Home / Projects / Spatial optimization
Optimization project

CNG stand location optimization using AHP + genetic algorithm

A methodology demonstration that evaluates existing auto-rickshaw stands along a real Sylhet corridor and searches for alternative stand locations under multicriteria and spacing constraints.

Temuki–Bondor Bazar corridorAHP consistency ratio: 0.0292100 generated corridor candidates

Problem statement

CNG auto-rickshaws are a major form of local public transport in Sylhet. This project asks a compact planning question: are three flexible stands along the Temuki–Bondor Bazar corridor positioned reasonably under selected demand, cost, road-width and junction criteria, or do alternative corridor locations score better?

The project is intentionally framed as a computational methodology demonstration rather than a deployment-ready transport recommendation.

Real spatial inputs

9 landmarks + 2 marketsSchools, hospital, bus stand and market anchors
5 corridor path pointsPoints tracing the Temuki–Bondor Bazar route
5 existing standsTwo fixed endpoints + three flexible stands
8 junctionsReal junction coordinates used in proximity scoring
9 cost anchorsHigh / medium / low relative cost tiers
100 candidatesInterpolated points along the corridor path

AHP weighting + constrained GA search

AHP

Weight four criteria

Demand, cost, road width and junction proximity are weighted using pairwise comparisons and the eigenvector method. The reported consistency ratio is 0.0292.

SPACE

Generate candidates

Create 100 candidate locations along the real corridor using Haversine distance and linear interpolation.

SCORE

Evaluate each location

Compute criterion scores from landmark proximity, cost-tier anchors, road-width classes and junction distances.

GA

Select three flexible stands

Optimize locations with hard constraints of at least 400 m spacing between stands and at least 75 m clearance from junctions.

Demand score ─┐ Cost score ────┼── AHP weighted score ───┐ Road width ────┤ ├── Genetic Algorithm ── best 3 feasible locations Junction score ┘ │ constraints: spacing ≥ 400 m; junction clearance ≥ 75 m

Key finding

The GA placed its nearest alternative to Rikabi Bazar only 69 m away, indicating that the existing location is close to the optimized result even though its baseline AHP score was the lowest among the existing flexible stands. In contrast, GA-suggested alternatives for Modina Market and Pathantula were each more than 1.2 km away, indicating potential relocation opportunities within the assumptions of this model.

Interpretation: a low multicriteria score does not automatically imply a bad existing location. The constrained search suggested that Rikabi Bazar may be geographically constrained near its current position, which is a useful distinction between ranking and optimization.

Limitations

Demand and cost are proximity-based proxies rather than measured ridership and land-value data. Road width is simplified into a two-tier segment model, and junction proximity simplifies a more complex safety relationship. Therefore the outputs demonstrate the AHP–GA workflow; they should not be interpreted as an operational transport-planning recommendation without field validation and stronger behavioral/economic data.

View GitHub repository ↗Back to projects