Forgentics Take-Home Optimization Challenge

This project solves the Forgentics truck-loading problem. The goal is to assign metal plate orders to trucks so that:

• Every order ships today (no splits allowed).
• Each truck has a maximum capacity of 24.0 tons.
• All orders are delivered to the same customer (routing is ignored).
• The total number of trucks is minimized.

The input file is provided as orders.csv (or .xlsx) and the output is loads.csv.

---

Approach

Step 1 – Import and Setup

The program is written in Python 3 and uses:

• pandas for reading/writing CSV/Excel.
• OR-Tools CP-SAT solver for the optimization.
• Built-in utilities (Path, json, etc.) for clean file handling.

Step 2 – Read and Validate Orders

The code reads the orders file and normalizes the data.
It checks that:

• All sales order numbers are unique.
• Weights are positive and no order exceeds 24.0 tons.
  To avoid floating-point errors, weights are multiplied by 10 and converted into integers (so 23.9 tons becomes 239 units).

Step 3 – Heuristic Packing (First Solution)

Before optimization, the code builds a good initial packing:

• First-Fit Decreasing (FFD): orders are sorted from heaviest to lightest and placed in the first truck that fits.
• Best-Fit Decreasing (BFD): tries to minimize wasted space by filling the tightest remaining truck.
• The better of FFD and BFD is kept.
• A “squeeze pass” then attempts to eliminate the last truck by moving its items into earlier trucks.

This heuristic often reaches the minimum number of trucks immediately and also provides a warm start for the solver.

Step 4 – Exact Optimization (CP-SAT)

The packing is then proven optimal using CP-SAT:

• Decision variables assign each order to one truck.
• Constraints ensure:
  • Each order is assigned exactly once.
  • No truck exceeds 24.0 tons.
• Objective is to minimize the number of trucks.
  The heuristic solution is provided as a hint to speed up solving. The solver confirmed that 10 trucks is both feasible and optimal for this dataset.

Step 5 – Output and Validation

The final assignment is written to loads.csv with columns:

• load_id,so_numbers,total_weight_tons
  • 1,"[SO-10001,SO-10012,...]",23.9
  • ...
• so_numbers is a bracketed list with no spaces after commas.
• A console summary is also printed showing each load and weight.
• Automatic checks confirm:
  • Every order is included exactly once.
  • No truck exceeds 24.0 tons.
  • Total weight matches the input.

---

Results

• Minimum trucks used: 10 (proven optimal).
• Output: loads.csv with the exact load assignments.

---

How to Run

1. Open the notebook or script in Google Colab (or any Python 3 environment).
2. Upload orders.csv or orders.xlsx.
3. Run the cells step by step.
4. The program prints the load summary and generates loads.csv in the working directory.

---

Notes

• The code is deterministic (same input -> same output).
• Integer scaling avoids rounding problems in capacity checks.
• The approach combines a practical heuristic for speed and an exact solver for proof of optimality.