Route optimisation is a planning method that works out which load each of one or more vehicles will take, in which order and along which roads it will visit the stops, by calculating distance, time and cost together. Software takes vehicle capacity, delivery time windows, driver break rules, road restrictions and tolls as constraints and proposes the trip with the lowest total cost.
What problem does route optimisation solve?
A navigation app takes one vehicle from point A to point B by the fastest road. Route optimisation, by contrast, deals with many vehicles, stops and rules at the same time: which load goes to which vehicle, in which order are the stops visited, where does the driver take a break? In operations research this question is called the Vehicle Routing Problem (VRP). Its single-vehicle form reduces to the travelling salesman problem. With just 10 stops there are 3,628,800 possible sequences; because this number grows exponentially as stops are added, software does not try every possibility but uses heuristic algorithms to find a good solution quickly.
Which data are taken into account?
- Distance and time matrix: Kilometres between stops and driving times that vary with the time of day and traffic.
- Time windows: A factory accepting goods only between 08:00 and 16:00, a port's cut-off time, booked unloading slots.
- Vehicle capacity: Weight, volume, pallet spaces or loading metres (ldm); equipment requirements such as reefer, ADR or lowbed.
- Driver rules: In the EU, under Regulation (EC) No 561/2006 and the AETR, daily driving is limited to 9 hours (10 hours twice a week), and 4.5 hours of continuous driving must be followed by a break of at least 45 minutes. Details are in the driver working and rest times entry.
- Restrictions: Weekend and public-holiday truck bans, ADR tunnel codes, bridge weight and height limits.
- Cost items: Fuel, motorway and bridge tolls, ferry or Ro-Ro tickets, driver accommodation.
How is route optimisation done?
- Clean the data: Addresses are converted into coordinates and wrong locations are corrected. If the coordinates of the office building have been entered instead of the factory gate, the plan will not work on the ground.
- Separate the constraints: Decide which rules are hard (bans, capacity) and which are soft (preferred times). Soft rules enter the model with penalty points.
- Choose the objective: Total kilometres, number of vehicles, duration of the longest tour or total cost; the plan takes shape around whichever is to be minimised.
- Test the solution: The algorithm proposes a plan and the planner corrects it with local knowledge: the queue at the border crossing, the consignee's lunch break, a trailer that cannot turn in a narrow street.
- Update live: Once the trip has started, deviations are spotted from GPS vehicle tracking data and the remaining stops are re-sequenced. Daily vehicle tracking and reporting passes the same data on to the customer.
The difference between urban distribution and an international trip
| Criterion | Multi-stop distribution | Long-haul international trip |
|---|---|---|
| Number of stops | Dozens of stops a day | Usually 1–5 loading and unloading points |
| Main difficulty | Stop sequence and time windows | Border crossing, transit permit, driving time, vessel schedule |
| Decision variable | Sequencing and vehicle assignment | Route corridor, transit country, break and overnight location |
| Source of uncertainty | Traffic | Customs and border waiting times |
| Replanning | Several times a day | Event-driven, throughout the trip |
On an international trip the shortest route is often not the cheapest. When deciding whether a truck from Türkiye to Central Europe should take the Bulgaria–Serbia corridor or a Ro-Ro connection to Trieste, transit permits, tolls, driver visas and the sailing schedule are weighed together. Truck driving bans in Europe can also determine the arrival day: a vehicle that cannot cross the border on Friday afternoon may be held until Monday.
Common mistakes
- Minimising kilometres only. A toll motorway may be shorter but more expensive overall.
- Leaving loading and unloading times out of the model.
- Leaving statutory breaks out of the plan. The arrival time on paper will then not be met.
- Working with bad data. AI-assisted planning tools also produce faulty plans from wrong addresses and missing constraints.