VaccineRouting — Full Pipeline with Min‑max, Gini & D‑Index Equity

(vacstep4 project)

A three‑phase Gurobi model that plans an equitable mobile‑vaccination campaign.

Phase	Role
1 – Weekly sizing	Pick stop locations & decide weekly dose totals
2 – Vehicle routing	Build 3‑day routes for each week (2 vehicles)
3 – Integrated plan	Produce one 12‑day, two‑vehicle schedule

The model can enforce three independent equity rules:

Policy	How to enable / disable in Program.cs
Min‑max (≥ 0.7)	keep / comment the line m.AddConstr(E, GRB.GREATER_EQUAL, 0.7, …)
Gini	uncomment the block marked gini (3 constraints)
D‑Index	block marked D‑Index (3 constraints, on by default)

---

Repository layout

.
├── .gitignore
├── README.md
├── vacstep4.sln          # Visual Studio solution
└── vacstep4/             # C# project root
    ├── Program.cs        # all three phases
    ├── vacstep4.csproj
    ├── origindata/       # raw text data (kept in repo)
    │   ├── qit_20*.txt   # instant vaccination rate  q(i,t)
    │   ├── cil_20.txt    # capacity if group i served at location l
    │   ├── djl_20.txt    # travel minutes from origin j to stop l
    │   ├── vit_20*.txt   # value per dose  v(i,t)
    │   └── nis_20.txt    # scenario demand n(i,s)
    └── output/           # **created at runtime** – solutions land here

Note bin/ and obj/ are build artefacts and are ignored by Git.

---

Quick start

# build the solution
dotnet build vacstep4.sln -c Release

# run (≈ 5 min per phase on a laptop)
dotnet run --project vacstep4 -c Release

Results appear in:

• vacstep4/output/phase1_result.txt
• vacstep4/output/week*.txt (one per week, Phase 2)
• vacstep4/output/phase3_result.txt

If a phase stops at its time‑limit you still get the best incumbent solution together with its MIP gap.

---

Tunable parameters (edit Program.cs)

Constant	Meaning	Default
D_BOUND	upper bound on D‑Index (0 = off, 1 = no cap)	0.25
P1_TimeLimit, P2_TimeLimit, P3_TimeLimit	solver limits (s)	300
StopServiceMin	minutes spent at each stop	30
DayBudgetMin	minutes available per vehicle per day	480
Prob[]	scenario probabilities	1/3, 1/3, 1/3
P1_theta, P3_theta	utility weight in the objective	6

Tips

• Lower m.Parameters.MIPGap (Phase 2) or increase the time limits if you need tighter optimality.
• Increase NoRelHeurTime (Phase 1 & 3) when the first feasible solution is slow to appear.

---

Reading the output

Variable	Meaning
Phase 1
W_i_t_s	doses for group i in week t (scenario s)
A_l_t	1 if stop l is chosen in week t
Phase 2 (week k)
y_mjlt	1 if vehicle m travels j → l on day t
W_g_v_t_s	doses for reduced group index g by vehicle v on day t
Phase 3	same names, now for the full 12‑day horizon

---

Common issues

Symptom	Probable cause	Fix
Status = INF_OR_UNBD	equity caps too tight	relax D_BOUND or disable extra equity blocks
First feasible solution slow	heuristics need more time	raise NoRelHeurTime
Day‑budget exceeded	djl_20.txt not in minutes or DayBudgetMin too low	check both

---

Credits

Code developed by Zeynep Aydin, with guidance from

• Doç. Dr. Eda Yücel
• Prof. Sibel Salman
• Betül Kayışağoly

We gratefully acknowledge the Gurobi Optimizer academic license.

---

Feel free to open issues or pull requests if you extend or improve the model.