RouteSense

Step 1

Create Delivery Network

Build your delivery network by clicking the map or auto-generating stops on real roads.

Interactive Map

Click on the map to add delivery stops. Stops are automatically placed on accessible roads. The first point will be the Depot.

Auto-Generate Network

Stops will be automatically placed on accessible roads. Large networks (>20 stops) may take a moment to generate.

Number of Stops:

Generation Method:

Current Network

No stops created yet. Click on the map or generate a network.

Step 2

Configure & Run Simulation

Set parameters and choose a routing algorithm, then run the simulation.

Time Configuration

Start Time:

Service Time per Stop (minutes):

Time Window Width (minutes): Creates time windows around expected arrival times

Traffic Configuration

Traffic Intensity:

Base Speed (km/h):

Fuel Rate (cost per km):

Cost Parameters

Alpha (travel time weight):

Beta (fuel cost weight):

Lambda (late penalty weight):

Optimisation Settings

Algorithm:

Max Iterations:

Simulation Mode:

Analysis

Compare Algorithms

Run multiple algorithms side-by-side on the same network and parameters.

Select Algorithms to Compare

All algorithms will run on the same network with identical parameters.

Greedy Nearest Neighbour Time Window Greedy Nearest Insertion Hill Climbing Simulated Annealing Adaptive Simulated Annealing Q-Learning (MDP)

Evaluation

Performance Evaluation

Reproduce the term-project write-up's experiments directly in the browser. Generates small (4-stop) and medium (6-stop) scenarios, runs each algorithm, and renders tables + charts. Export everything as a formatted Excel workbook.

Run Experiments

Runs the full battery: baseline comparison, Q-Learning learning-rate sensitivity, seed stability (3 identical runs), and traffic sensitivity. Network generation uses the site's grid generator and is rate-limited by Nominatim (~1 s per stop), so the full run takes a minute or two.

No run yet.

Table II — Baseline Comparison

Total cost and late-stop counts for each algorithm across both scenarios, with percentage deltas against Nearest Neighbor and Time-Window Greedy baselines.

Method	Small Total Cost	Medium Total Cost	Small Late Stops	Medium Late Stops	vs NN %	vs TWG %
Run the experiments to populate this table.

Table I — Q-Learning Learning-Rate Sensitivity

The Q-Learning learning rate is hardcoded to 0.1 in app.js. This experiment varies max_iterations (500 / 1000 / 2000) as the closest exposed proxy.

Alpha (LR proxy)	Max Iterations	Final Route Cost	Travel Time (min)	Late Stops	Route
Run the experiments to populate this table.

Seed Stability

Three identical Q-Learning runs on the small scenario. Identical costs + identical routes indicate deterministic behaviour under the current seed settings.

Run	Total Cost	Travel Time (min)	Fuel Cost	Late Stops	Route Order
Run the experiments to populate this table.

Traffic Sensitivity

Total cost and late-stop counts for each algorithm under Low / Medium / High traffic intensities.

Scenario	Algorithm	Traffic	Total Cost	Late Stops
Run the experiments to populate this table.

Output

Simulation Results

View key metrics, the optimised route, and the full execution log.

Total Cost

Travel Time

Fuel Cost

Late Stops

Route Details

Run a simulation to see route details here.

Route Visualisation

Execution Log

Simulation logs will appear here.