Speaker
Description
Waste collection is considered as a first objective in planning a sustainable waste management,
and provision of substantial collection coverage in urban and rural areas should be satisfied
before investing in more sophisticated infrastructure. Therefore, good waste collection service
is essential to upgrade further the next levels of a municipal solid waste system. One of the
ways to enhance performance on waste collection is to plan effective routing of trucks. The
study formulated a Time-Dependent Capacitated Arc Routing Problem with Green Routing (TD-
CARP-GR) to model the waste collection routing problem of Parang, Marikina City using three
scenarios: baseline, parameter, and collection scenarios. This study focused on the door-to-
door method of waste collection and its transport to a Materials Recovery Facility (MRF) in
Parang, Marikina City. The objective is to build an optimization problem of waste collection and
minimize the distance taken by the truck with their corresponding GHG emissions and fuel
consumption. A physics-based algorithm called Electromagnetism-like (EM-Like) is used for the
solution process and is executed on a directed, multigraph-based representation of the road
network, where edges contain the demand points and nodes will be the intersections and dead-
ends. The simulation results revealed that the particle movement boundary and the
initialization phase where feasible solutions are generated to start the algorithm, affects the
search process. Moreover, using Lilliefors, Kruskal-Wallis, and post-hoc test Tukey-Kramer, the
particles were observed to achieve diversity and cooperation in the task to minimize the
obtained distance values. The Shannon Entropy measure also quantified the amount of
information contained during the search when two parameter scenarios are compared and
showed that the traditional formulation of particle movement in [0,1] obtained better search
quality. However, the algorithm suffers from slow convergence rate of solutions, as seen from
the huge difference between the randomly generated initial solutions and city-block based
initialization of which the latter encourages more intelligent way of permuting objects as it
tends to order streets that are closer to each other. Therefore, more iterations are needed
leading to higher computational expense to eventually reach the range of values generated
from the better initialization method implemented in the city-block based distance. Moreover,
subjecting the algorithm with a local search procedure significantly improved the obtained
values per iteration, with the highest improvement of 21 percent. This trend is evident to all the parameter scenarios. The study also strengthens the recommendation of the implementation
of capacity considerations since there is poor correlation between cost and distance in the
baseline scenario. For the capacitated scenario, if the cost per meter that the service needs are
to be calculated, the TDCARPK100 gives 0.02198 PHP/meter, CARPK50 of 0.02123 PHP/meter,
and baseline of 0.02591 PHP/meter. Thus, it can be concluded that capacity considerations of
truck are important to attain better efficiency and cost savings. In conclusion, the EM-like with
local search and selection strategy mechanism displayed significant minimization abilities of
distance values.
| Conference Topic Areas | Track7: Smart Operations and Maintenance |
|---|
