IEEE Computational Intelligence Magazine - May 2022 - 89

operational status can be described as
pulling, holding, coasting, or braking.
Fernández-Rodríguez et al. [60] proposed
a multiobjective PSO to minimize
the running time and energy consumption
of trains. Cao et al. [61]
scheduled the coasting time of trains by
GA to optimize punctuality, passenger
comfort, and energy consumption. Feng
et al. [62] used GA to manage the timetable
design and speed control simultaneously
to maximize the utilization of
regenerative braking energy.
E. Intelligent Route Scheduling
Route scheduling aims to allocate vehicles
and assign each vehicle a driving
route to achieve certain objectives. The
well-known vehicle routing problem
(VRP) is a typical route scheduling
problem, in which a delivery company
should design the driving route of each
truck/lorry for delivering the goods
from the depot to the customers and
finally going back to the depot. The
route scheduling is related to the service
quality and profit of delivery companies
from the business perspective. City
VRPs [63] consider both the benefit of
delivery companies (e.g., service quality
and route cost) and the benefit of citizens
(e.g., traffic congestion and environmental
pollution). In this regard,
route scheduling is also considered from
the citizen perspective.
In recent years, some new VRPs are
studied to meet the demands in smart
cities, including multi-echelon VRPs,
green VRPs, VRPs with EVs, VRPs with
crowdshipping, and VRPs with drones.
Multi-echelon VRPs are common in
city logistics, in which the delivery process
is divided into multiple steps rather
than directly delivering the goods from
the depot to the customers [64]. To
optimize the route cost, Yan et al. [65]
proposed a graph-based evolutionary
algorithm that incorporated the fuzzy
logic to enhance the search efficiency. In
addition, Wang et al. [66] proposed a
hybrid algorithm of PSO and GA to
minimize the total cost of a multi-echelon
logistics network. Green VRPs [67],
[68] focus on environmental pollution.
Poonthalir and Nadarajan [69] proposed
a PSO-based algorithm with a greedy
mutation operator to optimize not only
the route cost but also the fuel consumption
for reducing emissions of
greenhouse gas. Li et al. [70] proposed an
improved ACO algorithm to optimize
four objectives: revenue, cost, transportation
time, and CO2
emissions.
With the development of EVs, certain
studies considered the scenario of
using EVs to transport goods. The
usage of EVs can also reduce vehicle
exhaust emissions and improve the air
quality in cities. Thus, the VRPs with
EVs have connections to the green
VRPs. There are two critical issues: one
is that the charging of EVs requires a
lot more time than the refueling of traditional
petrol vehicles; the other is the
time-of-use (TOU) electricity price.
TOU electricity price means the electricity
price is time-varying throughout
the day, which is common in the current
electricity market. Considering
these two issues, Yang et al. [71] proposed
a GA-based algorithm and Miao
et al. [72] proposed a DE-based algorithm
to generate route schemes for
optimizing charging cost and battery
loss. Li et al. [73] further considered the
carbon taxes that aimed to reduce
greenhouse gas emissions. An ACObased
algorithm is proposed to optimize
the charging cost, payment for EVs and
employees, service quality, queuing time
of EVs for charging, and cost for carbon
taxes. Note that for readers who are
interested in VRPs with EVs, the
resources (including a benchmark set)
in https://mavrovouniotis.github.io/
EVRPcompetition2020/ can help them
start a study.
Crowdshipping, derived from sharing
economy, is a new delivery mode that
allows ordinary people to join the delivery
process using their own vehicles
[74]. To minimize the total cost of
crowdshipping delivery, Wu et al. [75]
proposed a scale-adaptive fitness evaluation
method and embedded it into GA.
Feng et al. [76] proposed a GA-based
evolutionary multitasking algorithm that
can solve multiple instances simultaneously
while obtaining promising results
through knowledge transfer.
Drones can also be used to deliver
goods, which can reduce labor cost and
improve delivery efficiency. Due to the
low carrying capacity of drones, drones
usually cooperate with traditional vehicles
(e.g., trucks) and undertake the lastmile
delivery. To deal with this new
scene, simulated annealing-based algorithms
are proposed in [77] and [78];
Shao et al. [79] considered the route
length of drones and the number of
landing depots along the route to formulate
an objective function and proposed
a hybrid algorithm of ACO and
A) algorithm; Das et al. [80] proposed an
ACO-based multiobjective algorithm to
minimize the route cost and to maximize
the service quality simultaneously.
In the future, large-scale highly
dynamic delivery systems will be promising
in smart cities. Some studies have
focused on this topic. To deal with the
challenge of large-scale delivery
demand, Xiao et al. [81] used the wellknown
NSGA-II [82] to divide the
large-scale VRPs into a set of subcomponents;
Jabir et al. [83] proposed a
hybrid algorithm of ACO and the variable
neighborhood search. To deal with
the challenge of dynamic delivery
demand, Jia et al. [84] proposed a setbased
PSO algorithm; Xiang et al. [85]
proposed an ACO-based algorithm that
used a pairwise proximity learning
method to help the algorithm adapt to
the change of delivery demand.
F. Intelligent Car Service
Car service in smart cities includes the
following three services: taxi, carpool,
and parking. These services are considered
from the business perspective since
the operators of these services want to
earn profits. In addition, these services
are typically used by citizens and are
thus also considered from the citizen
perspective. In the domain of taxis, taxi
dispatch is a critical problem to be
solved. Traditionally, taxi drivers do not
know where passengers are currently,
and they typically seek passengers based
on their own experience. Thus, some
passengers tend to wait a long time for a
taxi, and some taxi drivers tend to
drive without finding passengers. Taxi
MAY 2022 | IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE 89
https://mavrovouniotis.github.io/EVRPcompetition2020/ https://mavrovouniotis.github.io/EVRPcompetition2020/
IEEE Computational Intelligence Magazine - May 2022

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - May 2022
