Clarke Wright Performance Benchmarks

In this vignette we discuss performance benchmarks of the Clarke-Wright algorithm implemented in this package by comparing the Clarke-Wright solutions of a set of problem instances with their optimal value.

The problem instance data

The problem instances we are basing our measurements on are provided courtesy of CVRPLIB and can be divided into five groups1 of different characteristics:

The data provides the problem instance as well as the optimal solution.

Relation to optimal and trivial solution

The idea of this indicator is to measure where the Clarke-Wright solution sits in between the optimal solution and the trivial solution2. Let \(\gamma\) be the cost of the Clarke-Wright solution, \(\omega\) the cost of the optimal solution, and \(\alpha\) the cost of the trivial solution. The measure \(\xi\) is defined to be \[\xi := 1 - \frac{\gamma - \omega}{\alpha - \omega}\,.\] The measure can move between zero and 1. It is zero if the solution is the trivial solution, and it is one if the solution is the optimal solution.

Evaluating this for CVRPLIB sample data yields the following graph.

The mean value over all problem instances is \(\overline{\xi}=96.7\%\).

Relative deviation of optimal solution

We can also simply measure the relative deviation of the Clarke-Wright cost to the optimal cost, \[\rho := \frac{\gamma - \omega}{\gamma}\,,\] which measures how much savings we miss by using the Clarke-Wright solution instead of the optimal solution.

The mean value over all problem instances is \(\overline{\rho}=11\%\), i.e. on average, we miss out on around \(11\%\) savings taking the Clarke-Wright solution compared to the optimal one.


  1. The naming is directly taken from CVRPLIB↩︎

  2. By trivial solution we mean the solution where each site is assigned exactly one route (which goes back and forth).↩︎