OPTIMIZATION MODEL FOR MUNICIPAL SOLID WASTE COLLECTION ROUTES CONSIDERING ACTUAL CONTAINER FILL LEVELS

Abstract

The paper proposes a two-stage algorithmic approach to planning municipal solid waste collection: first, a set of container sites to be serviced is formed from current fill-level measurements; then the visit sequence is optimized with a genetic algorithm. The route is encoded as a permutation of unique site identifiers (giant tour), and a separate Split decoder divides the permutation into feasible trips while accounting for vehicle capacity. This representation eliminates repeated depot markers in the chromosome and, by construction, prevents duplicate sites and the formation of autonomous subtours.

The method was tested in a reproducible synthetic experiment with 30 sites. At a 70% fill threshold, 9 sites were active. The effects of selection and of the optimizer were evaluated separately. Threshold selection reduced the static route length from 184.85 to 52.20 km, although this comparison corresponds to different service volumes. On the same set of 9 sites, the genetic algorithm shortened the route from 52.20 to 28.52 km and reached the exact solution found by dynamic programming in all 30 independent runs. The results concern a computational prototype using synthetic data and are not an estimate of actual savings or sanitary effects for Bishkek.