Optimization in Control and Robotics

Lab 2. Discrete Optimization

Activity 1
The code below is a simple implementation of the Dijkstra()algorithm.

function [P, di] = Dijkstra(G, s, d)

% This is an implementation of the Dijkstra´s algorithm, which finds the minimal
% cost path between two nodes.

% the inputs of the algorithm are:

% G: the graph object

% s: source node index (s in (1:n))
% d: destination node index (s in (1:n))

% the outputs of the algorithm are:

% P: the path starting at s and ending at d

% di: the minimal distance between s and d in the graph G

n=size(G.Nodes,1); % # of nodes
S=s; % S is the set of analyzed nodes.
C=setdiff(1:n,s); % C is the rest of nodes
dist(1:n)=inf; % it stores the shortest distance between s and i
prev(1:n)=nan; % it stores the previous node in the shortest path
dist(s)=0;

while S(end)~=d
candidate=neighbors(G,S(end)); % candidate if connected with the last node entered in S
candidate=setdiff(candidate,S); % candidate if not already in S
for i=candidate'
distp=dist(S(end))+G.Edges.Weight(findedge(G,S(end),i));
if distp<dist(i)
dist(i)=distp; prev(i)=S(end);
end
end
[a,b]=min(dist(C));
S=[S,C(b)];
C=setdiff(1:n,S);
end

% build the path

P=d;
while P(end)~=s
P=[P,prev(P(end))];
end
P=fliplr(P);
di=dist(d);
end

• Apply it to calculate the shortest path between the two points in the maze of the figure.
 To do this, it is necessary to convert the maze into a graph and apply the Dijkstra
algorithm. We propose two ways to do this: Break down the maze into square cells and:

1. Any cell is a node that can be connected to at most 4 nodes (cells). Two nodes
are connected by an edge if there is not a wall between them. The weights
of all edges are 1.

2. Any cell surrounded by at most one wall is a node. In these cells there is more
than one possible path (crosspaths). The edges in this graph no longer always
have weight equal 1, but it is the minimum distance of a path in the maze
connecting the two nodes that does not pass through another node.

In ATENEA you can find the edges for the first approach (Graph1.txt) and the edges and
weights for the second approach (Graph2.txt). Apply the Dijkstra algorithm using both
approaches and compare the results.

• Compare the previous result with the one obtained using function
shortestpath()of MATLAB.
• Translate the shortest path problem using the binary linear programming formulation
and solve it using function intlinprog() of MATLAB. Compare the results with the
previous methods.
Activity 2
Let’s consider 15 of the most important cities in US. The table below contains the distance by
road between all pairs of these cities. The data of this table is in the file distances.txt

Atlanta

Boston

Chicago

Dallas

Denver

Houston

Las Vegas

Los Angeles

Miami

New Orleans

New York

Phoenix

Francisco
San

Seattle

D.C.
Washington
Atlanta 0 1095 715 805 1437 844 1920 2230 675 499 884 1832 2537 2730 657
Boston 1095 0 983 1815 1991 1886 2500 3036 1539 1541 213 2664 3179 3043 440
Chicago 715 983 0 931 1050 1092 1500 2112 1390 947 840 1729 2212 2052 695
Dallas 805 1815 931 0 801 242 1150 1425 1332 504 1604 1027 1765 2122 1372
Denver 1437 1991 1050 801 0 1032 885 1174 2094 1305 1780 836 1266 1373 1635
Houston 844 1886 1092 242 1032 0 1525 1556 1237 365 1675 1158 1958 2348 1443
Las Vegas 1920 2500 1500 1150 885 1525 0 289 2640 1805 2486 294 573 1188 2568
Los Angeles 2230 3036 2112 1425 1174 1556 289 0 2757 1921 2825 398 403 1150 2680
Miami 675 1539 1390 1332 2094 1237 2640 2757 0 892 1328 2359 3097 3389 1101
New Orleans 499 1541 947 504 1305 365 1805 1921 892 0 1330 1523 2269 2626 1098
New York 884 213 840 1604 1780 1675 2486 2825 1328 1330 0 2442 3036 2900 229
Phoenix 1832 2664 1729 1027 836 1158 294 398 2359 1523 2442 0 800 1482 2278
San Francisco 2537 3179 2212 1765 1266 1958 573 403 3097 2269 3036 800 0 817 2864
Seattle 2730 3043 2052 2122 1373 2348 1188 1150 3389 2626 2900 1482 817 0 2755
Washington D.C. 657 440 695 1372 1635 1443 2568 2680 1101 1098 229 2278 2864 2755 0

We want to select a set of scheduled services among these 15 cities so that we can go from any
city to any other city using some of these services (probably making more than one transfer
through other cities).
A scheduled service consists of a road transport from one city to another. Always the return
service exists, that means, for instance, if there is a service from San Francisco to Los Angeles,
there exist also the service from Los Angeles to San Francisco. That is equivalent to consider
the graph as an undirected graph.
Which is the selection that minimize the sum of distances of all services?
• Solve the problem by implementing the Prim algorithm.
• Solve the problem using the function minspantree() of MATLAB;

The code below is a simple implementation of the prim()algorithm.

function [mcost,E] = prim(D)

[n,n] = size(D); % The matrix is n by n, where n = # nodes.
C = [1]; nC=1; % C= nodes selected. nC= #C
E = []; nE=0 % k is the number of edges selected
S = [2:n]';nS=n-1; % S= nodes not selected. nS=#S
mcost=0;
while nC < n
mincost = Inf;
for i=1:nC
for j=1:nS
ni = C(i); nj = S(j);
if D(ni,nj) < mincost
mincost = D(ni,nj);
u = ni; v = nj; % Save nodes and edge selected
end
end
end
mcost=mcost+mincost;
E=[E;[u,v]]; eE=nE+1; % Add this edge to tree.
C=[C;v];nC=nC+1;
S=setdiff(S,v); nS=nS-1;
end
end

Activity 3
Transportation companies face daily problems in logistics. Let’s consider the following ‘simple’
logistics problem:

An airline cargo company has an airplane that flies from the UK to Spain on a daily basis
to transport some cargo. Before the flight, it receives bids for deliveries from (many)
customers. Each bid contains the weight of the cargo item to be delivered, the amount
costumers are willing to pay and some other observations. The airline is constrained by
the total amount of weight the plane is allowed to carry. The company must choose a
subset of the packages (bids) to carry on the plane in order to maximize the total profit,
taking into account the weight limit that they must respect and all other constraints.

Bid Weight total Observations

(tons) price
1 2 200 € This bid is compulsory.
2 15 500 € Bids 2 and 9 cannot be selected together.

3 15 600 € Bid 3 can only be chosen if both bids 2 and 6 are

selected together.
 4 42 1000 € Bids 4 and 5 must be chosen together, or both
or none.
5 15 300 €
6 10 600 € It is compulsory to choose exactly two bids out
of these three bids: 6 7 and 8.
7 20 800 €
8 25 800 €
9 5 300 € Bids 2 and 9 cannot be selected together and bid
9 can only be chosen if bid 8 is already selected.
10 16 600 €
11 15 600 € It has to choose at least one of these two bids.
12 23 1000 €
If the airplane capacity is 𝑊 = 100 tons, which bids have to be chosen in order to
maximize the profit?

Activity 4
Solve the following instance of the minimum cost flow, that is, calculate the flow at each of the
20 directed edges to supply goods from supply nodes 1 and 3 to demand nodes 2 and 4
satisfying the capacity constraints and that minimize the total cost.
n=9; % # nodes
m=20; % # edges
% 1 1 1 1 1 1 1 1 1 1 2
% 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 edge
S=[1 3 3 3 5 5 5 6 6 6 7 7 7 7 7 8 8 9 9 9]; % starting node
E=[5 7 8 9 4 6 7 2 5 7 2 4 5 6 9 7 9 2 6 8]; % ending node
C=[3 6 4 5 9 5 4 6 6 4 5 4 7 3 5 5 3 5 2 5]; % cost
W=[9 4 3 2 4 2 5 4 5 4 5 4 5 2 4 5 4 3 5 3]; % capacity
% 1 2 3 4 5 6 7 8 9 node
B=[9 -7 6 -8 0 0 0 0 0 ]; % supplies and demands

a) Compute using as short a code as possible the parameters of the optimization

problem (objective function, matrix and RHS of linear constraints and lower and
upper bounds).

b) Check that it is not necessary to use integer constraints, that is, compare the solution
using integer and real variables.
Activity 5
Use binary integer programming to solve the classic traveling salesman problem. This problem
involves finding the shortest closed tour through a set of points.

a) Generate a set of 30 random points inside the square of opposite vertices (0,0)
and (10,10).

n=30; % number of points

P=zeros(n,2); % inicialize point's coordinates
D=zeros(n,n); % inicialize distance matrix
rectangle('Position',[0 0 10 10])
axis([-1 11 -1 11])
P=rand(n,2)*10; % generate n random points
hold on
plot(P(:,1),P(:,2),'rx') % draw the points as red crosses
for i=1:(n-1) % calculate distance matrix
for j=(i+1):n
D(i,j)=norm(P(i,:)-P(j,:));
D(j,i)=D(i,j);
end
end

b) Consider 𝑛2 binary variables 𝑥𝑖𝑗 and solve the following binary linear problem:
𝑛

min ∑ 𝑑𝑖𝑗 · 𝑥𝑖𝑗

𝑖=1
𝑛

st ∑ 𝑥𝑖𝑗 = 1 𝑖 = 1…𝑛
𝑗
𝑛

st ∑ 𝑥𝑖𝑗 = 1 𝑗 = 1…𝑛
𝑖
c) Add suitable variables and constraint to solve the TSP and depict the solution.