
some guidelines that teams can use to help design their drive train so that it will turn.
One goal of this paper is to provide the detailed math to illustrate how this problem is
solved. This is presented for the following reasons: a) to better understand the details of
how the system works; b) to have an idea of how engineering principals can be used to
solve a “mysterious” problem such as this; and c) to use as an exercise for training new
team members to see if the problem can be solved independently and verified against this
paper.
If the detailed math is not of interest, you may skip to the Conclusions and Extensions
sections. These sections will contain some more qualitative guidelines and rulesof
thumb to help you with designing a drive train that will turn.
Details
When considering the physics of the system, one should keep in mind that this system is
modeled with standard engineering practices. In other words, the physics aren’t “exact”.
It is not the purpose to find an “exact” solution, since that would take much more work
than is necessary, and is probably not possible to begin with. Therefore, some
assumptions and simplifications are made (friction is a good example). As the punch line
of the old engineering joke goes: “it isn’t exact, but it’s close enough for practical
purposes.”
A schematic drawing of a basic fourwheeldrive robot is shown below. The forces
acting on each wheel are shown. F
??
is the force on each wheel that is provided by the
motors, while R
??
is the reaction force (due to friction) from the robot attempting to turn.
The subscript corresponds to the location: LF = leftfront, RR = rightrear, etc.
Other notation:
L
WB
= length of the wheel base
L
TW
= length of the track width
COM = center of mass
COA = center of area; geometric center of the robot;
L
CY
= distance from the center of area to the center of mass (note that it is assumed that
the robot is left/right symmetric).
M
T
= the resultant turning moment
Static Analysis
We will be determining what is necessary to make the robot just start to turn. Therefore,
this can be treated as a static (rather than dynamic) problem. It is common for rigid body
dynamics problems to use the center of mass as the reference point of the body.
Therefore, moments will be taken about the center of mass.
F
LF
F
LR
F
RR
F
RF
R
LF
R
RR
R
LR
R
RF
C
OA
C
OM
L
TW
L
WB
L
CY
M
T

.

\

=
2
,
2
WB TW
OA
L L
C
The first step will be to determine the moments about the center of mass. The
contribution of each wheel is shown below (remember that Moment = Force X distance;
in this case the distance is the perpendicular distance from the force to the center of
mass).
We will assume that the gearing is low enough that the drive train is traction limited. See
the section “Proper Gearing” to see how to make sure your drive is geared low enough to
allow it to turn.
Assuming that the drive train is traction limited, the forces at each wheel will be the
normal force at that wheel multiplied by the coefficient of friction between the wheel and
the ground. Here, we assume that the wheel can have different coefficients of friction in
the longitudinal (x) direction and the lateral (y) direction. Therefore the forces at each
wheel can be written as follows:
In the above equations, N
R
is the normal force on a rear wheel and N
F
is the normal force
on a front wheel.
Now, substitute the friction relationships into the wheel moment equations:
R y RR R x RR
F y RF F x RF
F y LF F x LF
R y LR R x LR
N R N F
N R N F
N R N F
N R N F
µ µ
µ µ
µ µ
µ µ
= =
= =
= =
= =

.

\

− − =

.

\

+ − =

.

\

+ − =

.

\

− − =
CY
WB
RR
TW
RR RR
CY
WB
RF
TW
RF RF
CY
WB
LF
TW
LF LF
CY
WB
LR
TW
LR LR
L
L
R
L
F M
L
L
R
L
F M
L
L
R
L
F M
L
L
R
L
F M
2 2
2 2
2 2
2 2
Now that the moment contribution of each wheel is known, sum the moments to find the
resultant turning moment:
Collecting like terms, the moment equation can be rewritten as follows:
The first term can now be simplified using the distributive property to remove the factor
of 2. The term in the brackets can also be simplified by using the distributive property.
Using the distributive property again on the term in the brackets, the above equation can
be rearranged to get equation 1:
At this point, a relationship needs to be determined between the normal forces N
R
and N
F
,
the robot weight W, and the location of the center of mass. The following picture is a
side view of the robot. Notice that the force depicted on the front is 2 N
F
. This is the
case since there are 2 front wheels – not one. Likewise for the rear.
(
¸
(
¸

.

\

− + 
.

\

+ + 
.

\

+ + 
.

\

− −
+ + + =
CY
WB
R y CY
WB
F y CY
WB
F y CY
WB
R y
TW
R x
TW
F x
TW
F x
TW
R x T
L
L
N L
L
N L
L
N L
L
N
L
N
L
N
L
N
L
N M
2 2 2 2
2 2 2 2
µ µ µ µ
µ µ µ µ
( )
(
¸
(
¸

.

\

+ + 
.

\

− − + =
CY
WB
F CY
WB
R y F R
TW x
T
L
L
N L
L
N N N
L
M
2
2
2
2 2 2
2
µ
µ
( )  
CY F WB F CY R WB R y F R TW x T
L N L N L N L N N N L M 2 2 + + − − + = µ µ

.

\

− − =

.

\

+ − =

.

\

+ − =

.

\

− − =
CY
WB
R y
TW
R x RR
CY
WB
F y
TW
F x RF
CY
WB
F y
TW
F x LF
CY
WB
R y
TW
R x LR
L
L
N
L
N M
L
L
N
L
N M
L
L
N
L
N M
L
L
N
L
N M
2 2
2 2
2 2
2 2
µ µ
µ µ
µ µ
µ µ
( ) ( ) ( )   eqn(1) 2
F R CY F R WB y F R TW x T
N N L N N L N N L M − − + − + = µ µ
Using a static analysis again, take the sum of the forces in the z direction:
Rearrange the above equation to get equation 2:
Now take the sum of the moments about the center of area:
Rearrange the above equation to get equation 3:
Recall equation 1:
L
CY
2N
R
L
WB
2N
F
Front
W
COA
0 2 2
: 0
= − +
=
∑
W N N
F
F R
z
( )
CY F R WB
CY
WB
F
WB
R
WL N N L
WL
L
N
L
N
= −
= − −
=
∑
0
2
2
2
2
: 0 M
COA
(2) eqn
2
W
N N
F R
= +
(3) eqn
WB
CY
F R
L
L
W N N = −
( ) ( ) ( )   eqn(1) 2
F R CY F R WB y F R TW x T
N N L N N L N N L M − − + − + = µ µ
Now, substitute equations 2 and 3 into equation 1 to get the following:
Rearrange the above equation to get the following:
Lastly, rearrange further to get the final turning moment equation:
Making the Robot Turn
In order to make the robot turn, the final turning moment must be greater than zero.
Therefore the following relationship must be true:
The relationship in equation 4 can be satisfied a number of ways. A few ways to satisfy
the relationship are given below.
 You can move the COM toward the front or rear of the robot. As L
CY
gets larger, the
term in the parentheses will move toward a negative number. Thus, the right side of the
inequality becomes smaller, allowing the left side to be greater than the right side; i.e. the
relationship will be satisfied.
 Assume the worstcase COM: the COM is at the COA (i.e., L
CY
= 0). Equation 4 then
simplifies as follows:
(
¸
(
¸
− − =
WB
CY WB
y
TW x
T
L
L L
W
W L
M
2
2
2 2
µ
µ
* * * * * * 4
2
* * * * * *
2
(
¸
(
¸


.

\

− − =
WB
CY
WB y TW x T
L
L
L L
W
M µ µ
! ! ! turn! robot to for true be must   (4) eqn 4
2


.

\

− >
WB
CY
WB y TW x
L
L
L L µ µ
(5) eqn
WB y TW x
L L µ µ >
(
¸
(
¸
− − =
WB
CY
CY WB y TW x T
L
L
W L
W
L
W
L M 2
2 2
µ µ
Now assume that you are using rubber wheels such that the coefficient of friction is the
same in all directions. The friction coefficients in equation 5 then cancel. Then, the
following relationship must be satisfied:
Therefore, by making the track width larger than your wheelbase, the robot will turn.
 Another method of satisfying equation 5 is to use a wheel with friction characteristics
that are different in the x and y directions. For instance, you can use holonomic wheels.
You can also use wheels that have a tread pattern that is rounded on the sides of the
wheel, but sharp in the direction of travel. This way, the wheel bites into the carpet
providing a high driving friction, but slides easily sideways.
Proper Gearing
To ensure that the robot will turn, the driving force generated by the wheels must be
enough to overcome the frictional resistance force. To simplify things (and give some
safety margin), we should select our gears such that the driving force at EACH wheel is
enough to overcome the frictional resistance force at that wheel. Therefore, the moment
about the COM at EACH wheel should be greater than zero.
In order to determine the frictional resistance at each wheel, we need to determine the
normal force at each wheel. Recall equations 2 and 3:
Notice that this is simply a system of two equations with two unknowns. Solve for the
normal forces (left as an exercise for the reader):
This shows that if Lcy is positive, the normal force will be greater at the rear of the robot.
This should make sense since if Lcy is positive, the center of mass is toward the rear of
the robot, which means that there should be more normal force on the rear wheels.
WB TW
L L >
(3) eqn
WB
CY
F R
L
L
W N N = −
(2) eqn
2
W
N N
F R
= +
WB
CY
F
WB
CY
R
L
L W W
N
L
L W W
N
2 4
2 4
− =
+ =
Therefore, to be sure that the worst case is used, the analysis will be performed for a rear
wheel (Lcy is assumed to be positive for this example).
To ensure turning, the moment about the COM for the leftrear wheel must be greater
than zero:
Substitute in the equation for the rear normal force:
Rearrange and multiply out the two terms in parentheses:
Simplifying:
Lastly, solve for the force:
Therefore, you can use the above equation to determine the minimum force necessary at
the wheels to make the robot turn. Using this force, along with the motor characteristics
and the diameter of your wheels, you can determine the gear ratio that will satisfy this
minimum force. Of course, you should use a gear ratio that provides considerably more
than this minimum force, since this force is just barely enough to make the robot start to
turn.
0
2 2
> 
.

\

− −
CY
WB
R y
TW
LR
L
L
N
L
F µ
0
2 2 4 2
> 
.

\

−


.

\

+ −
CY
WB
WB
CY
y
TW
LR
L
L
L
L W W L
F µ


.

\

− + − >
WB
CY CY CY WB
y
TW
LR
L
WL WL WL WL L
F
2 4 4 8 2
2
µ


.

\

− >
WB
CY WB
y
TW
LR
L
L L
W
L
F
2
4 2 2
µ
turn order to in satisfied be must ip relationsh This * * * * * *
4
* * * * * *
2


.

\

− >
WB
CY WB
TW
y
LR
L
L L
L
W
F
µ
Extensions
Tank Treads
The fundamental difference between tank treads and 4wheeldrive is that the force on
the treads is distributed over the length of the treads, rather than at four discrete points.
In order to take the effects fully into account, the forces and moments need to be
integrated over the length of the tread.
The details of the math will be skipped for this, jumping straight to the conclusions
instead.
The good news is that the 4wheeldrive equations can still be used. The trackwidth
(L
TW
) is still the distance between the two treads. However, the wheelbase (L
WB
) is not
simply the length of the contact patch of the tread. Instead, L
WB
is 2/3 of the length of the
contact patch of the treads. Other than this, the equations can still be used.
Conclusions
1) When designing a drive train, begin by being sure that the robot will turn. To do this,
be sure that the following relationship is true:
2) If you don’t like math, use the following rules of thumb:
 Make the track width greater than the wheel base (L
TW
> L
WB
)
 If possible, reduce the lateral friction coefficient while keeping the longitudinal
friction high (i.e., use holonomic wheels or choose a good wheel tread pattern).
 Try to move the center of mass slightly away from the center of the robot. Use
caution to not move the COM far enough so that the robot becomes unstable.
3) Once you satisfy the equation in step 1), you should have values for the important
dimensions (L
TW
, L
WB
, and L
CY
). Plug these numbers into the following equation to
determine the force necessary at each wheel to make the robot turn:


.

\

− >
WB
CY
WB y TW x
L
L
L L
2
4 µ µ


.

\

− >
WB
CY WB
TW
y
LR
L
L L
L
W
F
2
4
µ
4) Using the force calculated in step 3), a minimum gearing can be determined such that
the motors can produce the necessary force at the wheel.
Keep in mind that these equations are generated such that the robot will just barely begin
to turn. When you are creating your design, some safety margin should be included into
the calculations. For instance, if you determine in step 3 that you need to produce 50 lb
of force at each wheel, you might consider that your gearing should produce 100 lb of
force at the wheels, just to be safe. I would suggest a safety factor of at least 2 for the
gearing.
You should also keep in mind that your drive train and gearing will not be 100%
efficient. Therefore, when determining your gearing, be sure to take efficiency into
account.
Final Conclusion
The physics used to develop these equations and relationships are not exact. However,
by carefully making a few simplifying assumptions, the problem can become easy
enough to solve, and useful results can be achieved. By using the above equations and
rules of thumb, you should be able to be successful in designing a fourwheeldrive
system.
. Therefore. It is common for rigid body dynamics problems to use the center of mass as the reference point of the body. Therefore. WB 2 2 Static Analysis We will be determining what is necessary to make the robot just start to turn. this can be treated as a static (rather than dynamic) problem.Other notation: LWB = length of the wheel base LTW = length of the track width COM = center of mass COA = center of area. MT = the resultant turning moment FLF RLF RRF LWB FLR COA LCY RLR MT COM FRF RRR FRR LTW L L COA = TW . geometric center of the robot. LCY = distance from the center of area to the center of mass (note that it is assumed that the robot is left/right symmetric). moments will be taken about the center of mass.
we assume that the wheel can have different coefficients of friction in the longitudinal (x) direction and the lateral (y) direction. See the section “Proper Gearing” to see how to make sure your drive is geared low enough to allow it to turn. the forces at each wheel will be the normal force at that wheel multiplied by the coefficient of friction between the wheel and the ground. NR is the normal force on a rear wheel and NF is the normal force on a front wheel. in this case the distance is the perpendicular distance from the force to the center of mass). The contribution of each wheel is shown below (remember that Moment = Force X distance. Assuming that the drive train is traction limited. M LR = FLR M LF = FLF M RF = FRF M RR = FRR LTW L − RLR WB − LCY 2 2 LTW L − RLF WB + LCY 2 2 LTW L − RRF WB + LCY 2 2 LTW L − RRR WB − LCY 2 2 We will assume that the gearing is low enough that the drive train is traction limited. substitute the friction relationships into the wheel moment equations: . Now. Therefore the forces at each wheel can be written as follows: FLR = µ x N R FLF = µ x N F FRF = µ x N F FRR = µ x N R RLR = µ y N R RLF = µ y N F RRF = µ y N F RRR = µ y N R In the above equations.The first step will be to determine the moments about the center of mass. Here.
. the moment equation can be rewritten as follows: MT = µ x LTW 2 (2 N R + 2 N F ) − µ y 2 N R LWB − LCY + 2 N F LWB + LCY 2 2 The first term can now be simplified using the distributive property to remove the factor of 2. The following picture is a side view of the robot. and the location of the center of mass. Notice that the force depicted on the front is 2 NF. Likewise for the rear. a relationship needs to be determined between the normal forces NR and NF. the above equation can be rearranged to get equation 1: M T = µ x LTW ( N R + N F ) − µ y [LWB ( N R + N F ) − 2 LCY ( N R − N F )] eqn(1) At this point. sum the moments to find the resultant turning moment: LTW L L L + µ x N F TW + µ x N F TW + µ x N R TW 2 2 2 2 L L L L − µ y N R WB − LCY + µ y N F WB + LCY + µ y N F WB + LCY + µ y N R WB − LCY 2 2 2 2 MT = µx NR Collecting like terms. This is the case since there are 2 front wheels – not one.M LR = µ x N R M LF = µ x N F M RF = µ x N F M RR = µ x N R LTW L − µ y N R WB − LCY 2 2 LTW L − µ y N F WB + LCY 2 2 LTW L − µ y N F WB + LCY 2 2 LTW L − µ y N R WB − LCY 2 2 Now that the moment contribution of each wheel is known. The term in the brackets can also be simplified by using the distributive property. M T = µ x LTW (N R + N F ) − µ y [N R LWB − 2 N R LCY + N F LWB + 2 N F LCY ] Using the distributive property again on the term in the brackets. the robot weight W.
take the sum of the forces in the z direction: ∑F z = 0: 2N R + 2N F −W = 0 Rearrange the above equation to get equation 2: NR + NF = W 2 eqn (2) Now take the sum of the moments about the center of area: ∑M 2NR COA = 0: LWB L − 2 N F WB − WLCY = 0 2 2 LWB ( N R − N F ) = WLCY Rearrange the above equation to get equation 3: NR − NF = W LCY LWB eqn (3) Recall equation 1: M T = µ x LTW ( N R + N F ) − µ y [LWB ( N R + N F ) − 2 LCY ( N R − N F )] eqn(1) .Front LCY COA W 2NF LWB 2NR Using a static analysis again.
Assume the worstcase COM: the COM is at the COA (i.You can move the COM toward the front or rear of the robot. the term in the parentheses will move toward a negative number.e. .Now. As LCY gets larger. the final turning moment must be greater than zero.. i. Equation 4 then simplifies as follows: µ x LTW > µ y LWB eqn (5) . substitute equations 2 and 3 into equation 1 to get the following: M T = µ x LTW L W W − µ y LWB − 2 LCY W CY 2 2 LWB Rearrange the above equation to get the following: MT = µ x LTW W 2 L L2 − µ yW WB − 2 CY LWB 2 Lastly. rearrange further to get the final turning moment equation: ****** MT = W 2 L2 µ x LTW − µ y LWB − 4 CY LWB ****** Making the Robot Turn In order to make the robot turn.. Therefore the following relationship must be true: µ x LTW > µ y LWB − 4 L2 CY LWB eqn (4) . LCY = 0). the relationship will be satisfied. A few ways to satisfy the relationship are given below.e. . the right side of the inequality becomes smaller.must be true for robot to turn!!!! The relationship in equation 4 can be satisfied a number of ways. Thus. allowing the left side to be greater than the right side.
we need to determine the normal force at each wheel. For instance. the wheel bites into the carpet providing a high driving friction. Then. the driving force generated by the wheels must be enough to overcome the frictional resistance force. Solve for the normal forces (left as an exercise for the reader): NR = NF = W W LCY + 4 2 LWB W W LCY − 4 2 LWB This shows that if Lcy is positive. Recall equations 2 and 3: W 2 NR + NF = eqn (2) NR − NF = W LCY LWB eqn (3) Notice that this is simply a system of two equations with two unknowns. but sharp in the direction of travel. This should make sense since if Lcy is positive. Proper Gearing To ensure that the robot will turn. In order to determine the frictional resistance at each wheel. the moment about the COM at EACH wheel should be greater than zero. the robot will turn. The friction coefficients in equation 5 then cancel. the normal force will be greater at the rear of the robot. we should select our gears such that the driving force at EACH wheel is enough to overcome the frictional resistance force at that wheel. the center of mass is toward the rear of the robot. This way. . To simplify things (and give some safety margin). the following relationship must be satisfied: LTW > LWB Therefore. . which means that there should be more normal force on the rear wheels. by making the track width larger than your wheelbase. Therefore.Now assume that you are using rubber wheels such that the coefficient of friction is the same in all directions. You can also use wheels that have a tread pattern that is rounded on the sides of the wheel.Another method of satisfying equation 5 is to use a wheel with friction characteristics that are different in the x and y directions. but slides easily sideways. you can use holonomic wheels.
you can determine the gear ratio that will satisfy this minimum force. Of course. the moment about the COM for the leftrear wheel must be greater than zero: FLR LTW L − µ y N R WB − LCY > 0 2 2 Substitute in the equation for the rear normal force: FLR W W LCY LWB LTW − µy + 4 2 L 2 − LCY > 0 2 WB Rearrange and multiply out the two terms in parentheses: FLR WL LTW WLCY WLCY WL2 CY > µ y WB − + − 8 2 4 4 2 LWB Simplifying: FLR LTW µ yW LWB L2 CY > 4 −L 2 2 WB Lastly. the analysis will be performed for a rear wheel (Lcy is assumed to be positive for this example). you should use a gear ratio that provides considerably more than this minimum force. . solve for the force: * * * * * * FLR > µ yW LWB LTW L2 − CY * * * * * * This relationship must be satisfied in order to turn 4 LWB Therefore. you can use the above equation to determine the minimum force necessary at the wheels to make the robot turn. since this force is just barely enough to make the robot start to turn.Therefore. along with the motor characteristics and the diameter of your wheels. Using this force. to be sure that the worst case is used. To ensure turning.
be sure that the following relationship is true: µ x LTW > µ y LWB − 4 L2 CY LWB 2) If you don’t like math. use holonomic wheels or choose a good wheel tread pattern). Plug these numbers into the following equation to determine the force necessary at each wheel to make the robot turn: FLR > µ yW LWB LTW L2 − CY 4 LWB .Extensions Tank Treads The fundamental difference between tank treads and 4wheeldrive is that the force on the treads is distributed over the length of the treads. Other than this. LWB is 2/3 of the length of the contact patch of the treads.Make the track width greater than the wheel base (LTW > LWB) . you should have values for the important dimensions (LTW. However. The trackwidth (LTW) is still the distance between the two treads. LWB. The good news is that the 4wheeldrive equations can still be used. 3) Once you satisfy the equation in step 1). To do this.e.Try to move the center of mass slightly away from the center of the robot.. The details of the math will be skipped for this.If possible. In order to take the effects fully into account. begin by being sure that the robot will turn. Instead. use the following rules of thumb: . jumping straight to the conclusions instead. Use caution to not move the COM far enough so that the robot becomes unstable. rather than at four discrete points. and LCY). Conclusions 1) When designing a drive train. the equations can still be used. . reduce the lateral friction coefficient while keeping the longitudinal friction high (i. the forces and moments need to be integrated over the length of the tread. the wheelbase (LWB) is not simply the length of the contact patch of the tread.
the problem can become easy enough to solve. However. When you are creating your design. just to be safe. if you determine in step 3 that you need to produce 50 lb of force at each wheel. You should also keep in mind that your drive train and gearing will not be 100% efficient. . by carefully making a few simplifying assumptions. you might consider that your gearing should produce 100 lb of force at the wheels. By using the above equations and rules of thumb. a minimum gearing can be determined such that the motors can produce the necessary force at the wheel.4) Using the force calculated in step 3). For instance. Therefore. I would suggest a safety factor of at least 2 for the gearing. Final Conclusion The physics used to develop these equations and relationships are not exact. some safety margin should be included into the calculations. Keep in mind that these equations are generated such that the robot will just barely begin to turn. be sure to take efficiency into account. and useful results can be achieved. you should be able to be successful in designing a fourwheeldrive system. when determining your gearing.