Capability
The Optimisation Engine
The optimisation engine is the part of Optrak that works out the routes. You give it the orders to be delivered or collected, the vehicles and drivers you have, and the rules they have to work to. It finds the lowest cost set of routes it can that gets the work done without breaking any of those rules. It doesn’t go about that the same way for every customer: the methods it uses are chosen to suit the kind of work your operation actually runs.
Last updated: 2026-08-11
No Two Operations Are Difficult in the Same Way
A fleet making many small drops around a city, against tight delivery windows, is a different problem from one running a few full loads across the country. A method that does well on one can do a poor job on the other.
So Optrak doesn’t have a single method. It has around thirty, each good at a different part of the job, and it combines them into a sequence that suits the work you run. That sequence is a short set of instructions rather than something built into the software, so a different approach can be tried out and measured against your own past work before anyone commits to it. The methods themselves go into what each one is for in technical detail.
How It Finds a Good Plan
There are far more ways to arrange the work than anyone could ever check, so the engine doesn’t try to look at all of them. Instead it:
- Builds a plan — gives every order a vehicle and a sensible place in that vehicle’s run.
- Improves it — tries small changes, such as moving a drop to another vehicle or calling at two sites the other way round, and keeps the ones that save money.
- Shakes it up — small changes stop helping after a while, so it takes a chunk of work back out of a good plan and puts it in differently. That often turns up arrangements the small changes would never have reached.
- Keeps going until the time you have allowed runs out, then gives you the best plan it found.
It runs many attempts at once, and they feed off each other rather than working in isolation. How the optimisation works goes over the same ground at more length, with what goes in and comes out, and what the optimiser means when it says one plan is better than another.
The Rules Are Checked as It Goes, Not at the End
Every time the engine considers a change, it checks the plan would still work in practice: that the load will fit and come off in the right order, that the driver is qualified and within their hours, that the vehicle can get into the site, that deliveries still land inside their windows, and — for an electric vehicle — that it has enough charge to finish. Anything that breaks one of those is thrown out there and then.
That is why an Optrak plan doesn’t need checking over for problems afterwards. It never produces one that has them.
Set Up for Your Operation, Not for an Average One
Most routing software arrives with one set-up for everybody. Optrak works out which of its methods suit you by trying them on your own past work and seeing which produce the best plans — and does that again when things change. This is the part of Optrak that uses machine learning: it chooses and sets up the methods. It does not write the routes.
Charging Is Part of the Route
If a round is too long for a single battery charge, the engine decides where to stop and charge as it works the route out — fitting the stop around the driver’s break, and taking on only as much charge as the rest of the round needs. Left until afterwards, you tend to find out too late that the round won’t work.
In This Section
- How the Optimisation Works What Optrak's optimiser actually does. There are more possible plans than anyone could check, so it builds plans, improves them, shakes them up, and keeps the cheapest one that works — checking as it goes that the plan holds up in the yard and on the road.
- Heuristic Catalogue and Optimisation DSL Optrak's optimiser runs around thirty heuristic families composed at runtime via a small scripting language — not a hardcoded pipeline. New strategies are scripts, not new builds, and the catalogue is designed against the actual fleet data model rather than a textbook CVRP simplification.
- Parallel Heuristics, Selection, and Acceptance Optrak's optimisation phases run heterogeneous heuristics in parallel over a shared solution population, with pluggable selection and acceptance strategies. Inter-thread visibility means a breakthrough on one core feeds every other within the same phase: the cooperative parallel pattern of Crainic and Toulouse, applied to heterogeneous heuristics rather than copies of one.
- Autotuning: Fitting the Optimiser to Your Operation Optrak works out which of its planning methods, and which settings, produce the best plans on your own work — automatically, by trying them against real days from your operation and keeping what wins. It is done when you are set up, and again when the operation changes or Optrak gains new methods, so the optimiser is fitted to how you run today rather than how you ran on day one.
- In-Route Charging: Charging Stops Put In While the Round Is Built, Not Checked Afterwards Optrak puts charging stops into a round while it is building it, rather than checking afterwards whether one is needed. It weighs what the electricity costs, how long the stop takes, and how much charge the stretches either side will use, against the cost of doing the work on diesel — which brings rounds that are too long for one charge into reach.
Frequently Asked Questions
Do I need to understand any of this to use Optrak?
No. Matching Optrak to your operation is done when you are set up, and again if the operation changes. There is nothing here a planner has to adjust from one day to the next. This section is for people who want to know how it works, or who are weighing Optrak against what they already use.
Is this “AI”?
Partly, and it is worth being precise about which part. Your routes are worked out by well-established mathematical methods that have been developed and tested over decades — not by AI. What AI does is choose which of those methods to use for your operation, and set them up. No AI writes your routes. AI in route optimisation explains where the line sits and why.
How is this different from other routing software?
What it takes into account while it is working. Most routing software is built around a simplified version of the problem — deliveries with a quantity, vehicles with a capacity — and deals with everything else afterwards. Optrak works with the real detail from the start: what will physically fit on the vehicle and come off in the right order, which drivers are allowed to drive what, how long roads actually take at different times, and how much charge an electric vehicle has left. The data model covers what it knows about your operation.
What happens when my operation changes?
Optrak is set up again. A new depot, a contract with a different pattern of drops, or vehicles moving from diesel to electric all change which methods produce the best plans. See autotuning.
See Optrak on Your Own Routes
Book a demo and we'll model your fleet, depots and constraints — and show the savings on real data.