352 Z. Wang and J.-B. Sheu / Transportation Research Part B 122 (2019) 350–36 4
Table 1
The publications related to the truck and drone routing problem.
Refs. Problem Approach Drone capacity Drone and
truck
Objective Solved
instances
Murray and Chu (2015) 1-Truck;
1-Drone
Heuristic 1-Customer 1:1 Min returning
time of vehicles
10 or 20 nodes
Wang et al. (2017) M-Truck;
N-Drone
Worst-case
analysis
1- Customer M:1 Same as above –
Ponza (2016) 1-Truck;
1-Drone
Heuristic 1- Customer 1:1 Same as above Up to 200
nodes
Carlsson and
Song
(2017)
1-Truck;
1-Drone
Analysis base
on heuristics
1- Customer 1:1 Min traveling
time
Up to 100
nodes
Agatz et al. (2018) 1-Truck;
1-Drone
Heuristic 1- Customer 1:1 Min logistics
cost
10 nodes
Ha et al. (2018) 1-Truck;
1-Drone
Heuristic 1- Customer 1:1 Min operational
cost
Up to 100
nodes
Ham (2018) M-Truck;
N-Drone
Heuristic M- Customer No rendezvous Min maximum
time
Up to 100
nodes
Chang and Lee (2018) 1-Truck;
M-Drone
Heuristic 1- Customer M:1 Min total time Up to 100
nodes
This paper M-Truck;
N-Drone
Exact algorithm M-Customer M:N Min logistics
cost
15 nodes
Ha et al. (2018) also considered TSPD but with the objective of minimizing operational cost including transportation cost
and the cost incurred by the waiting time of a vehicle for another. Chang and Lee (2018) studied the routing problem with a
truck and several drones, in which the truck works as a carrier that carries drones to some centers where drones can fly for
serving customers. Mathew et al. (2015) studied the heterogeneous delivery problem with a truck and a drone and multiple
street vertices are defined for trucks to stop and deploy drones. Dayarian et al. (2018) considered a home delivery system
in which a drone is used to resupply a delivery truck regularly. Mourelo Ferrandez et al. (2016) compared the truck-drone
delivery system with standalone truck or drone systems.
Wang et al. (2017) and Poikonen et al. (2017) studied the VRPD in which a fleet of trucks, each equipped with a number
of drones, delivers parcels to customers. They derived worst case bounds for the ratios of the total delivery times with or
without drones. Different from our VRPD, they assumed that a drone must return to the truck that it launches from. The
feature that a drone may be associated with multiple trucks is thus prohibited. Ham (2018) considered two different types of
drone tasks: drop off and pickup. After a drone finishes a delivery, it can fly to a customer or depot for pickup. He considered
multiple depots, multiple trucks, and multiple drones and developed a constraint programming approach. This work did not
consider the rendezvous of two types of vehicles, making the problem different from ours. Campbell et al. (2017) provided a
strategic analysis for the design of hybrid truck-drone delivery systems using continuous approximation modeling techniques
to derive general insights.
The related literature and the work of this paper are concluded in Table 1 , from which it can be observed that this paper
differs from the literature in vehicle number, solution approach, drone capacity, mapping relationship between trucks and
drones. Because of these differences, the existing methods cannot be applied directly.
3. Problem statement and arc-based model
In this section, we define the VRPD and propose an arc-based integer programming model.
3.1. Problem statement
The VRPD is defined in a graph G = ( N, A ): N is the node set, containing the depot node, a set of customer nodes C = { c
1
,
c
2
, …, c
n
}, and a set of docking hub nodes O = { o
1
, o
2
, …, o
m
}; and A = {( i, j )| i, j ∈ N, i = j } is the arc set. For the notational
convenience, let o
s
and o
t
represent the origin and termination depot respectively. A set of trucks K and a set of drones D
are initially located at the depot and docking hub nodes, ready for serving customers.
A docking hub node serves as a transfer station for drones since it has turned out to be a preferable way in the drone
delivery industry ( Morgan, 2017 ). The station is generally designed for storing and maintaining some backup drones and
supporting drones to land. In pilot experiments, a drone can fly from any nodes but can only land at a docking hub or
depot, not a customer. This is because the landing of a drone usually needs special conditions, such as a spacious area and
a special docking device at the space that can exchange information with the drone in order to guide it to land accurately
and safely ( Hern, 2014 ). These conditions cannot be easily realized at a customer site because of safety or privacy. Instead
of landing at a customer, a drone in VRPD can do deliveries by a parachute airdrop ( Mogg, 2017 ).
A drone can be loaded with up to L
D
weight units of customer parcels. A truck can carry at maximum L
R
drones through
an arc and supply up to L
S
drones at a hub with customer parcels and corresponding attachments so that the drones can
fly from the hub for deliveries ( L
R
> L
S
). Furthermore, a truck can also be loaded with at maximum L
T
weight units of