eoanodea / daddd-assignment

This project is a set covering problem for unmanned aerial vehicle (UAV) to cover a set of targets. The problem is formulated as a mixed integer linear programming (MILP) model and solved by Gurobi solver.

🛰️ Unmanned aerial vehicle set covering problem

✏️ Introduction

Individual assignment for the Data Analytics & Data Driven Decisions course.

This project is a set covering problem for unmanned aerial vehicle (UAV) to cover a set of targets. The problem is formulated as a mixed integer linear programming (MILP) model and solved by Gurobi solver.

Based on the Paper Unmanned aerial vehicle set covering problem considering fixed-radius coverage constraint, the model is extended to only consider the fixed-radius coverage constraint.

The idea behind this problem is to asssit in a region during disaster recovery. Where demand_points are people in need of aid within a certain region, how many UAVs are required to cover all demand points, based on the radius of the UAVs (coverage_radius).

📦 Installation

The following packages are required to run the code:

• Gurobi (you also need a licence)
• NetworkX
• Matplotlib

You should also have some sort of Python environment manager installed, such as Anaconda.

📚 How it works

The scp.ipynb file manages all aspects of solving this problem. It starts by generating a set of demand points within a networkx graph, based on the specified regiion width & height, and the number of demand points.

Each demand point is plotted randomly on the graph, the output looks something like this:

Secondly, we define decision variables, the objective function (to minimize the number of UAVs required), and the constraints, based on the model in the paper:

Sets

• $N$ = set of demand points
• $M$ = set of flight position of UAVs

Parameters

• $b^x_j$ = flight position of UAV $j$ on x-coordinate
• $b^y_j$ = flight position of UAV $j$ on y-coordinate
• $\alpha_{ij}$ = binary feasibility of UAV $j$ covering demand point $i$

Decision Variables

$$ y_j = \begin{cases} 1,\quad \text{ if UAV $j$ is used} \\ 0,\quad \text{otherwise} \end{cases} $$

$$ x_{ij} = \begin{cases} 1,\quad \text{ if demand point $i$ is covered by UAV $j$} \\ 0,\quad \text{otherwise} \end{cases} $$

$$ \forall i \in N \ \forall j \in N$$

Objective function

$$ \begin{alignat}{3} & \min \sum_{j \in M}y_j \\ & \text{s.t.} x_{ij} \le y_j \quad && \forall i \in N, \forall j \in M \\ & \sum_{j \in M} \alpha_{ij}x_{ij} \ge 1 \quad && \forall i \in N \\ & x_{ij} \in {0, 1} \quad && \forall i \in N, \forall j \in M \\ & y_j \in {0, 1} \quad && \forall j \in M\end{alignat} $$

Distance

Distance between each demand point is calculated using the Euclidean distance formula: $$\sqrt{(a^x_i - b^x_j)^2 + (a^y_i - b^y_j)^2}$$

$\alpha = 1$ if distance is less than or equal to the coverage radius, $0$ otherwise.

Finally, the model is solved using Gurobi solver, and the output looks something like this:

In this particiular case, 5 UAVs are required to cover all demand points.

Note: You may notice the number of UAVs flucuate between each run. This is because the demand points are randomly generated within the region every time, and the solver may find a different solution each time.