Optimization
Redraw collection routes from today's data, then decide how many trucks you need.
We think a city should redraw its waste collection routes from today's truck and weighbridge data before it decides how many trucks, sensors or sites it needs. Waste has grown faster than forecast, so routes drawn years ago may no longer match where people live. The World Bank's What a Waste 3.0 reports that the world produced 2.56 billion tonnes of municipal waste in 2022, and that the figure could reach 3.86 billion tonnes by 2050 under business as usual. Managing it already costs more than US$250 billion a year. extendfuture would model how waste is generated, collected, sorted and disposed of, so a city can plan routes, recycling and landfill capacity with evidence.
How can a city optimize garbage collection?
Measure how much waste each area produces and how trucks collect it today, then let optimization software plan the routes. We would combine truck GPS records, weighbridge totals and street maps from OpenStreetMap, forecast waste by area and day, and use open-source routing tools such as OR-Tools or VROOM to plan routes within truck capacity and shift limits. AI agents, software that sets up runs and drafts summaries, compare the schedules. Supervisors and crews review the routes before anything changes.
Plan for today's waste, not an old forecast
The 2018 edition of What a Waste did not expect global municipal waste to reach 2.56 billion tonnes until 2030. It reached that level in 2022. Collection still misses many people, and the World Bank puts collection rates as low as 31% in Sub-Saharan Africa and 67% in South Asia. The UN's 2026 SDG report also cites a UNEP estimate of 1.05 billion tonnes of food wasted in shops, food service and homes in 2022.
Start from the truck data you already have
A first model can start from records that the city or its contractors already hold. It would follow waste from the bin through collection, transfer, sorting, recycling, composting, incineration or landfill.
- Streets from OpenStreetMap and building footprints from open datasets
- Population and household data from the national statistics office, to estimate waste by area
- Truck GPS tracks, weighbridge records and bin fill data held by the city or its contractors
- Recycling and composting plant capacity, and the landfill space that remains
- Waste composition surveys that show the share of food, paper, plastic and other material
Measure the saving instead of assuming it
We would estimate waste per household and area, check those estimates against weighbridge totals, and forecast them by day and season with statistical models. Today's routes are rebuilt from truck GPS data and compared with optimized routes on the same days, so the saving is measured, not assumed. Facility scenarios, such as a new transfer station or composting site, are tested for distance, cost and capacity over the years ahead. Each forecast is checked against what happened in a past year before it is used.
Know when the landfill runs out before it does
A city should see the end of its landfill space years ahead, along with the options that delay it. Waste managers and city finance teams would get plans they can test before they adopt them.
- Collection routes and schedules that fit truck capacity, crew shifts and street access
- Areas with no regular collection, with the cost of reaching them
- Options for transfer stations, recycling and composting sites, with distance, cost and capacity for each
- A forecast of landfill space under each scenario, including when it runs out
- A plain-language summary of each option, checked by an analyst
Crews try new routes before they replace the old ones
Route changes affect crews and residents, so new routes run on a sample of days with supervisors before they replace the old ones. The model uses waste totals by area, not the contents of one household's bin. The World Bank counts 18 million urban waste workers worldwide. Plans that change where recyclables go are checked for their effect on waste workers' income, including informal workers. Every run is logged, and city officials sign off every change.
Run the routing on open tools
A city should be able to rerun its own routing model without a license fee. Each tool below is open source or open data.
- OR-Tools, Google's open-source optimization suite, for vehicle routing and crew scheduling
- VROOM, an open-source route optimization engine for vehicle routing problems with capacities and time windows
- OSRM, a routing engine on OpenStreetMap data that computes travel times between all pairs of stops
- OpenStreetMap, for streets and facility locations
- statsforecast, for forecasting waste by area, day and season
- QGIS, for siting transfer stations and recycling centers
Do we need sensors in every bin?
No. A useful model starts from truck GPS tracks, weighbridge totals and population data. Sensors help where fill levels vary a lot, and the model can show where they would change routes enough to justify their cost.
Will optimized routes work on real streets?
They are tested first. Crews and supervisors know about narrow lanes, market days and access rules that no dataset holds, and their corrections go into the model.
Can it raise recycling rates?
It can test the options. The model compares separate collection, drop-off points and sorting capacity, and shows how much material each would keep out of landfill and at what cost.
Does this work where collection is informal?
Partly. Where informal collection is missing from official records, the model estimates it from household and disposal data and flags the uncertainty. Plans are checked for their effect on informal workers before they are adopted.
How is the saving measured?
Against today's routes on the same days. Truck kilometers, fuel, hours and tonnes collected are compared before and after, so the result is a measured difference rather than a forecast.
Talk to the people who build.
One call. An honest read on what AI can do for this, and the number it has to beat.