Trajectory Planning

UNIT 7 TRAJECTORY PLANNING

Structure
7.1 Introduction
Objectives
7.2 Joint Space Scheme
7.3 Cartesian Space Scheme
7.4 Point-to-Point Motion
7.5 Continuous Path Motion
7.6 Summary
7.7 Key Words
7.8 Answers to SAQs

7.1 INTRODUCTION
The goal of trajectory planning is to generate a function according to which a robot will
move. This function generation depends on the robot tasks, for example, to pick-up an
object from one point and to place it in another point, or to weld two metal pieces along a
curve. In the former case, initial and final points are assigned, i.e., it is a
“point-to-point” motion, whereas in the latter situation a finite sequence of points need to
be specified, i.e., a “Continuous-path” motion. A robot user typically specifies a number
of parameters to describe a point-to-point or continuous-path task. A trajectory planning
algorithm then computes the inputs for the motion controller of the robot. Trajectory
planning can be done either in the joint space, i.e., in terms of joint positions, velocities
and accelerations, or in Cartesion space (also called operational space), i.e., in terms of
the end-effector’s positions and orientations, and their velocities and accelerations.
Usually, the latter is preferred since it allows a natural description of the task the robot
has to perform. However, the control action of the robot is carried in the joint space.
Thus, a suitable inverse kinematics algorithm is to be used to reconstruct the time
sequence of the joint variables corresponding to the above sequence of Cartesion
variables.
Objectives
After studying this unit, you should be able to
• understand the meaning, and necessity of trajectory planning,
• types of trajectory generation schemes,
• how to generate a trajectory,
• choose a suitable trajectory, and
• calculate the path points.

7.2 JOINT SPACE SCHEME

In this section, methods of determine a desired trajectory of joint variables over time are
considered. In the joint space scheme, it is assumed that the initial and final joint
positions corresponding to initial and final configurations of the end-effector are known
from the solution of a position inverse kinematics problem explained in Unit 5. Then, the
basic problem is how to select a trajectory between the given initial and final joint
positions within the allowable time interval to move between them. Various methods of
solving the problem have been proposed in the literature. A simple method is based on
the polynomial function of time. Suppose, for a joint, initial angle is θ0 at time t = 0, and
the final angle is θf at time t = tf, i.e.
5
Trajectory Planning θ ( 0 ) = θ o and θ ( t f ) = θ f
. . . (7.1)
and Control of Robots
Additionally, to start and stop the end-effector according to some desired velocity and
acceleration, the following conditions must be satisfied :
θ&( 0 ) = θ&o ; θ&( t f ) = θ&f . . . (7.2)
&
θ&( 0 ) = &
θ&o ; &
θ&( t f ) = &
θ&f . . . (7.3)
Although many smooth functions can satisfy the constraints given by Eqs. (7.1)-(7.3), a
polynomial of time t is chosen here due to their ease of computation and simplicity of
expression. The lowest order that can satisfy all the six conditions, Eqs. (7.1)-(7.3), is
five, namely,
θ ( t ) = a o + a1 t + a 2 t 2 + a 3 t 3 + a 4 t 4 + a 5 t 5 . . . (7.4)
Differentiating Eq. (7.4) with respect to time yields
θ&( t ) = a1 + 2 a 2 t + 3 a 3 t 2 + 4 a 4 t 3 + 5 a 5 t 4 . . . (7.5)
&
θ&( t ) = 2 a 2 + 6 a 3 t + 12 a 4 t 2 + 20 a 5 t 3 . . . (7.6)
Substituting the conditions of Eqs. (7.1)-(7.3) into Eqs. (7.4)-(7.6), six equations with six
unknowns, a0 . . . a5, are obtained as
θ (0) = θ o = a o

θ ( t f ) = θ f = a o + a1t f + a 2 t 2f + a 3 t 3f + a 4 t 4f + a 5 t 5f

θ&(0 ) = θ&0 = a1

θ&( t f ) = a1 + 2 a 2 t f + 3 a 3 t 2f + 4 a 4 t 3f + 5 a 5 t 4f

&
θ&(0 ) = &
θ&0 = 2 a 2

&
θ&( t f ) = 2 a 2 + 6 a 3 t f + 12 a 4 t 2f + 20 a 5 t 3f . . . (7.7)

Solving Eq. (7.7), the unknown coefficients, a 0 − a 5 , that satisfy Eqs. (7.1)-(7.3) are
given by
1 &
a o = θ o ; a1 = θ&o ; a 2 = &θo
2

a3 =
1
2 t 3f
[20 ( θ f − θ o ) − ( 8 θ&f + 12 θ&o ) t f − ( 3 &
θ&o − &
θ&f ) t 2f ]
a4 =
1
2 t 4f
[30 ( θ o − θ f ) + (14 θ&f + 16 θ&o ) t f + ( 3 &
θ&o − 2 &
θ&f ) t 2f ]
a5 =
1
2 t 5f
[12 ( θ f − θ o ) − 6 ( θ&f + θ&o ) t f − ( &
θ&o − &
θ&f ) t 2f ] . . . (7.8)

Example 7.1
Fourth Order Polynomial
Referring to the end point accelerations, if & θ&o = &θ&f = 0 , and the relation
shown in Figure 7.1 is satisfied by θ o , θ f , θ&o , and θ&f that is, if
tf
θ f − θo = ( θ&o + θ&f ) . . . (7.9)
2
then a5 = 0, which implies that only a fourth-order polynomial is required
for θ (t).
Using combinations of fourth-order polynomials and straight lines,
trajectories for various cases can be determined fairly easily. Let us consider
two such cases.
6
 θ Trajectory Planning

θf
θ&0 θ&t
tf
( θ&0 + θ&f )
2
θ0

o tf t
tf / 2
Figure 7.1 : Boundary Conditions &
θ&o = &
θ&f = 0

(i) Initial and Final Positions are Known

A trajectory is developed that starts from rest at θo, passes through the
phases of acceleration, constant velocity and deceleration, and finally comes
to a completes stop at θf. This is depicted in Figure 7.2. For this purpose,
first the value of parameter Δ that denotes one-half of the acceleration and
deceleration period is chosen. Next, the auxiliary points θ02 and θf 1, as
shown in Figure 7.2, are determined by the following procedure : First two
points θ01 and θf 2 at time t = Δ and t = t f − Δ are taken as θ01 = θ0 and
θ f 2 = θ2 . Then θ01 and θ f 2 are connected by a straight line, and θ02 and
θ f 1 are determined as the points on the straight line at time t = 2 Δ and
t = t f − 2Δ . This is followed by the determination of the trajectory segments
between θ0 and θ02 , and between θ f 1 and θ f , using the fourth-order
polynomial obtained in Example 7.1 such that their velocities at boundary
points θ0 , θ02 , θ f 1 and θ f coincide with the composition of the straight
lines connecting θ0 , θ01 , θ f 2 and θ f , and the accelerations at these
boundary points are zero. The trajectory segment between θ02 and θ f 1 is
specified by the straight line. Thus, the whole trajectory is given by the bold
solid line in Figure 7.2.
θ

θf2 θf
θf1

θ02
θ01
θ0

O
t
Δ Δ Δ Δ

Figure 7.2 : Trajectory with Acceleration, Constant Velocity and Deceleration Phases

Example 7.2 7
Trajectory Planning Initial and Final Points are Known
and Control of Robots
Applying the above method to the two-link planar manipulator shown in
Figure 7.3, a trajectory that starts from rest at the initial position,
Figure 7.3(a), and comes to a complete stop at the final position,
Figure 7.3(b), in 2 sec is generated. Suppose Δ = 0.25 sec. Then, the first
joint angle, θ1, is obtained as follows : First, choose θ01 = 90o at time,
t = 0.25 sec, and θf2 = 60o at t = 1.75 sec. Straight line equation is then
obtained as
570
θ = − 20 t +
6
Hence, θ02 at t = 0.5 sec, and θf1 at t = 1.5 are found as
θ02 = 85o ; and θ f 1 = 65o

Now, for the 4th order polynomial between θ0 and θ02, the coefficients are
obtained as
a0 = 90; a1 = a2 = 0; a3 = − 80; and a4 = 80

Y Y
o
60

a2 = l
o
120

a1 = l

o
90
o
60
O O

X X

(a) Initial Position (b) Final Position

Figure 7.3 : Two-link Manipulator (a) Initial Position, (b) Final Position

tf
Note that, θ&
& = θ&
0
& = 0 , and θ − θ =
f f 0 (θ&f + θ&0 ) conditions are used to find
2
the coefficient values. For the interval, 0 < t ≤ 0.5, the joint trajectory is then
obtained as follows :

θ1 = 90 − 80 t 3 + 80 t 4

For the straight line between θ02 and θf1, i.e., 0.5 < t ≤ 1.5, it can be easily obtained
as
θ1 = 85 − 20 (t − 0.5)

Finally, the 4th order polynomial between θf2 and θf, i.e., 1.5 < t ≤ 2.0, can be
written as

θ1 = a0 + a1 (t − 1.5) + a2 (t − 1.5) 2 + a3 (t − 1.5)3 + a4 (t − 1.5) 4

whose coefficients are calculated as

a0 = 65; a1 = − 20; a3 = 80; and a4 = − 80

Hence, the final trajectory for θ1 is written as

8
 ⎧ 90 − 80 t 3 + 80 t 4 , 0 ≤ t ≤ 0.5
Trajectory Planning
⎪
θ1 (t ) = ⎨ 85 − 40 (t − 0.5), 0.5 < t ≤ 1.5
⎪ 3 4
⎩65 − 20(t − 1.5) + 80 (t − 1.5) − 80 (t − 1.5) , 1.5 < t ≤ 2.0
Similarly, the second joint angle, θ2, is given by
⎧ − 60 − 160 t 3 + 160 t 4 , 0 ≤ t ≤ 0.5
⎪
θ2 (t ) = ⎨ − 70 − 40 (t − 0.5), 0.5 < t ≤ 1.5
⎪ 3 4
⎩ − 110 − 40 (t − 1.5) + 160 (t − 1.5) − 160 (t − 1.5) , 1.5 < t ≤ 2.0
Figure 7.4 shows the resulting endpoint trajectory.
Y
t=0

t = 0.5

t = 1.0

o o
5
o 10 10
o
5

t = 1.5

O
t=2 X
Figure 7.4 : Endpoint Trajectory
(ii) Several Positions are Known
A trajectory that starts from rest at initial position θ0 passes through
intermediate positions θ1 and θ2, and stops at final position θf is generated.
First, it is considered that the trajectory need not pass exactly through θ1 and
θ2. In this case, auxiliary points are specified, namely,
θ02, θ11, θ12, . . . , θf1, as in Figure 7.5. Note that the values at point θ0 and θ01
are the same, and the values at θf2 and θf are same. Moreover, θ02 and θ11 are
on the straight line between θ01 and θ1. Points θ12, . . . , θf1 are also
determined in the same way. Then, the trajectory segments for
(θ0 , θ02 ), (θ11 , θ12 ), . . . , (θ f 1 , θ f ) are determined using fourth-order
polynomials, and those for (θ02 , θ11 ), (θ12 , θ21 ) and (θ22 , θ f 1 ) using
straight lines. The resulting trajectory is shown in Figure 7.5 by solid line.
Next, consider the case when the trajectory is required to pass exactly
through θ1 and θ2. This case can be dealt with by increasing the number of
auxiliary points. A scheme is shown in Figure 7.6(a). Auxiliary points are
determined as follows : As shown in Figure 7.6(b), θi2 (i = 1, 2) is the
midpoint between A (the intersection of the line, t = ti – Δ, with the straight
line between θi-1 and θI) and B (the intersection of the line, t = ti – Δ with the
straight line between θi and θi+1). Point θi3 is determined similarly. The
trajectory segments for
(θ0 , θ02 ), (θ11 , θ1 ), (θ1 , θ14 ), . . . , (θ f 1 , θ f )

are then determined by fourth-order polynomials, and those for

9
Trajectory Planning (θ02 , θ11 ), (θ14 , θ21 ) and (θ24 , θ f 1 )
and Control of Robots
by straight lines, as in the previous case.
θ

θf2
θf
θf1
θ1
θ12
θ11
θ22
θ21
θ2
θ02

θ01
θ0

t
Δ Δ Δ Δ Δ Δ Δ Δ

Figure 7.5 : Trajectory Passing near the Intermediate Points

θ

θ12 θ13
θ1 θf2
θf
θf1
θ11 θ14
θ21 θ24
θ22 θ23
θ2
θ02
θ01
θ0

O t
Δ
t1 t2 tf
(a)

B
θi
θi4
θi3
θi2
θil
A θi+l

θi−l
Δ
ti
(b)
Figure 7.6 : Trajectory Passing through the Intermediate Points
(a) Intermediate Points, (b) Determination of Points θi2 and θi3

In the basic problem of connecting θ0 and θf by a polynomial, which was

mentioned at the beginning of this section, if we consider the boundary
conditions on position and velocity only, i.e. Eqs. (7.1-7.2), and do not take
into account the continuity of acceleration, one can then simplify the

10
 polynomials. More specifically, a third-order polynomial of the following Trajectory Planning
form
θ (t ) = a0 + a1 t + a2 t 2 + a3 t 3 . . . (7.10)
can be used in such cases instead of a fifth-order one. The above constants
are determined as follows :
a = θ ; a = θ&
0 0 1 0

1
a2 = [3 (θ f − θ0 ) − (2θ&0 + θ&f ) t f ]
t 2f
1
a3 = [− 2 (θ f − θ0 ) + (θ&f + θ&0 ) t f ] . . . (7.11)
t 3f

In particular, if θ0 , θ f , θ&0 , and θ&f satisfy Eq. (7.9), then a3 = 0, and so θ (t)
becomes a second-order polynomial. Hence, for cases (i) and (ii) treated
above, one can also use second-order polynomials.
SAQ 1
(a) What is the order of a trajectory that has to satisfy position, velocity and
acceleration constraints at the initial and final points?
(b) What is the advantage of joint space scheme?

7.3 CARTESION SPACE SCHEME

When a trajectory between two configurations of the robot is generated by a joint
variable scheme, as described in the preceding subsection, it is sometimes difficult to
predict the motion of the end-effector in the Cartesion space. There are also cases in
which some specific end-effector trajectory is required. For example, in arc welding, the
electrode should follow a seam precisely. In such cases, a trajectory in terms of the
position and orientation variables of the end-effector is to be generated. To describe the
position of the end-effector, three Cartesian coordinates are required which can be
determined similar Eq. (7.4) or Eq. (7.10).
Example 7.3
Cartesian Position Trajectory
A trajectory for the planar problem in Example 7.2 is obtained here in terms
of the end-effector position variables, namely, r ≡ [ x, y ]T . Since from
T
⎡ 3 3⎤
Figure 7.3 the initial position r0 is, r0 ≡ ⎢ , ⎥ , and the final position rf,
⎣ 2 2⎦
is r f ≡ [1, 0]T , the trajectory r (t ) ≡ [ x (t ), y (t )]T is obtained using the
philosophy of Example 7.2 as
⎧ 3 4 (2 − 3) (t 3 − t 4 )
⎪ + 0 ≤ t ≤ 0.5
⎪ 2 3
⎪⎪ (5 3 + 2) (2 − 3) (t − 1.5)
x (t ) = ⎨ + 0.5 < t ≤ 1.5
⎪ 12 3
⎪ (10 + 3) (2 − 3) (t − 1.5) 4 (2 − 3) [(t − 1.5)3 − (t − 1.5) 4 ]
⎪ + + 1.5 < t ≤ 2
⎪⎩ 12 3 3

11
Trajectory Planning ⎧3 3 4
and Control of Robots ⎪ 2 − 4t + 4t 0 ≤ t ≤ 0.5
⎪
⎪5
y (t ) = ⎨ − (t − 0.5) 0.5 < t ≤ 1.5
⎪4
⎪1 3 4
⎪ 4 − (t − 1.5) + 4 (t − 1.5) − 4 (t − 1.5) 1.5 < t ≤ 2
⎩

Figure 7.7 shows the resulting endpoint trajectory.

t=0

t = 0.5

t = 1.0

t = 1.5

O
t = 2.0 X

Figure 7.7 : Endpoint Trajectory

As for the effector orientation of the end-effector one may have several
choices to represent it. For example, Euler angles of the end-effector can be
used which, however, is difficult to visualise because three angles are the
three rotations of the end-effector about three time-varying axes. Hence, an
alternative method based on the rotation about a fixed axis is presented.
Other methods like double axis rotation, etc. are also available in the
literature.
Single-axis Rotation Method
In this method, an orientation representation between two frames, FA and FB, B

having the same origin is given by the pair, [e, α], where e ≡ [e x , e y , e z ]T is
the equivalent axis of rotation and α is the equivalent angle of rotation. This
pair is used to obtain FB from FA by a single rotation about the unit vector e
B

by an angle α. Note that the rotation matrix, Q (e, α), is given by

⎡ ex2 Vα + Cα ex e y Vα − ez Sα ex ez Vα + e y Sα ⎤
⎢ ⎥
Q (e , α) = ⎢ ex e y Vα + ez Sα e 2y Vα + Cα e y ez Vα − ex Sα ⎥ . . . (7.12)
⎢ ⎥
⎢ ex ez Vα − e y Sα e y ez Vα + ex Sα ez2 Vα + Cα ⎥
⎣ ⎦

where Sα ≡ sin α, Cα ≡ cos α, and Vα ≡ 1 − Cα. The correctness of Eq. (7.12)

can be checked, e.g., by rotating frame FA about Z-axis by 90o. The rotated
frame say, FB, will then have the X-axis parallel to the Y-axis of FA. As a
B

result, the first column of matrix Q (e, α) is [0 1 0]T which is true from Eq.
(7.12) for e = [0 0 1]T and α = 90o. Similarly, other columns can be verified
as [– 1 0 0]T and [0 0 1]T.

12
 Now, given two orientations of a frame, say, FA and FB, the latter can be
B
Trajectory Planning
obtained by rotating the former by an angle α about e, i.e., α is to be varied
from 0 to αf about the fixed e. Hence, the boundary conditions are given by
α (0) = 0, α (t f ) = α f

α&(0) = α&0 , α (t f ) = α&f

α
&(0) = α
& &0 , α (t f ) = α
& &
&f

for 0 ≤ t ≤ tf. The trajectory for α is then obtained similar to θ in Section 7.2.
Example 7.4
Single-axis Rotation
Suppose that a trajectory that transfers the end-effector orientation from rest
at the initial orientation shown in Figure 7.8(a) to the final rest orientation
shown in Figure 7.8(b) in time, tf = 1, is to generated. The problem is solved
using the single-axis rotation method. It is easily seen from Figure 7.8 that
the rotation matrix Q representing the final orientation with respect to the
initial one is

⎡0 0 1 ⎤
Q = ⎢⎢1 0 0 ⎥⎥
⎢⎣ 0 1 0 ⎥⎦

Comparing this matrix with Eq. (7.12), i.e., the summation of the diagonal
terms gives, αf = 120o, while the element wise comparisons lead to
T
⎡ 1 1 1 ⎤
e≡⎢ , , ⎥
⎣ 3 3 3⎦
Z0 YF

Y0

XF

X0
ZF

(a) (b)

Figure 7.8 : Initial and Final Orientations of the End-effector

(a) Initial Orientation, (b) Final Orientation

If a fifth-order polynomial, Eq. (7.4), is used for interpolation, the trajectory

for rotation angle, α, is given by

α (t ) = 1200 t 3 − 1800 t 4 + 720 t 5 0 ≤ t ≤1

SAQ 2
(a) What are the difficulties of Cartesian space scheme?
(b) When one uses Cartesian space scheme?

13
Trajectory Planning
and Control of Robots 7.4 POINT-TO-POINT MOTION
During the point-to-point motion, a manipulator has to move from an initial configuration
to a final one in a given time tf, be it a joint or Cartesian motion space. In such cases, the
actual path is of no concern. The algorithm generates a trajectory which, in respect to the
above general requirement, is also capable to optimize some performance index. A
suggestion for choosing the index is the energy dissipation in the motors. Let I be the
moment of a rigid body about its rotation axis and it is taken from an initial value θ0 to a
final value θf in a time tf. Assuming that the rotation is executed through a torque τ
supplied by a motor, the trajectory generation problem based on the optimization is then
formalized as follows : Set θ&= ω, and determine the solution of the differential equation
given by
Iω
&= τ

subject to the condition

tf

∫ ω (t ) dt = θ f − θ0
0

so as to minimize the performance index

tf

∫ τ2 (t ) dt
0

It can be shown that the resulting solution is of the type

ω (t ) = at 2 + bt + c

Even though the joint dynamics cannot be described in the above simple manner, the
above polynomial function of joint speed, ω(t), to generate a joint trajectory represents a
valid solution for the problem at hand. Therefore, to determine a joint motion, θ(t), the
cubic polynomial must be chosen, i.e.,

θ (t ) = a3 t 3 + a2 t 2 + a1 t + a0 . . . (7.13)

resulting into a parabolic velocity profile

ω (t ) ≡ θ&(t ) = 3a3 t 2 + 2a2 t + a1

and a linear acceleration profile

&(t ) ≡ &
ω θ&(t ) = 6a3 t + 2a2

Since four coefficients are available, it is possible to impose, besides the initial and final
joint position values θ0 and θf, also the initial and final joint velocity values θ&0 and θ&f ,
which are usually set to zero. Determination of a specific trajectory is then given by the
solution to the following system of equations :
a0 = θ0

a1 = θ&0

a3 t 3f + a2 t 2f + a1 t f + a0 = θ f

3a3 t 2f + 2a2 t f + a1 = θ&f

14
that allows computing the coefficients of the polynomial in Eq. (7.13). If it is desired to Trajectory Planning
assign also the initial and final values of acceleration, six constraints have to be satisfied
and than a polynomial of at least fifth order is needed, as obtained in Section 7.1.
An alternative approach with time laws of blended polynomial type is frequently adopted
in industrial practices. It allows to verify whether the resulting velocities and
accelerations can be supported by the physical mechanical manipulator. In this case, a
trapezoidal velocity profile is assigned, which imposes a constant acceleration in the start
phase, a cruise velocity, and a constant deceleration in the arrival phase. The resulting
trajectory is formed by a linear segment connected by two parabolic segments to the
initial and final positions. As can be seen from the velocity profiles in Figure 7.9, it is
assumed that both initial and final velocities are nil and the segments with constant
accelerations have the same time duration. This implies an equal magnitude of & θ&0 in the
two segments. Notice also that the above choice leads to a symmetric trajectory with
1 1
respect to the average point θ m = (θ f + θ i ) at t m = t f . The trajectory has to satisfy
2 2
some constraints to ensure the transition from θ0 and θf in time tf. The velocity at the end
of the parabolic segment must be equal to the (constant) velocity of the linear segment,
i.e.

& θ − θc
θ&c tc = m . . . (7.14)
t m − tc

where θc is the value attained by the joint variable at the end of the parabolic segment at
time tc with constant acceleration &θ&c . It is then

1&
θ c = θ0 + θ&c tc2 . . . (7.15)
2
Combining Eqs. (7.14)-(7.15) and substituting θm and tm in terms of θf, θ0 and tf yields
&
θ&c tc2 − &
θ&c t f tc + θ f − θ0 = 0 . . . (7.16)

Usually, &
θ&c is specified and then for given tf, θ0, and θf, the solution for tc is computed
from Eq. (7.16), for tc ≤ tf, as

1
tf t 2f &
θ&c − 4 (θ f − θ0 )
tc = − . . . (7.17)
2 2 &
θ&
c

Acceleration is then subject to the constraint

4 θ f − θ0
&
θ&c ≥ . . . (7.18)
t 2f

When the acceleration, & θ&c , is chosen so as to satisfy Eq. (7.18) with the equality sign, the
resulting trajectory does not feature the constant velocity segment any more and has only
the acceleration and deceleration segments (triangular profile). Given θ0 , θ f and t f ,
and thus also an average transition velocity, Eq. (7.18) allows imposing a value of
acceleration consistent with the trajectory. Then, tc is computed from Eq. (7.17), and the
following sequence of polynomials is generated
⎧ 1& & 2
⎪θ c + θc t 0 ≤ t ≤ tc
⎪⎪ 2
θ (t ) = ⎨ θc + &
θ&c tc (t − tc ) tc < t ≤ t f − tc . . . (7.19)
⎪
⎪θ f − 1 &θ&c (t f − t ) 2 t f − tc < t ≤ t f
⎪⎩ 2
15
Trajectory Planning θ
and Control of Robots
θf

θm

θc
θo

O tC tf −tC tf t

(a) Position
θ

θc

O tC tf −tC tf t

(b) Velocity

θc

tf −tC

O tC t

−θc

(c) Acceleration
Figure 7.9 : Characterisation of a Time Law with Trapezoidal Velocity Profile

Figure 7.10 illustrates a representation of the motion time law obtained by imposing the
data : θ0 = 0, θ f = π, and & θ&c = 6π . The adoption of a trapezoidal velocity profile
results in a worse performance indeed compared to the cubic polynomial. The decrease
tf
is, however, limited. The term, ∫ τ2 dt , increases by 12.5% with respect to the optimal
0
case.

2
[ Rad ]

0 0.2 0.4 0.6 0.8 1

[S]

(a) Position
16
 5 Trajectory Planning

[ Rad/s ] 2

0 0.2 0.4 0.6 0.8 1

[s]
(b) Velocity

20

10

0
[ Rad/s∧2 ]

− 10

− 20

0 0.2 0.4 0.6 0.8 1

[s]

(c) Acceleration
Figure 8.10 : Time History of Position, Velocity and Acceleration with
a Trapezoidal Velocity Profile Time Law

7.5 CONTINUOUS PATH MOTION

In several applications, the path in a joint or Cartesian motion space is described in terms
of a number of points greater than two. For instance, even for the simple point-to-point
motion of a pick-and-place task, it may be worth assigning two intermediate points
between the initial and the final points. Suitable positions can be set for lifting off and
setting down the object, so that reduced velocities are obtained with respect to direct
transfer of the object. For more complex applications, it may be convenient to assign a
sequence of points so as to guarantee better monitoring on the executed trajectories.
Therefore, the problem is to generate a trajectory when N points, termed path points, are
specified and have to be reached by the manipulator at certain instants of time. For each
variable there are N constraints, and then one might want to use an (N − 1)-order
polynomial. This choice, however, has the following disadvantages :
• it is not possible to assign the initial and final velocities,
• as the order of a polynomial increases, its oscillatory behavior increases, and
this may lead to trajectories which are not natural for the manipulator,
• numerical accuracy for computation of polynomial coefficients decreases as
the order increases,
• the resulting system of constraint equations is difficult to solve, and
17
Trajectory Planning • polynomial coefficients depend on all the assigned points; thus, if it is
and Control of Robots desired to change a point, all of them have to be recomputed.
These drawbacks can be overcome if a suitable number of low-order interpolating
polynomials, continuous at the path points, are considered in place of a single high-order
polynomial. Accordingly, the interpolating polynomial of lowest order is the cubic
polynomial, since it allows imposing continuity on velocities at the path points. With
reference to the single joint variable, a function θ (t) is sought, formed by a sequence of
N − 1 cubic polynomials θk (t), for k = 1, . . . , N − 1, continuous with first derivatives.
The function θ (t) attains the values θk for t = tk (k = 1, . . . , N), and θ1 = θi, for t1 = 0,
θN = θf, for tN = tf. The θk’s represent the path points describing the desired trajectory at
t = tk (Figure 7.11). The following situations can then be considered :
• Arbitrary values of θ&(t ) are imposed at the path points.

• The values of θ&(t ) at the path points are assigned according to a certain
criterion.
• The acceleration &
θ&(t ) shall be continuous at the path points.
Only the first case is shown here. Other two can be found in the literature.
θ

θ2

θN = θf
ρ2

θ3 ρ3 ρN-l
ρ1 ∫∫

θl = θi

∫∫
tl = 0 t2 t3 tN = t f t

Figure 7.11 : Characterisation of a Trajectory on a given Path

obtained through Interpolating Polynomials
Interpolating Polynomials with Velocity Constraints at Path Points
This solution requires the user to be able to specify the desired velocity at each
path point. Then, the system of equations allowing computation of the coefficient
of N − 1cubic polynomials interpolating the N path points is obtained by imposing
the following conditions on the generic polynomial ρk (t) interpolating θk and θk + 1,
for k = 1, . . . , N − 1 :
ρk (tk ) = θk ; ρk (tk + 1 ) = θk

ρ&k (tk ) = θ&k ; ρ&k (tk + 1 ) = θ&k + 1

The result is N − 1 systems of four equations in the four unknown coefficients for
each cubic polynomial, which can be solved independent of the other. The initial
and final velocities of the trajectory are typically set to zero (θ&1 = θ&N = 0) and
continuity of velocity at the path points is ensured by setting
ρ&k (tk ) = ρ&k + 1 (tk + 1 )

for k = 1, . . . , N − 2. Figure 7.12 illustrates the time history of position, velocity

and acceleration obtained with the following data :
π
θ1 = 0, θ2 = 2π, θ3 = , θ4 = π, t1 = 0, t2 = 2, t3 = 3, t4 = 5
2
θ&1 = 0, θ&2 = π, θ&3 = − π, and θ&4 = 0
18
 Notice the resulting discontinuity on the acceleration, since only continuity of Trajectory Planning
velocity is guaranteed.

4
[ Rad ]

0 2 4 6
[s]

(a) Position

5
[ Rad/s ]

−5

0 2 4 6
[s]

(b) Velocity

20
[ Rad/s ]
2

−20

−40

0 2 4 6
[s]

(c) Acceleration
Figure 7.12 : Time History of Position, Velocity and Acceleration with a Time Law of
Interpolating Polynomials with Velocity Constraints at Path Points

SAQ 3
(a) Write the applications for point-to-point and continuous path planning?
(b) How one can avoid the use of higher order polynomials while several points
are specified?

19
Trajectory Planning
and Control of Robots 7.6 SUMMARY
In this unit, several ways to plan a trajectory for the robot motion control are presented.
Joint and Cartesian space schemes are explained, as well as the point-to-point and
continuous path motion. The latter two motions appear in pick-and-place and welding
like operations, respectively.

7.7 KEY WORDS

Trajectory : A path to be followed by a robot.
Joint Space : Trajectories for the joints of a robot.
Cartesian Space : Trajectories for the end-effector coordinates.

7.8 ANSWERS TO SAQs

SAQ 1
(a) Five. See explanations in Section 7.2.
(b) It is simple and no problem of singularity arises.
SAQ 2
(a) One needs to perform inverse kinematics to obtain the joint variables needed
to control a real robot. Besides one may not be certain about the existence of
singularities for which no joint variable can be calculated.
(b) In situations like obstacle avoidances, etc., Cartesian space scheme is very
suitable.
SAQ 3
(a) Pick an place, and painting/welding, etc., respectively.
(b) See Section 7.5.

20
 Trajectory Planning
FURTHER READING
Sciavicco, L., and Siciliano, B. (2000), Modelling and Control of Robot Manipulator,
Springer-Verlag, London.
Yoshikawa, T. (1990), Foundation of Robotics, The MIT Press, Massachusetts.

21