Smart Multi-Level Parking Optimization

An operations-research pipeline that assigns arriving vehicles to spots in a multi-level car park to minimize user inconvenience, benchmarking exact methods (MILP, Min-Cost Flow) against Genetic Algorithm and Simulated Annealing metaheuristics.

Project Overview

Developed an operations research system that optimizes vehicle-to-space assignment in multi-level parking facilities using constrained optimization, minimizing walking distance, congestion, and assignment penalties while respecting capacity and compatibility constraints. Compared exact optimization (MILP, Min-Cost Flow) with scalable metaheuristics (Genetic Algorithm, Simulated Annealing) on realistic simulated datasets, evaluating solution quality, optimality.

Key Features

MILP formulation and Min-Cost Flow (successive shortest-path) exact optimal baselines
Genetic Algorithm and Simulated Annealing metaheuristics sharing one encoding, objective, and a greedy warm-start
Constraint modeling for level capacity, vehicle-size and EV compatibility, ramp and elevator access, and EV charging
Synthetic multi-level infrastructure generator (layouts plus walking-distance matrices) over the Kaggle IIoT dataset
Benchmark suite comparing objective value, optimality gap, and runtime across instances
Discrete-event simulation of occupancy dynamics: level density, utilization, and queueing