Fewer miles.
Better loads.
Just the fleet you need.
LATTIMEX decides which vehicle serves each delivery, in what order, and how every box is placed inside the vehicle. We implement it in your operation, with your map, your fleet and your loading rules.
From your orders to the route sheet, in three steps
Load your orders
Import a CSV with destinations, sizes and weights, or pull them straight from your ERP, WMS or TMS.
LATTIMEX optimizes
The engine assigns vehicles, sequences stops over the street network and solves the 3D placement of every box. An independent verifier checks the result.
Out on the road
A route sheet per vehicle with stop-by-stop navigation, itemized costs and the loading diagram ready for the dock.
This is what a finished plan looks like
A real Planner run: 300 deliveries over the Querétaro street network, four vehicles and the three-dimensional placement of every box. Walk through the four screens.
We don't sell a license. We implement an operation.
Every fleet is different: interior dimensions, doors, fragile goods, delivery windows and costs. That is why LATTIMEX is delivered as an implementation project, with our team working on your data until the plan is out on the street.
Assessment with your data
We take a sample of your current orders and routes and measure the baseline: miles, vehicles, load factor and planning hours.
Configuration
We prepare your city's street map, the fleet catalog with dimensions and doors, loading rules, and your costs per kilometer and per driver.
Integration
We install the Planner and the engine on your computers and connect them to your ERP, WMS or TMS inside your network. CSV import works from day one.
Go-live and measurement
We train your dispatch team, support the first weeks and compare before and after on your own routes.
You invest in the implementation
We quote each project based on your fleet size, your city and the level of integration you need.
- Your city's street map and fleet configuration
- Planner and engine installed on your infrastructure
- Training and go-live support
- Before-and-after measurement with your orders
The engine is open source
Engine, local server and research are published under the MIT license. If your team can install it, use it for free; if you'd rather have us do it, that is what implementation is for.
See how to install it →What we add on top of the Planner for your operation
These features are not part of the free Planner: they are designed and integrated during implementation, with your providers and your rules.
Reference illustrations.
Install it for free on your computer
The engine, the local server and the map builder are on GitHub. The Planner opens in your browser and works with the engine you run: no accounts, no quotas and no data sent to anyone.
# 1 · Clone the repository $ git clone https://github.com/Erick-MF28/lattimex.git # 2 · Install dependencies $ pip install -r requirements.txt # 3 · Build the engine (C++) $ python -m lattimex build # 4 · Build your city's map $ python -m lattimex map --bbox 20.47,-100.52,20.82,-100.22 --name queretaro # 5 · Start the local server $ python -m lattimex serve # 6 · Open lattimex.com/planner → https://lattimex.com/planner
Use, modify and distribute it, commercially too, keeping the copyright notice.
Free to run your company, study or research. Redistribution or resale requires permission.
Maps are built with OpenStreetMap data, © OpenStreetMap contributors.
Requirements: Python 3.10 or newer; a C++ compiler or pip install ziglang; Chrome, Edge or Firefox.
Is the project useful to you? Donations fund more research and engine maintenance.
♥ Support the projectSoftware you can audit
Everything runs on your infrastructure: the map, your data, the engine and the verifier. Every solution is checked by an independent program and the engine is benchmarked against the state of the art. We also publish the cases where another method wins.
3D loading of 300 boxes
Four vehicles, at least 75 % of each box's base supported and zero overlaps in the run shown on this page.
3D packer in C++
We rewrote the packer in C++ and checked that it matches the Python version in 72 out of 72 test cases.
Average gap to optimal
On 60 A, B and E benchmark instances with 2 s of compute. HGS, the state-of-the-art reference, reaches 0.004 %. Read the analysis
Verifier conflicts
An independent program re-checks distances, capacity, coverage and the geometry of every box before the plan is delivered.
Traceable releases
Every engine release ships with a manifest of cryptographic fingerprints. We know which binary computed each route.
Parallel portfolio
The engine launches two configurations with two seeds each, in parallel on your computer's cores, and keeps the best. The final selection is deterministic.
Your data stays in your company
Addresses, customers, drivers and costs are processed on your own computers. Route planning does not depend on an external service.
Code you can review
The engine and the research are in an open repository: your IT team can audit how every plan is computed.
Fewer miles, less fuel, fewer planning hours
Operations research consistently reports improvements in these variables when moving from manual planning to optimized routing.
* Your own figures are measured during the assessment and go-live: we compare your current routes against the optimized ones, with your own orders.
Illustrative scenario: 20 vehicles, 150 km per vehicle per day, 300 operating days, 15 % average improvement, 10 km/L and two hours of planning per day.
Let's look at your operation with your data
In the assessment we measure your current routes and show you, with your own orders, how much miles, vehicles and loading change.