Dynamic Programming

I divide and conquer method divides the problem into

independent subproblem
I solve subproblem recursively and combine solutions to solve
original problem
I dynamic programming solves problem by combining solutions
to subproblems
I the subproblems are not independent, subproblems share
subsubproblems
I solve subsubproblem just once and then saves answer in table
- avoiding recomputing answer when subsubproblem is
encountered

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Dynamic Programming

I DP applied to optimization problems

I many possible solutions - find solution with optimal value
I called a solution an optimal solution as opposed to the
optimal solution
I since there may be several solutions that achieve the optimal
value
I development of DP algorithm into four steps
I characterize the structure of an optimal solution
I recursively define the value of an optimal solution
I compute the value of an optimal solution in a bottom-up
fashion
I construct an optimal solution from computed information

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

I manufacturing problem - two assembly lines

I chassis enters, add parts at assembly stations and finished
auto exists
I assembly line i(= 1, 2) - n stations (j = 1, 2, . . . , n)
I Si,j jth station on line i
I jth station on both lines doing same function
I stations built at different times with different technologies
I time required at each station varies
I ai,j assembly time required at station Si,j
I entry time ei for chassis to enter line i and exit time xi for
completed auto
I time to go from one station to the next in the same line is
negligible

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

I for a special order - automobiles to be manufactured as

quickly as possible
I need to switch partially built auto from one line to other
though it passes through the n stations in order
I time to transfer a chassis away from line i after gone through
station Si,j is ti,j
I The problem is to determine which stations to choose from
line 1 and which to choose from line 2 inorder to minimize the
total time for one auto
I brute force way of minimizing the time is infeasible when there
are many stations
I possible to compute Θ(n) given a list of stations from line 1
and line 2 to finish an auto
I 2n possible ways to choose stations
I enumerating all possible ways and compute time requires
Ω(2n ) time which is infeasible for large n
M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design
Assembly-line scheduling

I characterize the structure of an optimal solution

I let us consider fastest way for a chassis to get from starting
point through S1,j
I if j = 1 only one way that the chassis could have gone - easy
to determine
I for j = 2, . . . , n two choices
I one transfer chassis from S1,j−1 to S1,j
I another choice chassis could have come from S2,j−1 and
transferred to S1,j with transfer time t2,j−1
I say, fastest way through station S1,j is through S1,j−1
I the chassis must have taken a fastest way from the starting
point through S1,j−1
I if there were a faster way to get through S1,j−1 it can be
substituted to yield a faster way through S1,j
M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design
Assembly-line scheduling

I say, fastest way through station S1,j is through S2,j−1

I the chassis must have taken a fastest way from the starting
point through S2,j−1
I if there were a faster way to get through S2,j−1 it can be
substituted to yield a faster way through S1,j
I an optimal solution to a problem (finding the fastest way
through S1,j ) contains within it an optimal solution to
subproblems (finding the fastest way through either S1,j−1 or
S2,j−1 )
I this property is called as optimal substructure
I construct an optimal solution to a problem from optimal
solutions to subproblems

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

I similarly, fastest way through station S2,j is either

I fastest way through station S2,j−1 and then through S2,j or
I fastest way through station S1,j−1 and then through S2,j
I In general, To solve the problem of finding the fastest way
through station j of either line,
I solve the subproblems of finding the fastest ways through
station j − 1 on both lines
I second step is to construct recursive solution
I define the value of an optimal solution recursively in terms of
optimal solutions to subproblems
I let, fi [j] fastest possible time to get a chassis from starting
point through Si,j
I goal is to determine the fastest time f ∗ to get a chassis all the
way through the factory
M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design
Assembly-line scheduling

I chassis has to get all the way through station n on either line
1 or 2 and then exit
I the faster of these ways is the fastest way through the entire
factory

f ∗ = min(f1 [n] + x1 , f2 [n] + x2 )

f1 [1] = e1 + a1,1
f2 [1] = e2 + a2,1

I to compute fi [j] for j = 2, . . . , n

I for f1 [j] fastest way through S1,j

f1 [j] = min(f1 [j − 1] + a1,j , f2 [j − 1] + t2,j−1 + a1,j )

similarly
f2 [j] = min(f2 [j − 1] + a2,j , f1 [j − 1] + t1,j−1 + a2,j )

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling


e1 + a1,1 if j = 1
f1 [j] =
min(f1 [j − 1] + a1,j , f2 [j − 1] + t2,j−1 + a1,j ) if j ≥ 2

e2 + a2,1 if j = 1
f2 [j] =
min(f2 [j − 1] + a2,j , f1 [j − 1] + t1,j−1 + a2,j ) if j ≥ 2

I fi [j] gives the values of optimal solutions to subproblems

I for constructing an optimal solution
I li [j] line number whose station j − 1 is used in a fastest way
through Si,j
I l ∗ line whose station n is used in a fastest way through the
entire factory

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

line 1 S1,1 S1,2 S1,3 S1,4 S1,5 S1,6 exit

e1 = 2 → 7→ 9→ 3→ 4→ 8→ 4→ x1 = 3
&2 &3 &1 &3 &4

2% 1% 2% 2% 1%
e2 = 4 → 8→ 5→ 6→ 4→ 5→ 7→ x2 = 2
line 2 S2,1 S2,2 S2,3 S2,4 S2,5 S2,6 exit

j 1 2 3 4 5 6 j 2 3 4 5 6
f1 [j] 9 18 20 24 32 35 l1 [j] 1 2 1 1 2
f2 [j] 12 16 22 25 30 37 l2 [j] 1 2 1 2 2
I f ∗ = 38 and l ∗ = 1

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

line 1 S1,1 S1,2 S1,3 S1,4 S1,5 S1,6 exit

e1 = 2 → 7→ 9→ 3→ 4→ 8→ 4→ x1 = 3
&2 &3 &1 &3 &4

2% 1% 2% 2% 1%
e2 = 4 → 8→ 5→ 6→ 4→ 5→ 7→ x2 = 2
line 2 S2,1 S2,2 S2,3 S2,4 S2,5 S2,6 exit

j 1 2 3 4 5 6 j 2 3 4 5 6
f1 [j] 9 l1 [j]
f2 [j] 12 l2 [j]

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

line 1 S1,1 S1,2 S1,3 S1,4 S1,5 S1,6 exit

e1 = 2 → 7→ 9→ 3→ 4→ 8→ 4→ x1 = 3
&2 &3 &1 &3 &4

2% 1% 2% 2% 1%
e2 = 4 → 8→ 5→ 6→ 4→ 5→ 7→ x2 = 2
line 2 S2,1 S2,2 S2,3 S2,4 S2,5 S2,6 exit

j 1 2 3 4 5 6 j 2 3 4 5 6
f1 [j] 9 18 l1 [j] 1
f2 [j] 12 16 l2 [j] 1

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

line 1 S1,1 S1,2 S1,3 S1,4 S1,5 S1,6 exit

e1 = 2 → 7→ 9→ 3→ 4→ 8→ 4→ x1 = 3
&2 &3 &1 &3 &4

2% 1% 2% 2% 1%
e2 = 4 → 8→ 5→ 6→ 4→ 5→ 7→ x2 = 2
line 2 S2,1 S2,2 S2,3 S2,4 S2,5 S2,6 exit

j 1 2 3 4 5 6 j 2 3 4 5 6
f1 [j] 9 18 20 l1 [j] 1 2
f2 [j] 12 16 22 l2 [j] 1 2

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

line 1 S1,1 S1,2 S1,3 S1,4 S1,5 S1,6 exit

e1 = 2 → 7→ 9→ 3→ 4→ 8→ 4→ x1 = 3
&2 &3 &1 &3 &4

2% 1% 2% 2% 1%
e2 = 4 → 8→ 5→ 6→ 4→ 5→ 7→ x2 = 2
line 2 S2,1 S2,2 S2,3 S2,4 S2,5 S2,6 exit

j 1 2 3 4 5 6 j 2 3 4 5 6
f1 [j] 9 18 20 24 l1 [j] 1 2 1
f2 [j] 12 16 22 25 l2 [j] 1 2 1

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

line 1 S1,1 S1,2 S1,3 S1,4 S1,5 S1,6 exit

e1 = 2 → 7→ 9→ 3→ 4→ 8→ 4→ x1 = 3
&2 &3 &1 &3 &4

2% 1% 2% 2% 1%
e2 = 4 → 8→ 5→ 6→ 4→ 5→ 7→ x2 = 2
line 2 S2,1 S2,2 S2,3 S2,4 S2,5 S2,6 exit

j 1 2 3 4 5 6 j 2 3 4 5 6
f1 [j] 9 18 20 24 32 l1 [j] 1 2 1 1
f2 [j] 12 16 22 25 30 l2 [j] 1 2 1 2

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

line 1 S1,1 S1,2 S1,3 S1,4 S1,5 S1,6 exit

e1 = 2 → 7→ 9→ 3→ 4→ 8→ 4→ x1 = 3
&2 &3 &1 &3 &4

2% 1% 2% 2% 1%
e2 = 4 → 8→ 5→ 6→ 4→ 5→ 7→ x2 = 2
line 2 S2,1 S2,2 S2,3 S2,4 S2,5 S2,6 exit

j 1 2 3 4 5 6 j 2 3 4 5 6
f1 [j] 9 18 20 24 32 35 l1 [j] 1 2 1 1 2
f2 [j] 12 16 22 25 30 37 l2 [j] 1 2 1 2 2

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

line 1 S1,1 S1,2 S1,3 S1,4 S1,5 S1,6 exit

e1 = 2 → 7→ 9→ 3→ 4→ 8→ 4→ x1 = 3
&2 &3 &1 &3 &4

2% 1% 2% 2% 1%
e2 = 4 → 8→ 5→ 6→ 4→ 5→ 7→ x2 = 2
line 2 S2,1 S2,2 S2,3 S2,4 S2,5 S2,6 exit

j 1 2 3 4 5 6 j 2 3 4 5 6
f1 [j] 9 18 20 24 32 35 l1 [j] 1 2 1 1 2
f2 [j] 12 16 22 25 30 37 l2 [j] 1 2 1 2 2
I f ∗ = 38 and l ∗ = 1

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

line 1 S1,1 S1,2 S1,3 S1,4 S1,5 S1,6 exit

e1 = 2 → 7→ 9→ 3→ 4→ 8→ 4→ x1 = 3
&2 &3 &1 &3 &4

2% 1% 2% 2% 1%
e2 = 4 → 8→ 5→ 6→ 4→ 5→ 7→ x2 = 2
line 2 S2,1 S2,2 S2,3 S2,4 S2,5 S2,6 exit

j 1 2 3 4 5 6 j 2 3 4 5 6
f1 [j] 9 18 20 24 32 35 l1 [j] 1 2 1 1 2
f2 [j] 12 16 22 25 30 37 l2 [j] 1 2 1 2 2
I f ∗ = 38 and l ∗ = 1

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

fastestway(a, t, e, x, n)

1. f1 [1] ← e1 + a1,1 and f2 [1] ← e2 + a2,1

2. for j ← 2 to n
3. do if f1 [j − 1] + a1,j ≤ f2 [j − 1] + t2,j−1 + a1,j
4. then f1 [j] ← f1 [j − 1] + a1,j l1 [j] ← 1
5. else f1 [j] ← f2 [j − 1] + t2,j−1 + a1,j l1 [j] ← 2
6. if f2 [j − 1] + a2,j ≤ f1 [j − 1] + t1,j−1 + a2,j
7. then f2 [j] ← f2 [j − 1] + a2,j l2 [j] ← 2
8. else f2 [j] ← f1 [j − 1] + t1,j−1 + a2,j l2 [j] ← 1
9. if f1 [n] + x1 ≤ f2 [n] + x2
10. then f ∗ = f1 [n] + x1 l ∗ = 1
11. else f ∗ = f2 [n] + x2 l ∗ = 2

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design

Assembly-line scheduling

I entire procedure take Θ(n) time

I compute fi [j] and li [j] - fill in table entries
I to compute fi [j] we need f1 [j − 1] and f2 [j − 1]
I Construct fastest way
1. i ← l ∗
2. print ”line” i, ”station” n
3. for j ← n downto 2
4. do i ← li [j]
5. print ”line” i, ”station” j − 1

M. A. Zaveri, SVNIT, Surat Algorithm Analysis and Design