MASTER

NUMERICAL METHODS AND APPLIED STATISTICS

Project on :

"Plant-Pollinator interaction"
Zoubir Nadia Zahir Yasmine

Supervised by :
Pr.Mustapha Lhous
Mathematics department

2021-2022
 CONTENTS

Motivation 3

1 General Introduction : 5

2 Problem of optimal control for bilinear systems : 6

2.1 Characterisation of the optimal control . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.1 Preliminary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Optimal control for the problem Pϵ . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.3 Sequential characterization of the solution of the problem (P ) . . . . . . . . . 7
2.1.4 Optimal feedback control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.1 Wave equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 Heat equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.3 Transport equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 REGIONAL CONTROL PROBLEM FOR DISTRIBUTED BILINEAR SYSTEMS:

APPROACH AND SIMULATIONS 11
3.1 Regional quadratic control problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Regional minimum energy control problem . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Numerical approach and simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4 Model Analysis : 18

5 Phase Portraits : 19

6 Conclusion: 20

Appendix A : References 21

Appendix B : Terms and Definitions 22

2
 LIST OF FIGURES

√
3.1 Phase Portrait for K = 0.8 > 2 − 1 and K > 1
(i.e ∆ > 0) . . . . . . . . . . . . . . 14
√ 2
3.2 Phase Portrait for K = 0.8 > 2 − 1 and K > 1
(i.e ∆ > 0) . . . . . . . . . . . . . . 15
√ 2
3.3 Phase Portrait for K = 0.8 > 2 − 1 and K > 1
(i.e ∆ > 0) . . . . . . . . . . . . . . 16
√ 2
3.4 Phase Portrait for K = 0.8 > 2 − 1 and K > 1
2
(i.e ∆ > 0) . . . . . . . . . . . . . . 17

3
 MOTIVATION :

Distributed systems are dynamical systems whose state depends on time and evolves on a spatial
domain. They are frequently encountered in practical engineering problems. Examples of furnaces
for heating metal slabs or heat exchangers, aircrafts and robot arms or energy transmission lines.
Appropriate mathematical modelling of such systems yields most often non-linear partial differential
equations. Clearly such models involve using sophisticated mathematical methods, which require to
describe the process more accurately and to implement more effective control strategies. A wide liter-
ature is devoted to control of distributed systems and many notions have been studied and explored .

Infinite dimensional Bilinear systems analysis can formulate many real problems. The control-
lability is among the most important analysis notions, within there are many concepts as exact
controllability, approximate controllability, regional controllability and so on. Ball, Marsden, and
Slemrod, discussed the controllability for distributed Bilinear systems. Bradly, Lenhart, and Yong
treated the optimal control of the Velocity term in a Kirchhoff plate equation.

4
CHAPTER 1
GENERAL INTRODUCTION :

let Ω ⊂ Rn (n = 1, 2, 3) be a spatial domain with regular boundary ∂Ω.

For T > 0 and Π = Ω×]0, T [, Σ = ∂Ω×]0, T [, consider a bilinear system described here

= Ay(x, t) + u(t)By(x, t) Π
 ∂y(x,t)

 ∂t

(1) y(x, 0) = y 0 (x) Ω

y(x, t) = 0 Σ



Where :

• A generates a C semi-group (T (t)) of bounded linear operators on a Banach space L (Ω)

0
t≥0
2

• B : L (Ω) −→ L (Ω) is a linear bounded operator

2 2

• the control u(t) ∈ U = {u ∈ L ([0, T ]) | −M ≤ u(t) ≤ M } , M is a positif constant.

M
∞

System (1) is bilinear in the pair (u, y), and its solution y is a non-linear function with respect to u.
For a given y 0 ∈ L2 (Ω), (1) is written as
Z t
(2) yu (x, t) = T (t)y (x) +
0
T (t − s)u(s)Byu (x, s)ds
0

and solutions of (2) are called mild solutions of (1), the mild solution y of (1) associated to the
control u exists and the set of reachable states from an initial state y 0 is of dense complement in the
state space.

5
CHAPTER 2
PROBLEM OF OPTIMAL CONTROL FOR BILINEAR SYSTEMS :

2.1 Characterisation of the optimal control

2.1.1 Preliminary
In general, a problem of control aims to achieve a certain degree of performance for the system at
hand using suitable control laws among available options. If this is indeed feasible, then one usually
aims to achieve this performance while optimizing a certain criterion. A problem of optimal control
is an optimization problem on a reasonable set described by dynamic constraints. As an interesting
example, the question of describing the best control among those that allow to reach a desired state
with minimal cost or energy. Such problems arise in various applications, such as the optimization
of hydrothermal systems and non-smooth modeling in mechanics and engineering, etc.
The main goal of this chapter is to study a quadratic optimization problem with a restricted end-
point state. In the case of a bounded set of admissible control, we will characterize the optimal
control either for exactly or approximately attainable states. This problem can be formulated as an
optimization problem with endpoint constraint, which can also be approximated by a set of uncon-
strained problems. Moreover, if the steering control is scalar valued, then the optimal control can be
expressed as a time-varying feedback law.
Let us now consider the following assumptions:
(a) For all y ∈ X the mapping u 7→ B(u, y) is compact, (b) A is the infinitesimal generator of a
linear compact C0 - semigroup S(t).
Note that assumption (b) is systematically satisfied for U = R. The quadratic cost function J to
be minimized is defined by
Z T
rZT
J(u) = ∥y(t)∥2 dt + ∥u(t)∥2U 2 dt.
0 2 0
Here, r > 0 and u belongs to the set of admissible control

Uad = {u ∈ V /y(T ) = yd } ,

where V is a closed convex subset of L2 (0, T ; U ) and yd ∈ X is the desired state. The optimal control
problem may be stated as follows
min J(u)
(
(P)
u ∈ Uad

6
 CHAPTER 2. PROBLEM OF OPTIMAL CONTROL FOR BILINEAR SYSTEMS :

In order to solve the problem (P ), let us introduce the following auxiliary cost function

Jϵ (u) = ∥y(T ) − yd ∥2 + ϵJ(u),

where ϵ > 0, and let us consider the following optimal control problem

minϵ (u)
(
(Pϵ )
u∈V

Definition 1. - A target state yd ∈ X is approximately controllable for the system (1), if for all
ε > 0 there exists uε ∈ V such that ∥yuε (T ) − yd ∥ ≤ ε. - A target state yd ∈ X is exactly attainable
for the system (1), if there exists u ∈ V such that yu (T ) = yd .

The following lemma provides a continuity property of the solution y with respect to the control
u.

Lemma 1. If one of the assumption (a) or (b) hold,

then for any sequence (un ) ⊂ L2 (0, T ; U ) such that un → u in L2 (0, T ; U ), we have

lim sup ∥yn (t) − y(t)∥ = 0,

n→+∞ 0≤t≤T

where yn and y are the mild solutions of the system (1) respectively corresponding to un and u.

2.1.2 Optimal control for the problem Pϵ

The following result discusses the existence of the optimal control related to the auxiliary problem
Pϵ .

Theorem 1. Let one of the assumptions (a) or (b) hold. - If V = {u ∈ L2 (0, T ; U )/∥u∥U ≤ M } for
some M > 0, then there exists an optimal control for the problem (Pϵ ), which satisfies the following
formula: !−1
∥ϵru∗ (t) + (B (., y ∗ (t))∗ ϕ(t)∥U
u (t) = −
∗
+ ϵr B (., y ∗ (t))∗ ϕ(t),
M
where ϕ is the mild solution of the following adjoint system

ϕ̇(t) = −A∗ ϕ(t) − B ∗ (u∗ (t), ϕ(t)) − 2ϵy(t)

(

ϕ(T ) = 2 (y(T ) − yd )

B ∗ (u∗ (t), .) being the adjoint of the operator B (u∗ (t), .). - If V = L2 (0, T ; U ), then the control
defined by
1
u∗ (t) = − (B (., y ∗ (t))∗ ϕ(t)
ϵr
is a solution of the problem (Pϵ ), where ϕ is the mild solution of the adjoint system (7).

2.1.3 Sequential characterization of the solution of the problem (P )

Theorem 2. Assume that V is bounded and let yd be an approximately attainable state by a control
from V . Then the problem (P ) posses a solution. Moreover any weak limit value of (u∗n ) in L2 (0, T ; U )
is a solution of (P ).

7
 CHAPTER 2. PROBLEM OF OPTIMAL CONTROL FOR BILINEAR SYSTEMS :

Theorem 3. If Uad ̸= ∅, then there exists a solution u∗ of the problem (P ). Furthermore, any weak
limit value of the solution (u∗n ) of (Pϵn ) in L2 (0, T ; U ) is a solution of (P ).

2.1.4 Optimal feedback control

In this part we will try to express the optimal control u∗ of the problem (P ) as a time-varying
feedback law for the class of commutative bilinear systems with scalar control [15, 28]. Assume that
U = R, then we can write the system (1) as follows

ẏ(t) = Ay(t) + u(t)By(t)

(

y(0) = y0 ∈ X

where A : D(A) ⊂ X 7→ X is the infinitesimal generator of a linear C0− semi-group S(t), B is a

bounded linear operator and u ∈ V := L2 (0, T ).

Theorem 4. Assume that A and B commute with each other and that Uad ̸= ∅. Let v ∈ Uad and
let y0 ∈ X be such that S(T )y0 ∈
/ Ker(B). Then for any solution u∗ of the problem (P ), we have the
following formula
1ZT 2 ZTZT ∗ 2Z T ∗
u∗ (t) = v(s)ds + ⟨y (s), By ∗ (s)⟩ dsdα − ⟨y (s), By ∗ (s)⟩ ds
T 0 Tr 0 α r t

2.2 Examples
2.2.1 Wave equation
Let es consider the following wave equation
 ∂ 2 y(t,x)
 ∂t2

 = ∆y(t, x) + u(t, x)y(t, x), t ∈ [0, T ] Ω = (0, 1)

y(t, 0) = y(t, 1) = 0 t ∈ [0, T ]
y(0, x) = y0 (x) x∈Ω



where

• u ∈ L (0, T, L (Ω)),
2 2

• T > 4 max |x − x | for some x ∈ R\[0, 1],

x∈Ω 0 0

• the desired state z ∈ H (Ω) ∩ H (Ω) is such that

d
1
0
2
̸= 0) ∈ L∞ (Ω), where ⊮(zd ̸= 0)
∆zd
zd
⊮(zd
indicates the characteristic function of the set (zd =
̸ 0) := {x ∈ Ω/zd (x) ̸= 0}.
This system has the form of the system (1) if we take y(t) = (z(t), ż(t)), X = H01 (Ω) × L2 (Ω) with
⟨(y1 , z1 ) , (y2 , z2 )⟩X = ⟨y1 , y2 ⟩H 1 (Ω) + ⟨z1 , z2 ⟩L2 (Ω) and
0

0 I 0 0
! !
A= with D(A) = H01 (Ω) ∩ H (Ω) ×
2
H01 (Ω) and B = .
∆ 0 I 0
Here B is a compact linear bounded operator on X and A is the infinitesimal generator of a linear
C0 - semi-group S(t) of isometries (see [3], p.176). The quadratic cost function is given by
Z T
2 2
 rZT
J(u) = ∥z(t)∥H01 (Ω) + ∥ż(t)∥L2 (Ω) dt + ∥u(t)∥2L2 (Ω) dt,
0 2 0

8
 CHAPTER 2. PROBLEM OF OPTIMAL CONTROL FOR BILINEAR SYSTEMS :

where u(t) := u(t, · ) and z(t) := z(t, · ). According to [25], there exists a control v ∈ L2 (0, T ; L2 (Ω)
such that the corresponding solution zv of the system (1) verifies zv (T ) = zd . Then, according to
Theorem 5 there exists a control u∗ ∈ L2 (0, T, R), which guarantees the exact attainability of zd at
time T , and is a a/solutibn of thes problem (P ) with Uad = {u ∈ L2 (0, T, L2 (Ω)) /z(T ) = zd }.

2.2.2 Heat equation

In this part we study the optimal exact attainability for the reaction-diffusion equation. Let us
consider the following system

= ∆y(t, x) + u(t, x)y(t, x), in Q = Ω × (0, T ), T > 0


∂
 ∂t y(t, x)



y(t, 0) = y(t, 1) = 0, on (0, T )
y(0) = y0 in Ω



where Ω = (0, 1) and u ∈ L2 (0, T, U ) is a control function.

Case 1: Distributed control (U = L2 (Ω))

Assume that y0 , yd ∈ L2 (Ω) are such that

- for a.e. x ∈ Ω, yd y0 ≥ 0,
- for a.e. x ∈Ω, y0 (x) = 0 ⇐⇒ yd (x) = 0,
- a := ln yyd0 1(y0 ̸= 0) ∈ L∞ (Ω), where ⊮(y0 ̸= 0) indicates the characteristic function of the set
(y0 ̸= 0) := {x ∈ Ω/y0 (x) ̸= 0}.
- ∆y d
1
yd (yd ̸=0)
∈ L∞ (Ω),
- |yd | > 0 a.e. on some nonempty open subset O of Ω.
According to Theorem 2 in [24], there is a time T for which yd is exactly attainable for the system
(28) using a control v ∈ L2 (0, T, L2 (Ω)), so Uad ̸= ∅. Then, according to Theorem 5 , there exists a
control u∗ which guarantees the exact attainability of yd at time T , and is solution of the following
problem
min J(u)
(
n   o
= u ∈ L2 0, T, L2 (Ω) /yu (T ) = yd
u ∈ Uad
more precisely any weak limit of u∗n given by theorem 3 corresponding to sequence (ϵn ) gives an
optimal control u∗

Case 2: Scalar control (U = R)

Here, we have u(t, x) = u(t) ∈ R. Assume that y0 , yd ∈ L2 (Ω) are such that yd = λy0 with λ > 1 and
y0 > 0, a.e in Ω. According to Theorem II 4 and Remark 4 in [26], there is a time T for which yd is
exactly attainable for the system (28) using the control v(t) = T +(λ−1)t
λ−1
∈ L2 (0, T, R), so Uad ̸= ∅. By
Theorem 6 , there exists a feedback control u∗ ∈ L2 (0, T, R) which guarantees the exact attainability
of yd at time T , and is solution of the problem (P ) with Uad = {u ∈ L2 (0, T, R) y ∗ (T ) = yd }, and
satisfies the following formula
1 2 Z T Z T ( 2 2Z T ∗
u t) = ln(λ) +
(
y s) dsdα − ∥y (s)∥2 ds
T Tr 0 a r t

9
 CHAPTER 2. PROBLEM OF OPTIMAL CONTROL FOR BILINEAR SYSTEMS :

2.2.3 Transport equation

Let us consider the following transport problem

= − ∂x y(t, x) + u(t)y(t, x), t ∈ (0, T ), x ∈ Ω = (0, +∞)


∂ ∂
 ∂t y(t, x)



y(t, 0) = 0, t ∈ (0, T )
y(0, x) = y0 (x), x∈Ω



where u ∈ L2 (0, T ). Here the operator A = − ∂x

∂
with the domain D(A) = H01 (Ω) generates a C0 -
semi-group of isometries S(t) in X = L (Ω). Below, we will develop numerical simulation for the
2

example (30). For this end, we take r = 2, T = 9, y0 = x exp(−x) and

0, if x ≤ 9
(
yd (x) =
(x − 9) exp(9 − x), if x ≥ 9
n o
then the control v = 0 ∈ Uad = u ∈ L2 (0, T )/y ( T ) = yd . By Theorem 6 , there exists a feedback
control u∈ L2 (0, T ) which guarantees the exact attainability of yd at time T . Moreover u∗ is the
solution of the problem (P ) and satisfies the following formula

1 Z T Z T ( 2 Z T
u t) =
(
y s) dsdα − ∥y ∗ (s)∥2 ds.
T 0 a t

In the Figure 1, we compare numerically the two controls u∗ and v = 0 in term of the state at the
finite time T = 9. Moreover, we find J(u∗ ) = 1.2442 and J(v) = 2.25 ≈ 2J (u∗ ). We observe that
the desired state is exactly attainable either by using the optimal control u∗ or the control v = 0.
However, the control u∗ leads to a lower cost than the zero control.

Remark. Unlike the case of linear systems, the uniqueness of the optimal control of the quadratic
cost (2) is not guaranteed in general when dealing with bilinear systems, which is due to the lack
of convexity of the state w.r.t control. However, in the case of the quadratic cost function J(u) =
0 u (t)dt, the uniqueness of the optimal control is assured by the strict convexity of the cost J.
RT 2

10
CHAPTER 3
REGIONAL CONTROL PROBLEM FOR DISTRIBUTED BILINEAR
SYSTEMS: APPROACH AND SIMULATIONS

3.1 Regional quadratic control problem

Given T > 0, let us associate with (2) the problem

min u ∈ L2 [0, T ]Jϵ(u)

with
Jϵ (u) = ∥χω zu (T ) − zd ∥2L2 (ω)
Z T
+ϵ u2 (t)dt, ϵ > 0.
0

Proposition 1. 1. For u ∈ L2 [0, T ] and h ∈ L2 [0, T ], ∀t ∈ [0, T ], we have ∥zu+h (t) − zu (t)∥ = o(∥h∥)
as h → 0. 2. There exists u∗ ∈ L2 [0, T ] such that

Jε (u∗ ) = J ∗ = min v ∈ L2 [0, T ]Jε(v).

Let Z t
y(t) = U (t, s)h(s)Bzu (s)ds.
0
Then
∥zu+h (t) − zu (t) − y(t)∥ = o(∥h∥)
as h → 0, where (U (t, s))t≥s is the evolution operator generated by A + uB.

11
 CHAPTER 3. REGIONAL CONTROL PROBLEM FOR DISTRIBUTED BILINEAR
SYSTEMS: APPROACH AND SIMULATIONS

Theorem 5. A control which minimizes the problem (3) is given by

1
u(t) = − ⟨Bz(t), P (t)z(t) − U ∗ (T, t)χ∗ω zd ⟩
ϵ
where P is the selfadjoint and nonnegative operator solution of the following equation:

 dt ⟨P (t)y, z⟩ + ⟨P (t)y, (A + u(t)B)z⟩

 d

+⟨(A + u(t)B)y, P (t)z⟩ = 0, ∀y, z ∈ D(A)
P (T ) = χ∗ω χω∗



Here U ∗ (t, s) is the adjoint operator of U (t, s) and χ∗ω is the adjoint operator of χω . The minimum
is given by
Jε (u) = ⟨P (0)z0 , z0 ⟩ + 2 ⟨χω U (T, 0)z0 , zd ⟩
Z T
+ ∥zd ∥2L2 (ω) + ε u2 (t)dt
0

Proposition 2. 1. The sequence (Jϵ (uϵ ))ϵ>0 is decreasing as ϵ → 0. 2. The sequence

Z T !
u2ϵ (t)dt
0 ϵ>0
 
2
is increasing as ϵ → 0. 3. The sequence ∥χω zϵ (T ) − zd ∥ L2 (ω) ϵ > 0 is decreasing as ϵ → 0, and
∀ϵ > 0
∥χω zϵ (T ) − zd ∥ L2 (ω) ≤ ∥χωS(T )z0 − zd ∥L2 (ω) .
In particular, there exists a subsequence of (χω zϵ (T )− zd )ϵ>0 which converges weakly in L2 (ω).

3.2 Regional minimum energy control problem

Here let us go back to the problem (2), and consider the set

R(T ) = {zu (T )}
[

u∈L2 [0,T ]

of the states reachable at time T from z0 . We have the main result.

Theorem 6. Let uϵ be a solution of (3) and assume that Uad (ω) is nonempty. Then we have

uϵ → u⋆ as ϵ → 0 in L2 [0, T ]

and  
χω zϵ → χω zu∗ in C [0, T ]; L2 (ω) .
Moreover, u⋆ is a solution to the problem (2).

Proposition 3. Suppose that Uad (Ω) is nonempty and L2 (Ω) has an orthonormal basis (ϕn )n of
eigenfunctions of A. In addition, if A commutes with B, then the problem (2) has only one solution.

12
 CHAPTER 3. REGIONAL CONTROL PROBLEM FOR DISTRIBUTED BILINEAR
SYSTEMS: APPROACH AND SIMULATIONS

3.3 Numerical approach and simulations

Example
Let Ω =]0, 1[ and consider the bilinear system described by the following evolution equation:
 2
 ∂z
(x, t) = α ∂ ∂x
z(x,t)
2 + βz(x, t)
 ∂t


+γu(t)z(x, t) in Ω×]0, T [,



 z(x, 0) = z0 (x) in Ω,
z(0, t) = z(1, t) = 0 on ]0, T [,




where α, β and γ are positive real numbers. This equation may represent a simplified model of the
temperature distribution in a furnace. The system (21) looks like (1) with

∂2
Ā = α +β
∂x2
with the domain n o
D(Ā) = z ∈ H 2 (0, 1) | z(0) = z(1) = 0 .

The operator Ā admits a set of eigenfunctions ϕi ( · ) associated with the eigenvalues λi given by
√
ϕi (x) = 2 sin(iπx), λi = β − αi2 π 2 , i ≥ 1.

The solution (21) is approximated by

M
ai (t)ϕi (x).
X
z(x, t) ≃
i=1

Let z0 (x) = sin(πx), zd (x) = 8x(1 − x), α = 0.01, β = 0.01, γ = 0.02, ε = 0.0001 and T = 1.
Augmenting the truncation order M beyond 5 does not improve the simulation results.
Using the above algorithm for different regions of ω and after the 7-th iteration, we have (i) Case
of ω =]0.4, 0.6[ s see Figs. 1 and 2 , (ii) Case of ω =]0.6, 1[ : see Figs. 3 and 4 .

13
CHAPTER 3. REGIONAL CONTROL PROBLEM FOR DISTRIBUTED BILINEAR
SYSTEMS: APPROACH AND SIMULATIONS

√
Figure 3.1: Phase Portrait for K = 0.8 > 2 − 1 and K > 1
2
(i.e ∆ > 0)

14
CHAPTER 3. REGIONAL CONTROL PROBLEM FOR DISTRIBUTED BILINEAR
SYSTEMS: APPROACH AND SIMULATIONS

√
Figure 3.2: Phase Portrait for K = 0.8 > 2 − 1 and K > 1
2
(i.e ∆ > 0)

15
CHAPTER 3. REGIONAL CONTROL PROBLEM FOR DISTRIBUTED BILINEAR
SYSTEMS: APPROACH AND SIMULATIONS

√
Figure 3.3: Phase Portrait for K = 0.8 > 2 − 1 and K > 1
2
(i.e ∆ > 0)

16
CHAPTER 3. REGIONAL CONTROL PROBLEM FOR DISTRIBUTED BILINEAR
SYSTEMS: APPROACH AND SIMULATIONS

√
Figure 3.4: Phase Portrait for K = 0.8 > 2 − 1 and K > 1
2
(i.e ∆ > 0)

17
CHAPTER 4
MODEL ANALYSIS :

18
CHAPTER 5
PHASE PORTRAITS :

19
CHAPTER 6
CONCLUSION:

Only recently have mutualisms been subject to the same level of attention from mathematicians
and biologists. As a result, it is not surprising that our understanding of their evolution ramains
rudimentary. Further studies of the fate of mutualisms in response to environmental change are
therefore essential if the goal is to conserve higher units of biological organization. In this paper,
we have tried to develop a flexible model of plant and pollinator population dynamics that allows
representation of many kinds of interaction network structures. Our model is based on a mechanistic
representation of plant-pollinator interactions that accounts for the specific properties of this type
of mutualism. We have attempted to bridge the gap between mathematics and population bilogy.

20
 APPENDIX A : REFERENCES

F. REED HAINSWORTH. Foraging Efficiency and Parental Care in Colibri Coruscans

Gita BENADI. Linking Specialisation and Stability of Plant- Pollinator Networks

Gita BENADI. Requirements for plant coexistence through pollination niche partitioning

HAINSWORTH, F. R., AND L. L. WOLF. 197213. Energetics of nectar extraction in a small, high
altitude, tropical hummingbird, Selasphorus flammula. J. Comp. Physiol. 80:377-387.

Heng Huang, and Paolo D’Odorico1. Critical Transitions in Plant-Pollinator Systems Induced by
Positive Inbreeding-Reward-Pollinator Feedbacks

Holling, C.S. (1973). Resilience and stability of ecological systems. Annu. Rev. Ecol. Evol. Syst.
4, 1–23.

Morgan, M., Wilson, W., and Knight, T. (2005). Plant population dynamics, pollinator foraging,
and the selection of self-fertilization. Am. Nat. 166, 169–183.

J. theor. Biol. (1981) 91, 363-378. The Dynamics of a Plant-Pollinator Interaction

Judith L. Bronstein, Ulf Dieckmann and Régis Ferrière. Coevolutionary Dynamics and the
Conservation of Mutualisms

Kunin W, Iwasa Y. 1996. Pollinator foraging strategies in mixed floral arrays: density effects and
floral constancy. Theoret. Popul. Biol. 49, 232–263.

WILLIAM F. MORRIS, DIEGO P. VAZQUEZ AND NATACHA P. CHACOFF. Benefit and cost
curves for typical pollination mutualisms

21
 APPENDIX B : TERMS AND DEFINITIONS

Density-dependent factor : The case where the per capita population growth rate depends
statistically on the density of the population. When the slope of that relation is negative , we call
this negative density-dependence. Negative density dependence a characteristic of a population
undergoing intraspecific competition, where individuals of the same species compete for shared
resources and have negative effects on their demographic rates.

Density-independent factor : also called limiting factor, in ecology, any force that affects the
size of a population of living things regardless of the density of the population (the number of
individuals per unit area).

Holling’s disc equation : A method for calculating the functional response of predators to
increase prey density. The equation is based on laboratory experimental data simulating predation.
The efficiency with which the predator consumes the prey should decline as the prey density
increases, due to extra time spent handling thee prey. Thus the relationship between prey density
and numbers consumed be predators is not a straight line but a curve. This relationship was first
summarized by C.S.Holling as y = 1−abxT tax
where y is the number of discs removed, x is the disc
density, T t is the total experimental time, a is a constant desribing the probability og finding a disc
at a giving density, and b is the time taken to pick up a disk.

Mutisia : A genus of flowering plant in the mutisia tribe within the sunflower family.

Pplane.m : It allows the use of graphical interface to plot direction fields for differential
equations.This matlab package was written and developed by the professor of Rice University John
C. Polking.

22