APPLIED ALGORITHMS

CONTENTS

• General diagram of backtracking, branch and bound

• The problem of bus routes picking up and dropping off passengers
• Delivery truck route problem
APPLIED ALGORITHMS
BACKTRACKING, BRANCH AND BOUND

3 4
 GENERAL DIAGRAM OF BACKTRACKING, BRANCHING AND BOUND GENERAL DIAGRAM OF BACKTRACKING, BRANCHING AND BOUND
try(1) Enumeration problem
• The backtracking algorithm allows us to solve combinatorial enumeration problems and 1
combinatorial optimization problems try(k){ //Try out the possible values assigned to Xk
• The alternative is modeled by a sequence of decision variables X1, X2, . . ., Xn try(2)
try(2) try(2) for v in Ak do {

• Need to find for each variable Xi a value selected from a given discrete set Ai such that 2 3 . . . if check(v,k){

• The constraints of the problem are satisfied Xk = v;

. . . . . . . . . . . . . . .
• Optimize a given objective function . . . [Update a data structure D]

• Backtracking algorithm try(k) if k = n then solution();

else {
• Traverse through all variables (e.g. order from X1, X2, . . ., Xn), for each variable Xk : . . .
try(k+1);
• Traverse through all possible values that could be assigned to Xk, for each value v:
}
• Check constraints . . . . . . . . . . . . . . .
[Recover the data structure D]
• Assign Xk = v
}
• If k = n then record a solution to the problem
}
• Otherwise, consider the variable Xk+1
}

5 6

GENERAL DIAGRAM OF BACKTRACKING, BRANCHING AND BOUND GENERAL DIAGRAM OF BACKTRACKING, BRANCHING AND BOUND
try(1) Minimize optimization problem (Denote f* : optimal value) try(1) Minimize optimization problem (Denote f* : optimal value)
1 1
try(k){//Try out the possible values assigned to Xk try(k){//Try out the possible values assigned to Xk
try(2) try(2) for v in Ak do { try(2) try(2) for v in Ak do {
try(2) try(2)
2 3 if check(v,k){ 2 3 if check(v,k){
. . . . . .
Xk = v; Xk = v;
. . . . . . . . . . . . . . . . . . [Update a data structure D] . . . . . . . . . . . . . . . . . . [Update a data structure D]

try(k) if k = n then updateBest(); try(k) if k = n then updateBest();

else { else {
. . . . . .
if g(X1, X2, …, Xk) < f* then if g(X1, X2, …, Xk) < f* then
try(k+1); try(k+1);
. . . . . . . . . . . . . . . Calculate the lower . . . . . . . . . . . . . . . Do not further build
} }
bound the solution if
g(X1, X2, …, Xk) of [Recover the data structure D] [Recover the data structure D]
the objective
g(X1, X2, …, Xk)
} ≥ f* }
function for
solutions that are } }
being continued to } }
be built from here

7 8
DELIVERY TRUCK ROUTE PROBLEM DELIVERY TRUCK ROUTE PROBLEM

• A fleet of K identical trucks needs to be assigned to transport pepsi boxes from the central • Example N = 3, K = 2, Q = 10, d[1] = 3, d[2] = 2, d[3] = 1
warehouse (point 0) to delivery points 1,2,...,N and back to the warehouse. The travel distance from  There are 6 shipping options as below:
point i to point j is c(i,j)
• Each truck has a load capacity of Q (each trip can only transport a maximum of Q boxes)
Route[1] = 0 – 1 – 0 Route[1] = 0 – 1 – 2 – 0
• Each delivery point i has a required quantity of d[i] boxes, i = 1,…, N. Route[2] = 0 – 2 – 3 – 0 Route[2] = 0 – 3 – 0
• It is necessary to develop a transportation plan so that: Route[1] = 0 – 1 – 3 – 0 Route[1] = 0 – 2 – 0
• Each vehicle must be assigned Route[2] = 0 – 2 – 0 Route[2] = 0 – 3 – 1 – 0
• Each delivery point can only be delivered by exactly one vehicle and only once. Route[1] = 0 – 1 – 0 Route[1] = 0 – 2 – 1 – 0
• The total amount of boxes on the vehicle must not exceed the vehicle's load. Route[2] = 0 – 3 – 2 – 0 Route[2] = 0 – 3 – 0
• The total route length of the vehicles is the smallest
• Note: do not need to use all K vehicles.

9 10

DELIVERY TRUCK ROUTE PROBLEM DELIVERY TRUCK ROUTE PROBLEM

• Example N = 3, K = 2, Q = 4, d[1] = 3, d[2] = 2, d[3] = 1 • Design data

– y[k] first delivery point of the kth vehicle (y[k] {0, 1, 2,..., N}, where k = 1, 2,…, K)
 There are 4 shipping options as below:
• y[k] = 0 means the vehicle k will not be used for route planning
– x[i] is the next point of the delivery point i on the route (x[i] {0,1,2,…,N}, với i = 1, 2,..., N)
Route[1] = 0 – 1 – 0 Route[1] = 0 – 1 – 2 – 0
Route[2] = 0 – 2 – 3 – 0 Route[2] = 0 – 3 – 0 – Since the vehilces are identical, we can assume y[k] ≤ y[k+1], k = 1, 2,…, K-1
Route[1] = 0 – 1 – 3 – 0 Route[1] = 0 – 2 – 0 • If y[k] > 0 then y[k+1] > y[k]
Route[2] = 0 – 2 – 0 Route[2] = 0 – 3 – 1 – 0 – Variables associated with the partial solution:
Route[1] = 0 – 1 – 0 Route[1] = 0 – 2 – 1 – 0 • visited[v] = true if v has been visited by a vehicle
Route[2] = 0 – 3 – 2 – 0 Route[2] = 0 – 3 – 0 • load[k]: total amount of boxes on the vehicle k 1 2 5

• f: total length of the current partial solution

• f*: best distance that has been found 0 4 3 0

6 8 7

11 12
 DELIVERY TRUCK ROUTE PROBLEM DELIVERY TRUCK ROUTE PROBLEM

• Strategy to traverse • K = 2, N= 6
– Start by traversing values for the tuple (y[1],. . ., y[K])
– For each complete values of tuple (y[1],. . ., y[K]), start traversing values for the x[1,...,N] derived f
rom x[y[1]] 0
– Each time: try to assign x[v] = u for the kth vehicle:
• If u > 0 (not the starting point): try continuing to browse the value for x[u] still on the kth vehicle
• If u = 0 (starting point) then
(y[1], y[2]) 1
– If k = K (all routes for K vehicles are complete) and all delivery points are visited, then obtain
a solution to the problem
– otherwise, try continuing to assign the values for the route of vehicle (k+1)th, derived from
x[y[k+1]]
– Variable nbR: the number of vehicles that has been used to deliver 1 2 5
– Variable segments
• The number of segments (segment: connection between 2 0 4 3 0
consecutive delivery points on the route)
• when segments = N+nbR then obtain a solution to the problem
6 8 7

13 14

DELIVERY TRUCK ROUTE PROBLEM DELIVERY TRUCK ROUTE PROBLEM

• K = 2, N= 6 • K = 2, N= 6

0 0

(y[1], y[2]) 1 2 (y[1], y[2]) 1 2

15 16
 DELIVERY TRUCK ROUTE PROBLEM DELIVERY TRUCK ROUTE PROBLEM
• K = 2, N= 6 • K = 2, N= 6

0 0

(y[1], y[2]) 1 2 (y[1], y[2]) 1 2

5 5

4 4

17 18

DELIVERY TRUCK ROUTE PROBLEM DELIVERY TRUCK ROUTE PROBLEM

• K = 2, N= 6 • K = 2, N= 6

0 0

(y[1], y[2]) 1 2 (y[1], y[2]) 1 2

5 3 5 3

4 4 6

0 0

19 20
 DELIVERY TRUCK ROUTE PROBLEM DELIVERY TRUCK ROUTE PROBLEM
• K = 2, N= 6 • K = 2, N= 6

0 0

(y[1], y[2]) 1 2 (y[1], y[2]) 1 2

5 3 5 6

4 6 4

0 0 0

21 22

DELIVERY TRUCK ROUTE PROBLEM DELIVERY TRUCK ROUTE PROBLEM

• K = 2, N= 6 • K = 2, N= 6

0 0

(y[1], y[2]) 1 2 (y[1], y[2]) 1 2

5 6 5 6

4 3 4 3

0 0 0

23 24
 DELIVERY TRUCK ROUTE PROBLEM DELIVERY TRUCK ROUTE PROBLEM
• K = 2, N= 6 • K = 2, N= 6

0 0

(y[1], y[2]) 1 2 1 (y[1], y[2]) 1 2 1 3

5 3

4 6

0 0 0 0

25 26

DELIVERY TRUCK ROUTE PROBLEM DELIVERY TRUCK ROUTE PROBLEM

• K = 2, N= 6 TRY_X(s, k){// thử giá trị cho x[s] for v = 0 to n do {
if(s = 0) then{ if checkX(v,k) then {
if k < K then x[s] = v; visited[v] = true; f = f + c[s,v];
TRY_X(y[k+1],k+1);
0 load[k] = load[k] + d[v]; segments = segments + 1;
return if v > 0 then { if f + (n + nbR – segments)*Cmin < f* then TRY_X(v,k); }
} else{
[. . .] if k = K then {
(y[1], y[2]) 1 2 1 3 . . .
} if segments = n + nbR then updateBest();

checkX(v, k){ }else{

4 if v > 0 and visited[v] if f + (n + nbR – segments)*Cmin < f* then TRY_X(y[k+1],k+1);

then return false; }

. if load[k] + d[v] > Q }
. visited[v] = false; f = f – c[s,v];
then return false;
.
return true; load[k] = load[k] – d[v]; segments = segments – 1;
0 0 } }
}

27 28
 DELIVERY TRUCK ROUTE PROBLEM
checkY(v, k){ TRY_Y(k){ // thử giá trị cho y[k]
if v = 0 then return true; s = 0;
if load[k] + d[v] > Q then if y[k-1] > 0 then s = y[k-1] + 1;
return false; for v = s to n do {
if visited[v] = true then if checkY(v,k) then {
return false; y[k] = v;

THANK YOU !
return true; if v > 0 then segments = segments + 1;
} visited[v] = true; f = f + c[0,v]; load[k] = load[k] + d[v];

solve(){ if k < K then TRY_Y(k+1);

f = 0; f* = +; y[0] = 0; else { nbR = segments; TRY_X(y[1],1); }

for v = 1 to n do load[k] = load[k] – d[v]; visited[v] = false; f = f – c[0,v];

visited[v] = false; if v > 0 then segments = segments – 1;

TRY_Y(1); }

output(f*); }

} }

29 30