Transport optimisation is usually explained either as impenetrable mathematics or as a black box that produces plans nobody can question. This series takes the middle path: it works through an optimiser stage by stage, in the order the stages actually run, and explains each one as a product decision rather than only as an algorithm.
The sequence matters. Before anything can be routed, something has to define what "better" means — that is the objective function, and it is a policy choice disguised as arithmetic. Then orders have to be grouped, a first route built, and that route improved. Each improvement method eventually gets stuck, and each subsequent method exists to escape the specific trap the previous one fell into.
For reasons that become clear in the opening article, the guide throughout is My Name Is Earl. Every stage has an accompanying interactive simulation, so you can drive the algorithm rather than take its behaviour on trust.
How transport optimisation works
Transport optimisation finds better allocations, sequences and routes while respecting operational constraints such as capacity, geography, driver hours, service windows and customer commitments.
What a route optimiser actually does
An optimiser turns a business definition of “better” into an objective function, constructs a feasible starting plan, then improves it using local search or broader metaheuristic exploration. The result is a plan shaped by the objectives and constraints the system has been given.
Construction versus improvement
Construction methods build an initial route from nothing or from a simple seed. Improvement methods start with an existing plan and change part of it. The series follows that journey from objective function and order clustering through Clarke-Wright construction, Two-Opt and Three-Opt improvement, then simulated annealing, Tabu Search, genetic algorithms, ant colony optimisation and Large Neighbourhood Search.
Heuristics versus metaheuristics
Exact optimisation becomes impractical as operational problems grow in size and constraints. Heuristics use useful rules to find good plans quickly; metaheuristics provide broader search strategies for escaping the local traps where a straightforward improvement method gets stuck.
-
The series opener: why transport optimisation is a product problem before it is an algorithm problem, and why Earl Hickey's list is the right way to frame it.
-
Why transport optimisation has to start by defining what better actually means.
Interactive simulation: Weighted Objective Functions — An Interactive Transport Planning Model →
-
Before a route can exist, something has to decide which orders belong together, and that clustering decision constrains everything downstream.
Interactive simulation: Order Clustering — Comparing K-Means, DBSCAN and Graph Partitioning →
-
The Clarke-Wright savings algorithm from 1964, still running inside commercial transport optimisers, and why chaining stops beats visiting them one at a time.
Interactive simulation: Clarke-Wright Savings — An Interactive Vehicle Routing Simulator →
-
Two-Opt and Three-Opt formalise the planner's instinct that a route crossing its own path “looks wrong”, and then run straight into the local optimum.
Interactive simulation: Two-Opt and Three-Opt — An Interactive Route Improvement Simulator →
-
Simulated Annealing: why a routing engine sometimes has to accept a worse route in order to reach a much better one.
Interactive simulation: Simulated Annealing — An Interactive Route Optimisation Simulator →
-
Tabu Search escapes a local optimum using memory rather than chance: the optimiser keeps a list of what it has just tried and refuses to repeat it.
Interactive simulation: Tabu Search — An Interactive Route Optimisation Simulator →
-
Genetic algorithms stop improving one plan and instead run a whole population of competing plans that can breed.
Interactive simulation: Genetic Algorithms for Routing — An Interactive Evolution Simulator →
-
How pheromone trails, exploration, and evaporation help a routing system learn from successful routes.
Interactive simulation: Ant Colony Optimisation — An Interactive Routing Simulator →
-
How destroying and repairing part of a route helps an optimiser escape the traps that small improvements cannot reach.
Interactive simulation: Large Neighbourhood Search — An Interactive Routing Simulator →
-
From Clarke-Wright and 2-Opt to Simulated Annealing, Tabu Search, Ant Colony and Large Neighbourhood Search, what Earl taught us about solving real optimisation problems.