Two-echelon vehicle routing problem(2E-VRP) is a kind of vehicle routing problem in which freight from depot is
compulsorily delivered through intermediate depots(satellites). The first echelon is from depot to satellites, while the second
is from satellites to customers. This paper proposes a Memetic algorithm to solve the NP-hard problem in a bottom-up way.
The customers are assigned to the satellites reasonably by an improved optimal splitting algorithm MDVRP-Split. Then
the first-echelon problem is solved by using local search. The elitist produced by crossover is improved by local search.
Computational tests on some benchmark instances show that the proposed algorithm pays more attention on the balance
between solution quality and efficiency, and outperforms two existing algorithms for 2E-VRP.