What is system dynamics & controls competency?

Dynamic systems are all pervasive. Physical phenomena that can be

formulated using ordinary differential equations have a wonderful quality of
commonality between them. As a consequence, system dynamic competency
may not get restricted by a specific physical domain; be it mechanical,
electrical or even economics & finance! The respective domain knowledge is
necessarily required to derive the governing ODE of the physical phenomenon
but once it’s achieved, the discerning eye of an analyst should be able to
establish an equivalence between them.
All this is possible because of the essential form of an ODE which defines the
relationship between the rate of change of a physical parameter which is being
evaluated, the present state of that parameter & a time dependent inducing
signal. This inducing signal can be a combination of an actuating signal & a
disturbance signal.
Now this is where the control system competency steps in. The motivation
behind modelling, simulating & analyzing any dynamic physical system is to get
a desired parametric output from it & control logic development is all about
minimizing the error by producing the necessary actuating signal. This
becomes challenging due to the unknown mathematical model of the
disturbance force!
ODEs
Let’s elaborate on the general form of the ODE of a single variable linear time
invariant dynamic system:

( )
n i
d y
∑ ci
dt
i
+ay (t)=F act (t )−F dis (t )
i=1

The first term on the right denotes the summation of all the possible n order
time derivatives of the parameter y multiplied by their respective time
invariant coefficientsc i;
the second term denotes the instantaneous state of the parameter y
multiplied by a time invariant compounding factor a ;
These right-hand terms get equated to a time varying inducing signal term
which is the difference between the actuating signal and the disturbance
signal.
This ODE is linear as all the right-hand terms are of power 1.
In the mechanical domain, form of this ODE governs the displacement ( y ) of a
spring ( k ) mass ( m ) damper ( b ) system subjected to time varying forces ( F ):

( )
2
d y dy
m 2
+b +ky ( t ) =Fact (t)−F dis (t)
dt dt

In the electrical domain, the exact form of ODE governs the charge ( q ) flowing
in an inductor ( l )capacitor ( c )resistor ( r ) system subjected to a time varying
voltage ( V ):

( )
d q dq q ( t )
2
l 2
+r + =V (t )
dt dt c

Interestingly, in the financial domain also, a reduced form of this ODE governs
the accumulated sum of money ( S ) with respect to a compound rate of interest
(r )

( dSdt )+rS=0
Hence, there would be quite a few more such examples through which an
equivalence across domains can be established!

Solving ODEs
Once the governing ODE of a physical system has been derived using the basic
principles pertaining to the relevant domain, the obvious next step would be to
obtain a time dependent function of the desired parameter by solving the ODE.
It’s known by experience that the lesser mathematicians amongst us dread the
sight of ODEs! This apprehension is further accentuated if the system has
multiple parameters (akin to a multiple degrees of freedom system in multi
body dynamics domain) to be accounted for, which is usually the case with real
life systems. The same can be attributed to the intricate conventional methods
of solving the non-homogenous ODEs (that have inducing signal terms) like the
method of undetermined coefficients & variation of parameters. On top of this,
in the case of a multi variable system, extensive linear algebra concepts
involving matrices, determinants, eigenvalues & eigenvectors also need to be
applied to get the solution. Therefore, if given a choice, one would prefer
solving an algebraic equation than a differential equation.
Now, this is where Laplace transform emerges as a powerful method of solving
non-homogenous ODEs. We all must have studied about Laplace transforms in
our under-graduation maths course but I wonder how many of us grasped its
application. Let’s first define the operation performed by this transform:
∞
F ( s )=L ( f )=∫ e
−st
f ( t ) dt
0

So, what this transform essentially does is to convert a differential equation

(time domain) to an algebraic one (s domain) hence enabling us to solve this
equation by purely algebraic manipulations. It can be shown, on a more
detailed study, that the algebraic form of the ODE of a single variable LTI
dynamic system is:
n

∑ ( c i si Y ( s ) )+ aY ( s )=L ( F act (t )−F dis ( t ) ) +initial condition terms

i=1

The solution of this algebraic equation is inversely transformed to obtain the

solution in the time domain. The real strength of this method is manifested in
the case of a multi variable coupled linear system. It gets solved without
resorting to the conventional decoupling (eigenvalue-eigenvector) method.

Solving ODEs: Doosra!!

It was highlighted in the previous section that life can be made easier for us
lesser mathematicians by converting an ODE to an algebraic one by applying
Laplace transforms. But there would still remain a predisposition to get
overwhelmed by the tediousness of the steps involved in obtaining a solution
via this route as well (It was rightly brought to notice in one of the comments!).
Therefore, many would wonder about the possibility of getting a lead to the
solution before attempting to solve an intricate ODE. The good news is that it is
definitely possible! And the approach for achieving that would be (which
should always be the rule while approaching any intricate problem) to first
focus on solving the simplest possible ODE, that is, a first order single variable
homogenous linear time invariant ODE:

( dydt )−ay ( t )=0

Its solution is ascertained from basic integral calculus:

 y ( t ) = y ( 0 ) eat

Now, from this & using principal of superposition, it can be concluded that a
general solution to a homogenous LTI n order single variable ODE would be:
n
y ( t ) =∑ k i e λ ti

i=1

k i depends on initial conditions;

λ i are the roots of the algebraic equation:

n

∑ ( c i si Y ( s ) )−aY ( s )=0
i=1

and depend on the value of coefficients c i & the compounding factor a .

The fact being implied here is that irrespective of the order & number of
variables of a homogeneous LTI ODE, it’s solution will always be of exponential
nature. Inferences that can be drawn on further scrutinising this:
 A pure negative real λ pulls the parameter ( y ) asymptotically towards
zero with time.
 A pure positive real λ pushes the parameter ( y ) asymptotically towards
infinity with time.
 A pure imaginary λ causes a sinusoidal oscillatory behaviour of the
parameter ( y ) with time.
 If λ is a combination of negative real & imaginary parts, it’ll result in an
exponential decaying sinusoidal oscillatory behaviour of the parameter
( y ) with time.
 If λ is a combination of positive real & imaginary parts, it’ll result in an
exponential rising sinusoidal oscillatory behaviour of the parameter ( y )
with time.
These inferences form the core of the role of a control system designer which
is to make apt adjustments to the value of coefficients c i & the compounding
factor a in order to extract the desired behaviour out of any dynamic system.

The concept of transfer function

As stated before, the general form of the non-homogeneous ODE of a single
variable linear time invariant dynamic system is:

( )
n i
d y
∑ ci
dt
i
+ay (t)=F induced (t)
i=1

On applying Laplace transform, its analogous algebraic form becomes:

n

∑ ( c i si Y ( s ) )+ aY ( s )=Finduced ( s ) +initial condition terms

i=1

The initial conditions terms can be nullified if the initial conditions are chosen
to be zero. Hence the eventual equation (on rearranging) is:

(∑ ( )
n
c i si ) + a Y ( s ) =Finduced ( s )
i=1

To put this in perspective, F induced is input to the system and gets subjected to a

(∑ ( )
n

transfer function, 1/ c i si ) + a , which results in an output of parameter Y .

i=1

Y ( s) 1
=

(∑ ( )
F induced ( s ) n
ci si ) + a
i=1

Let’s pictorially depict this by means of a block diagram:

Please bear in mind that F induced =F act−F dis.

Control system
The objective of any control system is to generate an optimum input actuation
signal ( F act) that extracts the desired parametric output ( Y desired ) from the system
by neutralising the effect of the disturbance signal ( F dis).

The most effective strategy for accomplishing this objective is by evaluating the
error between the desired and output parametric values & feeding that into
the controller which in turns generates the actuation signal.
Let’s understand this feedback mechanism by means of a block diagram:

( )
n

The transfer function, 1/ ∑ ( c i si )+ a , is also called as plant model.

i=1

Control system example problem

Let’s now apply all the concepts discussed so far to solve a simple control
problem: Design an appropriate control logic to displace a body of mass ‘ m’
from its initial position of x=0 to x=x desired while encountering a constant
disturbance force ‘ F dis’.
Step 1: Determine the governing ODE of the system.
The governing ODE (from Newton’s 2nd law) is:
2
d x
m 2 =F act (t)−F dis
dt

Step 2: Convert the ODE to an algebraic equation using Laplace transform.

In order to implement this step, let’s revise a few basics of Laplace transform:
∞
L ( f ( t ) )=∫ e
− st
f ( t ) dt=F ( s )
0

(∫ )
t
F( s)
L f (t ) dt =
0
s

For nth order derivative of f ( t ); making the initial conditions to be 0:

∞
L ( f ( t ) )=∫ e
n −st n n
f ( t ) dt =s F ( s )
0

Laplace transform of a few important functions are as follows:

∞
k
L ( k )=∫ e
−st
kdt=
0 s
∞
1
L ( e )=∫ e e dt=
at −st at

0 s−a
∞
s
L ( cos ωt )=∫ e
−st
cos ωt dt= 2 2
0 s +ω
∞
ω
L(sin ωt )=∫ e
− st
sin ωt dt= 2 2
0 s +ω

The algebraic equation is:

F dis
m s X ( s ) =L ( F act ( t ) ) −
2
s

The actuation force F act (t ) is chosen proportional to the error ' x desired −x '
F act ( t )=k p (x desired −x )

Hence, the algebraic equation becomes:

m s 2 X ( s ) =k p ( X desired
s
F
−X ( s ) − dis
s )
On rearranging we get:

1 [ ( k p X desired−F dis ) /m ]
X ( s) = .
( )
s k
s2 + p
m

Step 3: Obtain the solution in time domain by using inverse Laplace

transform.

(√ )
t
( k p X desired−F dis ) kp
x (t )=
m ∫ sin m
t dt
0

[ (√ )]
x (t )=( X desired −F dis /k p ) 1 −cos
kp
m
t

We can neutralize the effect of F dis by choosing higher k p with respect to F dis
Hence, the eventual expression can be:

[
x (t )= X desired 1−cos ( √ )]
kp
m
t

The following conclusions are made:

 The response is oscillatory in nature with the amplitude being 2 times
the desired displacement. Hence, it’s not satisfactory.
 The frequency of oscillation is directly proportional to the square root k p.
Higher the chosen k p, quicker would be the response but higher would
be the oscillation rate.
For the response to be acceptable/satisfactory, it should be:
  necessarily bounded (attains a steady state value of x desired);
 preferably non-oscillatory (small amount of overshoot might be ok to
accommodate quicker response).
On inspecting the expression ' X ( s ) ' , it’s easily figured out that the role of ' 1/ s '
in X ( s ) is of integration. Hence, the part of X ( s ) which is the root cause of the
displacement response and needs to be redressed is:
[ ( k p X desired−F dis ) /m ]
(s + km )
2 p

Oscillatory response is due to the roots of s2 + ( kp

m )
being purely imaginary. It is
known, from a basic knowledge of quadratic equations, that for the response
to be acceptable, its roots have to be purely real and negative. In order to
achieve that, a non-zero ' s ' term needs to be introduced and a judicious
choice of that term will result in the desired solution.
Mathematically, a non-zero ' s ' is introduced by choosing F act (t ) as:
F act=k p ( x desired −x) + k d s (x desired−x )

This essentially defines F act as a function of both the error & its derivative and
forms a proportional derivative (PD) controller.
Practically, from a mechanical domain standpoint, it’s akin to introducing a
damper into a spring mass system and choosing such a damping coefficient
that makes this spring mass damper system overdamped!
Using this PD controller governed F act, the solution on rearrangement is:

1 kp
X ( s) = .
[( (
X desired 1+ s
kd
kp) )]
−
F dis
kp

( )
s m 2 kd k p
s +s +
m m

(
For the roots of s2 + s
kd k p
+
m m )to be real; its discriminant > 0:
2
k d >4 m k p
For the roots to be negative; the sum of roots ( )
−k d
m
<0 & product of roots

( km )> 0. As k & k
p
d p ¿ 0, this condition is always met.

Hence, as stated above, a judicious choice of k d & k p will extract the desired
response like:

The two primary acceptance criteria for the response are:

 stable (bounded output at steady state);
 zero/small steady state error (achievement of desired parametric value).
Stability
Stability is the tendency of a dynamic system’s response to return to zero on
being subject to an impulse (like a sudden blow by a hammer). A closed loop
system should output a bonded response for a bounded input. The necessary
and sufficient condition for a dynamic system to be stable is that all the roots
( λ i ) of the transfer function should be having negative ( ¿ 0 ) real parts. On the s-
plane, none of the roots should lie on the right half plane.
For an open loop system (plant):

Y (s ) 1
=k plant ( s )=
F impulse ( s ) P( s)

As F impulse ( s )=1, the impulse response Y ( s )=1/ P ( s ); The roots of P ( s ) are also
known as poles and all these poles should have negative real parts for the
plant to be stable.
For a close loop system:

Y (s ) k k
= controller plant
Y desired ( s ) 1+ k controller k plant

The roots of k controller k plant are called zeros whereas the roots of
1+k controller k plant are called as poles. All the poles of 1+k controller k plant should have
negative reals parts for the closed loop system to be stable.
Final Value Theorem and Steady State Error
Once the stability of the system gets established, the subsequent step is to
ascertain whether the desired parametric value has been achieved or not. In
other words, the steady state error needs to be zero or miniscule. This can be
evaluated in the s domain itself by making use of the final value theorem.
lim error ( t ) =lim sE ( s )
t →∞ s →0

Y desired ( s )
E ss =
1+ k controller k plant
Y desired
For a constant desired parametric value, Y desired ( s )=
s
Y desired
E ss =
1+ k controller k plant

This theorem is applicable to stable systems only. Anyways, steady state makes
no sense for an unstable system!

Root Locus: Part 1

This section will lay the basis for a graphical technique known as root locus for
tackling control problems. As the name suggests, this method involves plotting
all the possible roots (poles) of the characteristic equation of the closed loop
system on the s plane by varying a single controller gain parameter from zero
(which represents the open loop transfer function) to infinity. The advantage
of using the root locus method is that it allows the estimation of the gain by
visually selecting those poles from the plot which elicit the required total
response from the system.
Defining the term required total response: Stability and miniscule steady state
error (in that order) are the foremost response requirements as discussed in
the previous section. Once these are met, the transient response
characteristics like rise time, overshoot and settling time which have an
important bearing on the system’s overall performance need to be assessed.
Therefore, the combination of the steady state and transient state responses
define the total response of a dynamic system. A first order system’s transient
response will only exhibit rise time as it is devoid of any oscillations. A second
order underdamped system is the ideal system to showcase all the
characteristics of transient response. Its typical transfer function is:
2
ωn
2 2
s + 2 ζ ωn s +ωn

ζ is the damping ratio; ω n is the natural frequency.

The poles (roots) of this transfer function are:

s12=−ζ ω n ± j ( √ 1−ζ ) ω n
2

It is inferred that these poles are located at a radius ω n on the left hand s plane
and at an angle of cos−1 ζ with the negative real axis.
The total response is depicted below. It can be seen how the transient state
transitions to the eventual steady state.

Let’s define the transient response characteristics:

 Rise time (t r): This indicates the swiftness of response.
1
tr ∝
ωn √ 1−ξ 2

higher the frequency/lower the damping ratio, swifter would be the

response.
 Overshoot (M p ): The maximum amount the response overshoots the
steady-state or final value.
 −¿ ¿
M p=e

higher the damping ratio, lower would be the M p.

 Settling time (t s): The time taken for the system transients to decay.
4
t s=
ζ ωn
higher the damping ratio and natural frequency, lower would be the
settling time.

The conclusion that can be drawn from the above discussion is that due
to the contrary effect of damping ratio and natural frequency on the
transient response characteristics, a control system engineer has a
discerning task of selecting optimum values that elicit the required total
response from the system.
In the next section, the use of root locus in achieving the same will be
taken up.

Root Locus: Part 2

As described earlier the characteristic equation of a closed loop system is:
1+k controller k plant = 0

On substituting k plant with its polynomial transfer function 1/ P ( s ):

P ( s ) +k controller =0

For sketching the root locus, this characteristic equation needs to be formatted
to:
k controller
1+ =0
P (s )

For a n order plant:

n
P ( s )=∏ ( s−α i )
i=1

α denotes the poles of the transfer function P ( s )

k controller can be broken down to a general form:

 l

∏ ( s−z i )
i=1
k m

∏
i=1
( s− pi )

k is the design parameter of interest or the gain variable ( 0< k <∞ ); z and p are
the zeros & poles respectively of the controller. The characteristic equation
thus becomes:
l

1
∏ ( s−zi ) −1
i=1
n
. m
=
k
∏
i=1
( s−α i ) ∏ ( s− p i )
i=1

This equation provides an insight into the role played by a controller in

achieving the desired/required response from a dynamic system.
Mathematically, the controller modifies the open system transfer function by
adding appropriate zeros and poles. This essentially is the objective of the
root locus design method.
A most useful way of representing any point on the s plane is in polar form
M e , M being the length of the vector from the point to the origin and θ being
iθ

the anti-clockwise angle this vector makes with the positive real axis.
The characteristic equation implies that for any point lying on the root locus,
the product of vector lengths defined from that point to the zeros divided by
the product of vector lengths defined from that point to the poles, should
equate to a vector having a length of 1/k and making an angle of 180 ° (or its
odd multiples) with the positive real axis.
l

∏Mz i
−1
i=1
n m
=
k
∏
i=1
Mα ∏ M p
i=1
i i

θ=∑ zero angles−∑ pole angles=180 ° ( odd multiples )

Simply put, for a point to lie on the root locus it should satisfy the angle
condition. The magnitude happens to be an automatic consequence of that.
The snapshots above show the pole (X) zero (O) plot of a closed loop system.
Let’s look at the rules/guidelines for sketching the root locus without
calculations and which give intuitive insight into the behaviour of the control
system:
1. Number of branches: Branch is the path traversed by a pole as the gain
is varied. It equals the number of poles or zeros, whichever is greater.
2. Starting and ending points: Each root locus branch originates at the open
loop pole (zero gain) & terminates at the open loop zero (infinite gain). If
there are greater number of poles than zeros, then the branches from
the extra poles approach infinity (termed as infinite zero) and vice versa.
3. Symmetry: The root locus is symmetrical about the real axis as the
complex closed loop poles exist in conjugate pairs.
4. Real-axis segments: By making use of the angle condition, it’s found that
on the real axis, the root locus exists to the left of an odd number of
real-axis, finite open loop poles and/or finite open loop zeros (as
indicated by the blue lines in the snapshot below. The resultant root
locus sketch is placed next to that).

In the next section, rules involving calculations and leading to refinement of

the root locus sketch will be taken up.

Let’s look at the remaining basic root locus sketching rule:

5. Behaviour at infinity: In case of unequal number of poles and zeros
(normally the poles are greater in number than the zeros), the deficient
zeros are considered to be at infinity (termed as infinite zeros). The root
 locus approaches these infinite zeros along straight lines known as
asymptotes. The equation of these asymptotes is given by the real-axis
intercept, σ a and angle, θ a as follows:

σ a=
∑ finite poles−∑ finite zeros
¿ finite poles−¿ finite zeros

180 °( odd multiples)

θ a=
¿ poles−¿ zeros

This rule, specifically, and the other four generally would be better
understood by solving an example:
K ( s+3 )
k controller k plant =
( s )( s+1 )( s+2 ) (s+ 4)
Rearranging this to the desired format:
( s +3 ) −1
=
( s ) (s+1)(s +2)(s +4 ) K

Rule 1: The number of finite poles and zeros is four and one respectively.
Hence, there will be four branches (please keep in mind that a branch is
the path traversed by a pole as the gain is varied) of the root locus.
Rule 2: These branches will originate at 0, -1, -2 and -4 on the real axis.
One of the branches will terminate at the finite zero -3 on the real axis
whereas the other three will terminate at the infinite zeros.
Rule 3: Whatever locus is sketched above the real axis has to be mirror
imaged below it.
Rule 4: The portion of the real axis which forms part of the locus will be
in between 0 & -1, -2 & -3 and -4 & -∞ .
Rule 5: The number of asymptotes will equal the number of infinite
zeros – three. These three asymptotes will have a real-axis intercept of:
( 0 ) + (−1 ) + (−2 ) + (−4 ) −(−3 ) −4
σ a= =
4−1 3
at an angle of
180° ( 1 ) 180 ° ( 3 ) °∧180 ° ( 5 )
=60 ° ; =180 =30 0°
4−1 4−1 4−1

The resultant root locus sketch is shown below:

Salient observations:
 Four branches originate from the finite poles (marked by X).
 They move along the portion of the real axis forming part of the locus.
 The asymptotes – two black lines (60° and 300° respectively) & negative
real axis (180°) have a real-axis intercept of -4/3.
 One branch originating from -2 terminates at the finite zero (marked by
O). Another branch which originates from -4 moves towards infinite zero
along the negative real axis. The remaining two branches that originate
from 0 and -1 respectively, have no other choice but to move towards
each other before breaking away from a point (to be estimated by a rule
apparently for refinement) on the real axis and asymptotically approach
the infinite zeros via intercepting the imaginary axis (to be estimated by
another rule apparently for refinement).
So, it has been aptly demonstrated that the five basic rules are good enough to
sketch a reasonable root locus. The other rules for refinement will be taken up
on a need basis. We are well equipped to attack the simple control problem by
using the root locus technique!
Let’s employ the root locus technique to solve the simple control problem
discussed in an earlier section. The problem involved formulating an
appropriate control logic to displace a body of mass ‘m’ from its initial position
of x=0 to x=x desired .

The process begins by setting up the characteristic equation of the closed loop
system:
1+k controller k plant = 0
1
k plant= 2
ms

Selectingk controller=K , which defines the proportional controller gain.

The characteristic equation, on being reduced to the required root locus
format, becomes:
1 −1
=
ms
2
K

Sketching the root locus

Rule 1: The number of finite poles and zeros is two and zero respectively.
Hence, there are two branches of the root locus.
Rules 2 & 3: Both these branches begin at the origin of the s plane and
terminate at the infinite zeros symmetrically about the real axis.
Rule 4: No portion of the real axis forms part of the locus as both the finite
poles lie on the origin and hence no odd number of poles/zeros exist on the
real axis.
Rule 5: The number of asymptotes equal the number of infinite zeros – two.
These two asymptotes have a real-axis intercept of:
2 ( 0 )−( 0 )
σ a= =0
2−0
at an angle of
180° ( 1 ) °∧180 ° ( 3 )
=90 =270 °
2−0 2−0

The resultant root locus sketch is shown below:

Observations:
 The closed loop poles always lie on the imaginary axis and exist as
conjugate pairs.
 This solution in unaccepatable as imaginary poles signify a marginally
stable undamped oscillatory response. Increasing the gain leads to an
increase in the frequency of response which implies a lesser rise time.
The efficacy of the root locus as a design technique becomes evident on
realizing that the controller action, which empowers in achieivng the desired
system response, is founded on the idea of adding appropriate poles/zeros to
the existing locus.
In this example, the foremost objective is to manoeuvre the existing locus to
the left hand plane so that the closed loop poles have a negative real part
which results in a stable dynamic response. It is inferred that, to achieve this
objective, the two branches extending to infinity on the imaginary axis need to
terminate on the negative real axis. This can only be accomplished by adding a
zero on the negative real axis via the controller transfer function:
k controller=K (s + p);

Taking p=1, the characteristic equation becomes:

( s +1 ) −1
=
ms
2
K

Sketching the root locus

Rule 1: Although a finite zero has been added, there are still two branches of
the root locus.
Rules 2 & 3: The branches begin at the origin of the s plane but one of them
will terminate at -1 on the real axis whereas the other one will terminate at the
infinite zero.
Rule 4: The real axis portion which forms part of the locus is in between -1 & -∞
.
Rule 5: There’s only one asymptote which is the negative real axis.
Let’s look at step-by-step drawn sketches to better understand the overall
methodology:

 As the real axis portion to the left of -1 forms part of the locus, the
branches emanating from the origin have no choice but to diverge and
trace the ballpark path.

 The two branches converge and meet the real axis at a break-in point
(which is to the left of -1) from where one branch proceeds towards the
right to termainate at -1 whereas the other one proceeds to the left to
terminate at -∞ .

It’s evident from the ball park root locus sketch derived in the previous section
that by adding a zero to the dynamic system under consideration, all possible
solutions:
 meet the foremost requirement of stability as all the closed loop poles
for K ranging between 0 and ∞ lie on the left hand plane; hence have
negtive real parts.
 have a nil steady state error, E ss . It’s known:
E ss =lim
s →0 ( X desired
1+ k controller k plant )
 [ ( )]
2
ms
E ss =X desired lim 2
=0
s→ 0 m s + K ( s+ 1 )

Therefore, the primary accecpatance criteria for the dynamic resposne is

universally met. The optimal choice of gain is based on the assessment of the
transient response characteristics like rise time, overshoot and settling time.
These depend on the damping ratio, ζ and natural frequency, ω n (shown in an
earlier post).

The features of the ball park sketch that need refinement:

 the angle of departure of the two branches of the locus from the double
pole at the origin.
 the break-in point at which the two branches meet the real axis.
 the angle of arrival of the two branches of the locus at the break in
point.
The angle of departure/arrival is given by:
180° +360 ° ( l−1 )
q

q is the number of branches and l is an integer from 1 to q . For the given

problem, q=2; hence angle of departure/arrival equals 90 ° and 270 ° .
For locating the break-in point, it’s important to understand its mathematical
significance. After, meeting the real axis at the break-in point, the locus moves
towards the respective zeros which are located on the real axis with increasing
gain. Thus, the break-in point essentially is a point of local minimum of the gain
on the real axis.
2
−m s
K=
( s +1 )

dK s ( s+2 )
=−m =0
ds ( s+1 )2

On scrutinizing this solution, it’s found that s=−2 is the point of local
minimum.
The refined root locus sketch is shown below:
For 0< ζ <1 , the damping ratio is indicated by constant angle lines (light blue
coloured). The closed loop poles for a given ζ are located at the intersection of
the locus with the corresponding ζ line. These poles signify stable response
with gradually decaying oscillations. This response is characterised by
decreasing overshoot, increasing rise time and decreasing settling time with
increasing ζ .
For ζ ≥ 1, the closed loop poles are located on the locus at the real axis. These
poles signify stable response without any oscillations. This response is
characterised by no overshoot but increasing rise time with increasing ζ .
Hence, the optimal solution will be based on design requirement and
constraint specifics. Ideally, a swift response (low rise time) is desired but
that’ll also lead to high overshot and high settling time. So, the chosen gain has
to perform a balancing act between these response characteristics.
Occasionally, this choice might also be limited by hardware constraints.