Mathematical
Modeling
Mathematical Modeling
Week 07: Introduction to Dynamic Models
3
Outline of the course
• Part 1: Optimization Model
• Part 2: Dynamic Model
• Chapter 1: Introduction to Dynamic Model
• Chapter 2: Analysis of Dynamic Model
• Chapter 3: Simulation of Dynamic Model
• Part 3: Probability Model
4
Dynamic Models
• Many problems of practical interest involve process that evolve over
time.
• Dynamic models are used to represent the changing behavior of these
systems.
• Space flight
• Electrical circuits
• Population growth
• Military battles
• Spread of disease
5
Outline
• Steady state
• Dynamical system
• Discrete time dynamic systems
6
Steady state analysis: an illustrative example
In an unmanaged tract of forest area, hardwood and softwood trees compete for
the available land and water. The more desirable hardwood trees grow more
slowly but are more durable and produce more valuable timber. Softwood trees
compete with the hardwoods by growing rapidly and consuming the available
water and soil nutrients. Hardwoods compete by growing taller than the
softwoods can and shading new seedlings. They are also more resistant to disease.
Can these two types of trees coexist on one tract of forest land indefinitely, or will
one type of tree drive the other to extinction?
7
Step 1: Ask a question
• Variables: • Assumptions:
• 𝐻 = hardwood population (tons/acre) • 𝑔𝐻 = 𝑟1 𝐻 − 𝑎1 𝐻 2
• 𝑆 = softwood population (tons/acre) • 𝑔𝑆 = 𝑟2 𝑆 − 𝑎2 𝑆 2
• 𝑔𝐻 = growth rate for hardwoods • 𝑐𝐻 = 𝑏1 𝑆𝐻
(tons/acre/year) • 𝑐𝑆 = 𝑏2 𝑆𝐻
• 𝑔𝑆 = growth rate for softwoods • 𝐻 ≥ 0, 𝑆 ≥ 0
(tons/acre/year) • 𝑟1 , 𝑟2 , 𝑎1 , 𝑎2 , 𝑏1 , 𝑏2 are positive reals
• 𝑐𝐻 = loss due to competition for
hardwoods (tons/acre/year) • Objective: Determine whether
• 𝑐𝑆 = loss due to competition for 𝐻 → 0 or 𝑆 → 0
softwoods (tons/acre/year)
8
Step 2: Select a modelling approach
• We will model this problem as a dynamic model in steady state.
• We are given functions:
• Defined on a subset 𝑆 of 𝑅𝑛 . The functions 𝑓1 , … , 𝑓𝑛 represent the rate
of change of each variable 𝑥1 , 𝑥2 , … , 𝑥𝑛 respectively.
9
Step 2: Select a modelling approach
A point 𝑥1 , … , 𝑥𝑛 in the set 𝑆 is called an equilibrium point provide that
At this point, the rate of change of each of the variable 𝑥1 , … , 𝑥𝑛 is equal
to zero. And the system is at rest.
10
Step 2: Select a modelling approach
• Variable 𝑥1 , … , 𝑥𝑛 are called state variables
• 𝑆 is called the state space.
• Since the functions 𝑓1 , … , 𝑓𝑛 depend only on the current state
𝑥1 , … , 𝑥𝑛 of the system, knowledge of current state suffices to
determine the entire feature of the system.
• We only need to know where we are now, not how we got here.
• When we are at an equilibrium point, we say that the system is in
steady state.
11
Step 3: Formulate the problem
• Let 𝑥1 = 𝐻 and 𝑥2 = 𝑆 denote our two state variables, defined on the
state space
{ 𝑥1 , 𝑥2 : 𝑥1 ≥ 0, 𝑥2 ≥ 0}
• The steady-state equations are
12
Step 4: Solve the problem
Solving this system, we have three points
and a point is intersection of two following lines
13
Step 4: Solve the problem
Solving system
We get
14
Step 4: Solve the problem
15
Step 4: Solve the problem
• We have condition 𝑥1 > 0, 𝑥2 > 0.
• We consider additionally a reasonable assumption, the effect of
competition between members of the same species should be stronger
than the competition between species, 𝑎𝑖 > 𝑏𝑖
• Condition for coexistence is
16
Step 5: Answer the question
• The parameters 𝑟𝑖 (for 𝑖 ∈ {1, 2})measure
growth tendency, and the parameters 𝑎𝑖 and
𝑏𝑖 measure the strength of competition
within and between populations,
respectively. Thus, the ratios 𝑟𝑖 /𝑎𝑖 and 𝑟𝑖 /𝑏𝑖
must measure the relative strength of
growth versus competition.
• Let us try to go further. In the absence of
competition between species, the growth
rate is
𝑟𝑖 𝑥𝑖 − 𝑎𝑖 𝑥𝑖2 = 𝑥𝑖 𝑟𝑖 − 𝑎𝑖 𝑥𝑖
17
Step 5: Answer the question
• The ratio 𝑟𝑖 /𝑎𝑖 represents the equilibrium
population level in the absence of competition
between species, or the level at which the
population will stop growing of its own
accord. Similarly, if we neglect the factor of
competition within a population, the net
growth rate is
𝑟𝑖 𝑥𝑖 − 𝑏𝑖 𝑥𝑖 𝑥𝑗 = 𝑥𝑖 𝑟𝑖 − 𝑏𝑖 𝑥𝑗
• The ratio 𝑟𝑖 /𝑏𝑖 thus represents the level of
population 𝑗 necessary to put an end to
growth of population 𝑖.
18
Step 5: Answer the question
For each type of tree (hardwood and softwood), there are two kinds of limits to
growth. The first comes from competition with the other type of tree, and the
second comes from competition between trees of the same type under crowded
conditions. Thus, for each type of tree there is one point where growth will halt
itself due to crowding, and another point where the growth of one type of tree
will halt the growth of the other type due to competition. The condition for
coexistence of both types is that each type reaches the point where it limits its
own growth before it reaches the point where it limits the other’s growth.
19
The steady-state analysis
The steady–state analysis of this section leaves one important question
unanswered. Given that a dynamic model has an equilibrium solution, will we
ever get there? The answer depends on the dynamics of the model. An
equilibrium point
𝑥0 = (𝑥10 , … , 𝑥𝑛0 )
is said to be asymptotically stable (or just stable) if whenever the state
variables
(𝑥1 𝑡 , … , 𝑥𝑛 𝑡 )
pass sufficiently close to x0, they are drawn into the equilibrium. In other
words,
𝑥1 𝑡 , … , 𝑥𝑛 𝑡 → 𝑥0
20
Outline
• Steady state
• Dynamical system
• Discrete time dynamic systems
21
Dynamical system
• Dynamical system models are the most commonly used type of
dynamic model.
• In a dynamical system model the forces of change are represented by
differential equations.
• In this section we focus on the graphical method for obtaining
qualitative information about a dynamical system.
• The emphasis will be on questions of stability.
22
Dynamical Systems: An illustrative example
• The blue whale and fin whale are two similar species that inhabit the same
areas. Hence, they are thought to compete. The intrinsic growth rate of each
species is estimated at 5% per year for the blue whale and 8% per year for
the fin whale. The environmental carrying capacity (the maximum number of
whales that the environment can support) is estimated at 150,000 blues and
400,000 fins. The extent to which the whales compete is unknown. In the
last 100 years intense harvesting has reduced the whale population to
around 5,000 blues and 70,000 fins. Will the blue whale become extinct?
23
Step 1: Ask a question
• Variables: • Assumptions:
• 𝐵 = number of blue whales • 𝑔𝐵 = 0.05𝐵(1 −
𝐵
)
150000
• 𝐹 = number of fin whales 𝐹
• 𝑔𝐹 = 0.08𝐹(1 − )
• 𝑔𝐵 = growth rate of blue whale 400000
population (per year) • 𝑐𝐵 = 𝑐𝐹 = 𝛼𝐵𝐹1
• 𝑔𝐹 = growth rate of fin whale • 𝐵 ≥ 0, 𝐹 ≥ 0
population (per year) • 𝛼 is positive real
• 𝑐B = effect of competition on blue • Objective: Determine whether the
whales (whales per year)
dynamic system can reach stable
• 𝑐𝐹 = effect of competition on fin
whales (whales per year)
equilibrium starting from 𝐵 =
5000 and 𝐹 = 70000
24
Step 2: Select a modelling approach
We will model this problem as dynamic system.
• A dynamical system consists of 𝑛 state variable 𝑥1 , … , 𝑥𝑛 and a
system of differential equations
defined on the state space 𝑥1 , … , 𝑥𝑛 ∈ 𝑆, where 𝑆 is a subset of 𝑅𝑛 .
25
Step 2: Select a modelling approach
• The existence and uniqueness theorem of differential equation states
that if 𝑓1 , … , 𝑓𝑛 have continuous first partial derivatives in a
neighborhood of a point 𝑥0 = 𝑥10 , … , 𝑥𝑛0 there exists a unique solution
to this system of differential equations through this initial condition.
• It is best to think of a solution to a dynamical system as a path through
the state space. As long as differentiability assumptions are satisfied,
there is a path through each point, and paths cannot cross except at an
equilibrium.
26
Step 2: Select a modelling approach
• Let 𝑥 = 𝑥1 , … , 𝑥𝑛 , 𝐹 𝑥 = 𝑓1 𝑥 , … , 𝑓𝑛 𝑥 .
𝑑𝑥
• The dynamical system equations is =𝐹 𝑥
𝑑𝑡
• For a path 𝑥(𝑡), the derivative 𝑑𝑥/𝑑𝑡 represents the velocity vector.
• Hence, for every solution curve 𝑥(𝑡), we have that 𝐹(𝑥 𝑡 ) is the
velocity vector at each point.
• The points where 𝐹 𝑥 = 0 are the equilibria, and we will pay special
attention to the vector field nearby these points.
27
Step 3: Formulate the problem
• Let 𝑥1 = 𝐵, 𝑥2 = 𝐹, we write
where
The state space 𝑆 = { 𝑥1 , 𝑥2 : 𝑥1 > 0, 𝑥2 > 0}
28
Step 4: Solve the model
• From the dynamical system we sketch above graph
29
Step 4: Solve the model
• From the graph, we have 4 equilibrium
solutions
• (0,0)
• (150000, 0)
• (0, 400000)
• And a point whose coordinate depend on
𝛼 (note that the graph assume 400000 <
0.05/𝛼
• Due to the initial condition 𝑥1 0 =
5000, 𝑥2 0 = 70000, solution tends to
the fourth equilibrium point.
30
Step 5: Answer the question
• If 40000 > 0.05/𝛼, then two species could not coexist.
• In the absence of further harvesting, the whale
populations will grow back to their natural levels, and the ecological
system will
remain in stable equilibrium.
31
Sensitivity analysis
Let do sensitivity analysis for parameter 𝛼
• For any value 𝛼 < 1.25 × 10−7 there is a equilibrium 𝑥1 > 0, 𝑥2 > 0 at
where
32
Sensitivity analysis
33
Outline
• Steady state
• Dynamical system
• Discrete time dynamic systems
34
Discrete time dynamic systems: An illustrative example
• Astronauts in training are required to practice a docking maneuver
under manual control. As a part of this maneuver, it is required to bring
an orbiting spacecraft to rest relative to another orbiting craft. The
hand controls provide for variable acceleration and deceleration, and
there is a device on board that measures the rate of closing between
the two vehicles. The following strategy has been proposed for bringing
the craft to rest. First, look at the closing velocity. If it is zero, we
are done. Otherwise, remember the closing velocity and look at
the acceleration control. Move the acceleration control so that it
is opposite to the closing velocity (i.e., if closing velocity is positive,
we slow down, and we speed up if it is negative) and proportional
in magnitude (i.e., we brake twice as hard if we find ourselves closing
twice as fast). After a time, look at the closing velocity again and repeat
the procedure. Under what circumstances will this strategy be
effective?
35
Discrete time dynamic systems
36
Discrete time dynamic systems
37
Step 1: Ask a question
• Variables:
• 𝑡𝑛 = time of 𝑛𝑡ℎ velocity observation (sec)
• 𝑣𝑛 = velocity at the time 𝑡𝑛 (m/sec)
• 𝑐𝑛 = time to make 𝑛𝑡ℎ control adjustment (sec)
• 𝑎𝑛 = acceleration after 𝑛𝑡ℎ adjustment (m/sec2)
• 𝑤𝑛 = wait before 𝑛 + 1 𝑡ℎ observation (sec)
• Assumptions:
• 𝑡𝑛+1 = 𝑡𝑛 + 𝑐𝑛 + 𝑤𝑛
• 𝑣𝑛+1 = 𝑣𝑛 + 𝑎𝑛−1 𝑐𝑛 + 𝑎𝑛 𝑤𝑛
• 𝑐𝑛 ≥ 0, 𝑤𝑛 ≥ 0
• Objective: Determine whether 𝑣𝑛 → 0
38
Step 2: Select a modelling approach
• We will model this problem as a discrete-time dynamical system.
39
Step 2: Select a modelling approach
40
Step 2: Select a modelling approach
41
Step 2: Select a modelling approach
42
Step 2: Select a modelling approach
• Example: Let 𝑥 = 𝑥1 , 𝑥2 and
consider the difference equation
∆𝑥 = −𝜆 𝑥
Where 𝜆 > 0. What is the behavior of
solutions near the equilibrium point
𝑥0 = 0,0 ?
43
Step 2: Select a modelling approach
• Figure shows a graph of the vector field
𝐹(𝑥) = −𝜆𝑥 in the case where 0 < 𝜆 <
1. It is clear that 𝑥0 = (0, 0) is a stable
equilibrium. Each step moves closer to 𝑥0 .
Now let us consider what happens when 𝜆
becomes larger. Each of the vectors in
Figure will stretch as 𝜆 increases. For 𝜆 >
1 the vectors are so long that they
overshoot the equilibrium. For 𝜆 > 2 they
are so long that the terminal point 𝑥(𝑛 +
1) is actually farther away from (0, 0) than
the starting point 𝑥(𝑛). In this case 𝑥0 is an
unstable equilibrium.
44
Step 3: Formulate the problem
• We return now to the docking problem of the example Docking Maneuver.
Step 3 of the five-step method is to formulate the model. We are modeling
the docking problem as a discrete–time dynamical system. From the first
step, we have
𝑣𝑛+1 − 𝑣𝑛 = −𝑘𝑣𝑛−1 𝑐𝑛 − 𝑘𝑣𝑛 𝑤𝑛
• Hence, the change in velocity over the nth time step depends on both 𝑣𝑛
and 𝑣𝑛−1 . To simplify the analysis, let us assume that 𝑐𝑛 = 𝑐 and 𝑤𝑛 = 𝑤 for
all 𝑛. Then the length of each time step is
∆𝑡 = 𝑐 + 𝑤
seconds, and we do not need to include time as a state variable
45
Step 3: Formulate the problem
• We do, however, need to include both 𝑣𝑛 and 𝑣𝑛−1 . Let
𝑥1 𝑛 = 𝑣𝑛
𝑥2 𝑛 = 𝑣𝑛−1
Compute
∆𝑥1 = −𝑘𝑤𝑥1 − 𝑘𝑐𝑥2
∆𝑥2 = 𝑥1 − 𝑥2
The state space is 𝑥1 , 𝑥2 ∈ 𝑅2
46
Step 4: Solve the problem
• Step 4 is to solve the model. There is one
equilibrium point (0,0) found at the
intersection of the two lines:
𝑘𝑤𝑥1 + 𝑘𝑐𝑤2 = 0
𝑥1 − 𝑥2 = 0
• These are the steady–state equations
obtained by setting ∆𝑥1 = 0 and ∆𝑥2 = 0
• The figure shows a graph of the vector field
𝐹 𝑥 = (−𝑘𝑤𝑥1 − 𝑘𝑐𝑥2 , 𝑥1 − 𝑥2 )
• It appears as though solutions will tend
toward equilibrium, but it is hard to be sure. If
𝑘, 𝑐, and 𝑤 are large, then the equilibrium is
probably unstable, but once again it is difficult
to tell.
47
Step 5: Answer the question
• Step 5 is to answer the analysis question in plain English. Maybe we should just
say in plain English that we don’t know the answer. However, we probably can
do better than that. Let us report that a completely satisfactory solution is not
obtainable by elementary graphical methods. In other words, it will take more
work using more sophisticated methods to determine exactly the conditions
under which the proposed control strategy will work. It does seem that the
strategy will be effective in most cases as long as the time interval between
control adjustments is not too long and the magnitude of those adjustments is
not too large. The problem is complicated by the fact that there is a time delay
between reading the velocity indicator and adjusting the controls. Since the
actual closing velocity may change during this interval, we are acting on dated
and inaccurate information. This adds an element of uncertainty to our
calculations. If we ignore the effects of this time delay (which may be permissible
if the delay is small), we can then draw some general conclusions, which are as
follows.
48
Step 5: Answer the question
• The control strategy will work so long as the control adjustments are not too
violent. Furthermore, the longer the interval between adjustments, the lighter
those adjustments must be. In addition, the relationship is one of proportion.
If we go twice as long between adjustments, we can only use half as much
control. To be specific, if we adjust the controls once every 10 seconds, then
we can only set the acceleration controls at 1/10 of the velocity setting to
avoid overshooting the target velocity of zero. In order to allow for human
and equipment error, we should actually set the controls somewhat lower,
say 1/15 or 1/20 of velocity. More frequent adjustments require more
frequent observations of the closing velocity indicator and more
concentration on the part of the operator, but they do allow for the
successful administration of more thrusting power under control. Presumably,
this would be advantageous.
49
THANK YOU !
50