SUM NSM Dynamic Pricing for Demand Nudging — Demo

A demo of the dynamic-pricing model developed for the SUM project. The model and methodology are described in CSI_Report_D1_Mesut_Can_Koseoglu.pdf.

The core deliverable is the sum_nsm_dynamic_pricing Python package: it generates optimization-ready mobility instances from GTFS and GBFS data, runs OpenTripPlanner-based routing, solves the bilevel dynamic-pricing optimization model, and exports summary metrics.

What it does

1. Build station inputs from GTFS and GBFS (stations.json).
2. Build multimodal hubs with clustering (hubs.geojson).
3. Query OpenTripPlanner (OTP) for OD routes (od_routes.json).
4. Build the optimization instance file (Instance_*.json).
5. Run the bilevel optimization (optimization_instance_results.json).
6. Compute summary metrics (optimization_metrics.json).

If OTP is unavailable, route generation and the downstream steps are blocked.

Repository layout

Path	Description
sum_nsm_dynamic_pricing/	The Python package — the core deliverable.
demo_sum_nsm_dynamic_pricing.ipynb	Demo notebook: optimization and metrics on a generated instance.
data/sum_nsm_dynamic_pricing/geneva_1km_mon08h/	Sample Geneva inputs and outputs (stations, hubs, instance, results, metrics, web_demo/).
CSI_Report_D1_Mesut_Can_Koseoglu.pdf	Model document.

Requirements

• Python 3.10+
• Gurobi with an active license (gurobipy is a required dependency)
• OpenTripPlanner running and reachable from your machine

Install

From the repository root:

cd sum_nsm_dynamic_pricing
python -m pip install --upgrade pip
python -m pip install -e .

This installs the package in editable mode for local development.

Quick run

Run the example pipeline script (sum_nsm_dynamic_pricing/examples/geneva_pipeline.py):

python sum_nsm_dynamic_pricing/examples/geneva_pipeline.py

The script runs instance generation, optimization, and metrics export. It expects a GTFS zip at ./tpg_gtfs.zip and writes outputs to ./output_geneva.

For an interactive walkthrough of the optimization and metrics step, open demo_sum_nsm_dynamic_pricing.ipynb.

Minimal API example

from pathlib import Path

from sum_nsm_dynamic_pricing import (
    DonkeyRepublicProvider,
    InstanceParams,
    InstancePipeline,
    OptimizerParams,
    PipelineConfig,
)

config = PipelineConfig(
    center_lat=46.2044,
    center_lon=6.1432,
    radius_m=2000,
    output_dir=Path("./output_geneva"),
    gtfs_path=Path("./tpg_gtfs.zip"),
    gbfs_path=None,
    seed=42,
)

pipeline = InstancePipeline(config=config, gbfs_provider=DonkeyRepublicProvider())
params = InstanceParams()

instance_path = pipeline.run_all(params)
solution_path = pipeline.optimize(OptimizerParams(), instance_path=instance_path)
metrics_path = pipeline.summarize_optimization(solution_path, params)

A one-call run_pipeline(config, params) helper is also exported for the instance-generation step.

Core outputs

By default, outputs are written to the configured output_dir:

• stations.json
• hubs.geojson
• od_routes.json
• Instance_*.json
• optimization_instance_results.json
• optimization_metrics.json
• gurobi/ (solver logs and model artifacts)

A complete example output set is included under data/sum_nsm_dynamic_pricing/geneva_1km_mon08h/.

Build the package

From the repository root:

cd sum_nsm_dynamic_pricing
python -m pip install --upgrade build
python -m build

Build artifacts are generated in sum_nsm_dynamic_pricing/dist/. Install the wheel locally with:

python -m pip install dist/*.whl