What went before

• Last time we looked at two parts of the mathematics of robot

motion.
• Overall motion, making simplifying assumptions
– Robot as a body moving in a plane.
KINEMATICS II
– Forward and reverse kinematic models.
• Motion of individual wheels.
– Equations relating to wheel spin and slideways motion.
– Fixed and steerable standard wheels have constraints.
– Swedish and spherical wheels, as well as castors, do not have
constraints.

cis32.5-fall2006-parsons-lect05 2

Kinematic constraints
Today
• How does the design of a robot with M wheels constrain how the
robot moves?
• Take individual constraints on wheels, and use these to establish
constraints on the robot as a whole. – How does a differential drive robot move compared with a
• Gives us a more accurate kinematic picture of the whole robot. bicycle?
• Five categories of wheel:
– Robot as a constrained body moving in a plane.
• Tells us how the design of the robot constrains its ability to move. – Fixed standard
– Steerable standard
• Gives us precise notions of:
– Castors
– Mobility – Swedish
– Steerability – Spherical
– Maneuverability
• Only fixed and steerable standard wheels have any constraints.

cis32.5-fall2006-parsons-lect05 3 cis32.5-fall2006-parsons-lect05 4
 • A fixed standard wheel, radius r, polar coordinates l and α to
some reference point on the chassis, wheel at β to chassis, chassis
moving at ξ˙I has a rolling constraint:
• Similarly, we have a sliding constraint:
[sin(α + β) − cos(α + β)(−l) cos β] R(θ)ξ̇I − rϕ̇ = 0
[cos(α + β) sin(α + β)l sin β] R(θ)ξ̇I = 0
where ϕ̇ is the rate of rotation of the wheel about its axle and:

cos θ sin θ 0 

• This says that the wheel does not move perpendicular to its
plane of rotation.


R(θ) = − sin θ cos θ 0 
 



0 0 1 • We can write identical expressions for a steerable standard
 

translates from the global frame of reference to the local frame. wheel.
• This says that the wheel does not slip at its point of contact with
the ground.

cis32.5-fall2006-parsons-lect05 5 cis32.5-fall2006-parsons-lect05 6

From constraints on wheels to constraints on robots

• ϕf (t) is a vector that describes the rotation of the fixed wheels.
• ϕs(t) is a vector that describes the rotation of the steerable wheels.
• Consider we have a robot with N wheels.
• ϕ(t) collects these together:
– Nf are fixed standard wheels  

– Ns are steerable standard wheels ϕf (t) 

ϕ(t) = 


ϕs(t)


• Deal with this by considering fixed wheels all together, and

steerable wheels all together. • We can now write the rolling constraints of all the wheels in a
way analogous to the constraint for one wheel:
• βs(t) is a vector that describes the steering angles of all Ns
steerable wheels. J1(βs)R(θ)ξ̇I − J2ϕ̇ = 0
• βf is a vector that describes the orientation of all Nf fixed wheels.

cis32.5-fall2006-parsons-lect05 7 cis32.5-fall2006-parsons-lect05 8
 where J2 gives wheel radii, and J1 relates wheels to the motion
• We can do the same thing for sliding constraints, giving:
along their planes:
 
C1(βs)R(θ)ξ˙I = 0
 r1 0 . . . 0 
where

0 r2 . . . 0 
 


J2 = .. .. .. .. 
  


C1f 
C1(βs) = 
 


C1s(βs)
  
0 0
 

. . . rNf +Ns 
• These two expressions then summarize all the constraints on the
 
J1f 
J1(βs) =


root due to its wheels.


J1s(βs)
 

• J1f and J1s(βs) are relate wheels to motion for fixed and steerable • The sliding constraint (the second one) has the biggest impact on
wheels respectively. what a robot can do.

cis32.5-fall2006-parsons-lect05 9 cis32.5-fall2006-parsons-lect05 10

Maneuverability of a mobile robot • Sliding constraints require that:

• The mobility of a chassis is its ability to directly move in the C1f R(θ)ξ˙I = 0
environment.
and
• An unconstrained robot can move in any direction at any time, C1s(βs)R(θ)ξ˙I = 0
and therefore follow any path through the environment. • These put contraints on R(θ)ξ˙I , and thus on the motion that the
• There are two parts to mobility: robot can perform.
– Instantaneous motion • One view:
– Steering R(θ)ξ˙I must belong to the null space of C1(βs)
which is (roughly speaking) the difference between omni-drive meaning that motion is in some space N such that for any n ∈ N
and a car chassis.
C1(βs)n = 0
• The overall maneuverability of a chassis is a combination of
freedom from sliding constraints and steerability. • Another view: instantaneous center of rotation.

cis32.5-fall2006-parsons-lect05 11 cis32.5-fall2006-parsons-lect05 12
 Back to maneuverability

Instantaneous center of rotation • The range of possible motion is determined by the set of
independent constraints.
• Related to the rank of C1(βs).
• More constraints = greater rank of C1(βs)
• More constraints = less flexibility in mobility of the robot.

cis32.5-fall2006-parsons-lect05 13 cis32.5-fall2006-parsons-lect05 14

• A robot with one fixed standard wheel.

– One constraint. • Range of possible values for rank [C1(βs)]:
– C1(βs) has rank one.
0 ≤ C1(βs) ≤ 3
• A robot with two fixed standard wheels in differential drive:
• C1(βs) = 0. No constraints, and no standard wheels.
– Two constraints, but not independent:
– C1(βs) has rank one. • C1(βs) = 3. Fully constrained since only three dimensions to
constrain.
• A robot with two fixed standard wheels in bicycle drive:
• C1(βs) = 3. No motion possible.
– Two constraints.
– C1(βs) has rank two

cis32.5-fall2006-parsons-lect05 15 cis32.5-fall2006-parsons-lect05 16
 Degree of mobility • Differential drive robot:
– δm = 2
– Robot can change both orientation and position on its current
• The null space N of C1(βs) defines how the robot can move just
path just by changing wheel speed.
by changing wheel velocity.
• Bicycle drive robot:
• The dimensionality of N measures the degrees of freedom under
the robot’s control just by altering velocity. – δm = 1
• Define degree of mobility δm: – Robot can only change position on its current path by
changing wheel speed.
δm = dim N [C1(βs)] – Needs steerable wheel to change orientation.
= 3 − rank [C1(βs)]

cis32.5-fall2006-parsons-lect05 17 cis32.5-fall2006-parsons-lect05 18

Degree of steerability

• Degree of mobility captures what can be done just with changes

• 0 ≤ δs ≤ 2
in wheel velocity.
• What about steerable wheels? • δs = 0 implies no steerable wheels.

• They don’t have an instantaneous effect, but they do have an • δs = 1 implies one independent steerable wheel.
effect over time. – As in a car where two steerable wheels share one axle
• Degree of steerability δs • δs = 2 only possible if no standard wheels
δs = rank [C1s(βs)] – “Two-steer”
C1s(βs) is the “steerable” bit of C1(βs).
• The bigger the rank of C1s(βs), the more steerable the robot.

cis32.5-fall2006-parsons-lect05 19 cis32.5-fall2006-parsons-lect05 20
 Degree of manueverability
On to workspace
• The degree of maneuverability depends on both mobility and
steerability:
• What is important is how the robot can move in its environment.
δM = δm + δs • Degrees of freedom (DOF)
• Maneuverability includes both the degrees of freedom that can – Feature of the workspace
be manipulated instantaneously through changes in wheel • Differentiable degrees of freedom
velocity, and those that can be manipulated through steering.
– Feature of the robot
• As a result, robots with the same δM are notnecessarily
equivalent. – Number of independently achievable velocities
– Equal to degree of mobility δm.
• We can see that by looking at different wheel configurations.

cis32.5-fall2006-parsons-lect05 21 cis32.5-fall2006-parsons-lect05 22

Summary

• These notes have discussed robot kinematics.

• In particular, they have considered how we can develop
constraints on a whole robot from constraints on a set of wheels.
• Furthermore, they have discussed how this set of constraints on
a robot can be used to develop notions of mobility, steerability,
and maneuverability.

cis32.5-fall2006-parsons-lect05 23