> ## Documentation Index
> Fetch the complete documentation index at: https://docs.wherobots.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Last Mile Delivery

> How delivery operators, logistics platforms, and autonomous delivery teams can use Wherobots to map pedestrian networks, model service areas, and plan the final leg of the delivery journey.

The final leg of the delivery journey is the most expensive part, and often the least well mapped. While longstanding road networks are well covered, the sidewalks, crosswalks, and building approaches a courier or a delivery robot actually uses are usually missing.

Wherobots can build that missing layer from imagery, join it to address and building data, and model the service areas that your dispatch system plans against.

## How to use these solutions

Treat these notebooks as data engineering patterns as you evaluate Wherobots for your business or organization:

<CardGroup cols={2}>
  <Card title="Pedestrian Network Mapping" icon="person-walking" iconType="solid" href="/tutorials/example-notebooks/rasterflow-tile2net">
    Detect sidewalks, crosswalks, and pedestrian pathways from aerial imagery and turn them into routable geometry for couriers and sidewalk robots.
  </Card>

  <Card title="Delivery Zones & Service Areas" icon="clock" iconType="solid" href="/tutorials/example-notebooks/isochrones">
    Generate walk- and drive-time catchments around depots, dark stores, and lockers to size delivery zones and set realistic promise windows.
  </Card>

  <Card title="Address & Building Context" icon="location-dot" iconType="solid" href="/tutorials/example-notebooks/overture-maps">
    Resolve drop-off points against building footprints and points of interest so a stop lands at the right entrance, not the street centroid.
  </Card>

  <Card title="Road Coverage in Under-Mapped Areas" icon="road" iconType="solid" href="/tutorials/example-notebooks/rasterflow-chesapeake">
    Detect roads from imagery where commercial basemaps are thin, so rural and newly built routes are plannable.
  </Card>
</CardGroup>

## Key Capabilities

<AccordionGroup>
  <Accordion title="Pedestrian network mapping" icon="person-walking">
    Run RasterFlow's Tile2Net model over aerial imagery to detect sidewalks, crosswalks, and pedestrian paths across a whole service area. Vectorize the output into routable geometry, then inspect it for gaps and discontinuities to judge which segments carry a delivery robot or a courier on foot.

    <Card title="Detecting sidewalks with RasterFlow" icon="person-walking" href="/tutorials/example-notebooks/rasterflow-tile2net" horizontal>
      Turn detected pathways into vector geometry you can route on.
    </Card>
  </Accordion>

  <Accordion title="Corridor walkability scoring" icon="store">
    Score a retail corridor or neighborhood on the continuity of its pedestrian network, then join that score to demand and store locations to decide where walking beats driving for the final leg of the delivery journey.

    <Card title="Detecting sidewalks with RasterFlow" icon="person-walking" href="/tutorials/example-notebooks/rasterflow-tile2net" horizontal>
      Map sidewalks and crosswalks, then score the corridor for continuity.
    </Card>
  </Accordion>

  <Accordion title="Delivery zone and service area modeling" icon="clock">
    Generate isochrones from every depot, dark store, and locker to size zones by travel time rather than radius. Recompute them as the network changes to keep promise windows honest.

    <Card title="Making drive time isochrones" icon="clock" href="/tutorials/example-notebooks/isochrones" horizontal>
      Size delivery zones by travel time rather than radius.
    </Card>
  </Accordion>

  <Accordion title="Drop-off points and demand clusters" icon="location-dot">
    Join delivery addresses to building footprints and points of interest so stops resolve to real entrances, and cluster historical orders to find where density justifies a locker or a micro-hub.

    <Columns cols={2}>
      <Card title="Using Overture Maps" icon="map" href="/tutorials/example-notebooks/overture-maps" horizontal>
        Resolve stops against building footprints and places.
      </Card>

      <Card title="DBSCAN clustering" icon="circle-nodes" href="/tutorials/example-notebooks/clustering-dbscan" horizontal>
        Find where order density justifies a locker or micro-hub.
      </Card>
    </Columns>
  </Accordion>

  <Accordion title="Road coverage where basemaps are thin" icon="road">
    Apply RasterFlow road detection to imagery in rural and newly developed areas where commercial basemaps lag, so routes exist for addresses the map does not know about yet.

    <Card title="Detecting roads with RasterFlow" icon="road" href="/tutorials/example-notebooks/rasterflow-chesapeake" horizontal>
      Find roads in areas where commercial basemaps are thin.
    </Card>
  </Accordion>

  <Accordion title="Publishing coverage maps" icon="layer-group">
    Generate vector tiles and PMTiles directly from WherobotsDB so ops teams, partners, and customers can view live coverage, zone boundaries, and pathway networks in any web map, with no tile server to operate.

    <Columns cols={2}>
      <Card title="How to generate PMTiles" icon="layer-group" href="/tutorials/example-notebooks/pmtiles-railroad" horizontal>
        Publish a single-file tile archive of your coverage.
      </Card>

      <Card title="Creating vector tiles with VTiles" icon="map" href="/tutorials/example-notebooks/map-tile-generation" horizontal>
        Render zones and pathways in any web map, no tile server.
      </Card>
    </Columns>
  </Accordion>
</AccordionGroup>

## Wherobots Advantage

| Challenge                                            | How Wherobots Helps                                                          |
| ---------------------------------------------------- | ---------------------------------------------------------------------------- |
| Reliable sidewalk data does not exist for many areas | Detect the pedestrian network from imagery instead of buying new datasets    |
| Routing needs geometry, not pictures                 | Vectorize detections into pathways a routing engine can consume directly     |
| Coverage and zones change constantly                 | Airflow-orchestrated pipelines recompute isochrones and zones on a schedule  |
| City-scale imagery on every refresh                  | RasterFlow processes raster data at planetary scale with built-in mosaicking |

<Card title="Get started with Wherobots" icon="rocket" href="/get-started/wherobots-cloud/start-notebook" cta="Run your first notebook">
  Try Wherobots with your own geospatial data in minutes.
</Card>