PROJECT 09Optimization Algorithms & HeuristicsEvolutionary Heuristics

VRP SOLVER

Evolutionary heuristics benchmark comparing Genetic Algorithms and Differential Evolution on CVRP.

Source Repository

VRP SOLVER // EVOLUTIONARY HEURISTICS & REAL-TIME ROUTE PLAYBACK

GENETIC ALGORITHM VS. DE

COMPONENT 01

Problem Formulation

Defines customer demand nodes, fleet limits, and Euclidean travel cost matrices.

COMPONENT 02

PMX Crossover Engine

Guarantees valid routing permutations without duplicate customer node visits.

COMPONENT 03

Differential Evolution

Applies continuous vector perturbation adapted to discrete permutation spaces.

COMPONENT 04

Tkinter Route Canvas

Animates fleet trajectories in real time while plotting fitness convergence curves.

TECH STACK • CORE RUNTIME

PythonGenetic AlgorithmsDifferential EvolutionTkinterNumPyMatplotlib

DOMAIN DISCIPLINE

Algorithms & Optimization

Author: Martin Wassim

VERIFIED INTERFACE ARTIFACTS & ARCHITECTURAL DIAGRAMS

2 visual assets

Figure 1: VRP SOLVER — Visual EvidenceVerified Production Asset

Figure 2: VRP SOLVER — Visual EvidenceVerified Production Asset

01 / THE PROBLEM BEING SOLVED

Operational & Technical Friction

Logistics and fleet operations must minimize total vehicle travel distance while strictly obeying truck payload capacities and customer delivery demands. Exact mathematical solvers scale exponentially, making evolutionary heuristics essential.

02 / THE ARCHITECTED SOLUTION

Systemized Implementation

Engineered custom Genetic Algorithm (GA) and Differential Evolution (DE) optimization engines in Python with specialized permutation chromosome representations, repair operators, and animated convergence plotting.

03 / ACTUAL IMPLEMENTATION & WORKFLOW

How It Operates in Production

Built with Python and NumPy. Formulated specialized partially matched crossover (PMX) operators to maintain valid tour permutations, penalty fitness functions for capacity violations, and a real-time Tkinter vector canvas illustrating route evolution.

Conducted empirical benchmarks across problem instances of varying complexity (20 to 100 nodes), quantifying generational convergence speed, solution stability, and premature stagnation risks.

04 / SYSTEM ARCHITECTURE EXPLANATION

Data Flow & Infrastructure

Problem parameters (depot coords, customer demands, vehicle capacities) feed into the evolutionary engine. Populations undergo iterative selection, crossover, and mutation. The best chromosome per generation updates the Tkinter canvas and logs convergence metrics.

Architectural Highlight: Benchmarked specialized chromosome representations, PMX crossover, and continuous-to-discrete differential vector adaptations across 20 to 100 customer nodes.

05 / KEY ENGINEERING DECISIONS

Partially Matched Crossover (PMX): Guarantees offspring chromosome validity without producing duplicate customer visits.

Dynamic penalty fitness functions: Softly penalizes vehicles exceeding maximum capacity constraints, guiding search back into feasible search space.

Real-time visual convergence canvas: Rendered real-time route animations to give immediate spatial feedback on route partitioning behavior.

06 / CHALLENGES & MITIGATIONS

•Adapting continuous Differential Evolution mutation vectors to discrete permutation-based routing chromosomes.

•Avoiding premature convergence in local minima on highly clustered customer distributions.

07 / VERIFIED CAPABILITIES

Capacity-Constrained Vehicle Routing Problem (CVRP) mathematical formulation

Genetic Algorithm engine with tournament selection, PMX crossover, and inversion mutation

Differential Evolution engine with adaptive mutation vectors adapted to discrete permutation spaces

Interactive Tkinter visualization canvas rendering vehicle routes and customer clusters in real time

Generational fitness convergence plotting comparing solution distance and vehicle count

Customer demand matrix generator with adjustable fleet capacity limits

Return to Project Index Inquire About Similar Architecture