Flying UAVs Travelling Salesman

Problem with Two Depots

Thesis to be submitted in partial fulfillment of the

requirements for the degree

of

Integrated Masters of Technology in Industrial Engineering

Management

by

Aditya Jaiswal
20MF3IM01

Under the supervision of

Prof. Anand Jacob Abraham

DEPARTMENT OF INDUSTRIAL AND SYSTEMS ENGINEERING

INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR
 Department of Industrial and Systems
Engineering
Indian Institute of Technology, Kharagpur
India - 721302

CERTIFICATE

This is to certify that we have examined the thesis entitled Flying UAVs Travelling
Salesman Problem with Two Depots, submitted by Aditya Jaiswal(Roll Num-
ber: 20MF3IM01 ) a postgraduate student of Department of Industrial and Systems
Engineering in partial fulfilment for the award of degree of Integrated Masters of Tech-
nology in Industrial Engineering Management. We hereby accord our approval of it as
a study carried out and presented in a manner required for its acceptance in partial ful-
fillment for the Post Graduate Degree for which it has been submitted. The thesis has
fulfilled all the requirements as per the regulations of the Institute and has reached the
standard needed for submission.

Supervisor
Department of Industrial
and Systems Engineering
Indian Institute of Technology,
Kharagpur

Place: Kharagpur
Date:

2
 ACKNOWLEDGEMENTS

I would like to express my deepest gratitude to Prof. Anand Jacob Abraham for
providing invaluable guidance throughout the course of this project. Their insightful
feedback, constant encouragement, and motivation were instrumental in the completion
of this research.
I would also like to extend my thanks to the faculty members of the Department of
Industrial and Systems Engineering for their invaluable insights and advice. A special
thanks to Lord Sri Krishna, the Supreme Personality of Godhead, for the strength to
continue and to my family and friends for their unwavering support throughout this
journey.

Aditya Jaiswal
IIT Kharagpur
Date:

3
 ABSTRACT

The Multiple Flying Sidekicks Traveling Salesman Problem (mFSTSP) is a modern lo-
gistics problem involving a truck and multiple UAVs (drones) that work in tandem to
deliver parcels to customers in a timely manner. This thesis introduces a new variation
of the problem by considering two depots: a starting depot and an ending depot. The
objective is to minimize the total delivery time, or makespan, while ensuring efficient co-
ordination between the truck and UAVs. A mathematical formulation is developed using
mixed-integer linear programming (MILP), and a heuristic approach is implemented to
solve larger problem instances. Numerical experiments demonstrate significant improve-
ments in delivery times with the two-depot configuration compared to a single-depot
model.

Keywords: Multiple Flying Sidekicks, Traveling Salesman Problem, UAVs, drones,

mixed-integer linear programming (MILP), two depots.

4
Contents

1 Introduction 6
1.1 Problem Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Objective and Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Literature Review 7
2.1 The Classical Traveling Salesman Problem (TSP) . . . . . . . . . . . . . 7
2.2 UAV-Based Delivery Systems and FSTSP . . . . . . . . . . . . . . . . . 7
2.3 Variants of TSP with Depots and UAVs . . . . . . . . . . . . . . . . . . 8

3 Mathematical Formulation 9
3.1 Sets and Indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3 Decision Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4 Objective Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.5 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.5.1 Customer Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.5.2 Truck Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.5.3 UAV Sortie and Endurance Constraints . . . . . . . . . . . . . . . 10
3.5.4 Time Synchronization . . . . . . . . . . . . . . . . . . . . . . . . 10
3.5.5 Subtour Elimination . . . . . . . . . . . . . . . . . . . . . . . . . 10

4 Methodology 11

5 Results and Discussion 12

5.1 Numerical Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.1.1 Comparison with Single-Depot Model . . . . . . . . . . . . . . . . 12
5.1.2 Impact of UAV Endurance . . . . . . . . . . . . . . . . . . . . . . 12
5.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

6 Conclusion and Future Work 13

Bibliography 13

5
Chapter 1

Introduction

1.1 Problem Overview

The Traveling Salesman Problem (TSP) is one of the most studied combinatorial
optimization problems in which a salesperson must visit a set of cities, returning to
the starting city while minimizing the total distance travelled. The Multiple Flying
Sidekicks Traveling Salesman Problem (mFSTSP) is an extension of the TSP
that integrates Unmanned Aerial Vehicles (UAVs) or drones, alongside a delivery truck.
In mFSTSP, the truck can launch and retrieve UAVs to deliver parcels to customers,
reducing delivery time (makespan) compared to the truck-only TSP.
This thesis modifies the mFSTSP by introducing two depots, allowing the truck and
UAVs to start their route at one depot and finish at another. This modification better
reflects real-world scenarios where logistics centers are geographically dispersed.

1.2 Objective and Scope

This research aims to develop a model that minimizes the makespan in a two-depot
mFSTSP by optimizing the coordination between the truck and UAVs. Specifically, the
following goals are addressed:

• Develop a mathematical model for the two-depot mFSTSP.

• Implement a heuristic solution for large instances of the problem.

• Compare the performance of the two-depot model with the single-depot configura-
tion.

6
Chapter 2

Literature Review

The Traveling Salesman Problem (TSP) is a fundamental combinatorial optimization

problem. Over decades, it has laid the groundwork for more complex routing problems,
including those involving Unmanned Aerial Vehicles (UAVs) and multiple depots.
In this section, we review the foundational work on TSP, as well as recent advancements
in UAV-based delivery systems that extend these classical models.

2.1 The Classical Traveling Salesman Problem

(TSP)
The TSP involves finding the shortest possible route that visits a set of cities, starting and
ending at the same city, while minimizing the total travel distance or time. As outlined
by Laporte (1992), the TSP is NP-hard, meaning that finding an exact solution for
large instances is computationally challenging. Various approaches, such as branch-and-
bound, cutting planes, and dynamic programming, have been developed to solve small
to medium-sized instances exactly. However, for larger instances, heuristic algorithms
such as nearest-neighbor, Christofides’ algorithm, and local search techniques are more
practical.
Laporte’s review emphasizes the importance of approximate algorithms when dealing
with real-world problems, such as logistics and vehicle routing. These algorithms provide
near-optimal solutions within a reasonable time frame, making them suitable for large-
scale applications Laporte (1992).

2.2 UAV-Based Delivery Systems and FSTSP

With the advent of UAVs in logistics, there has been a growing interest in extending
classical TSP formulations to include drone-based delivery systems. The Flying Side-
kicks Traveling Salesman Problem (FSTSP), introduced by Murray & Chu (2019),
integrates UAVs into delivery operations, allowing a truck to launch and retrieve drones
for parcel delivery. This system significantly reduces delivery times, as UAVs can deliver
packages while the truck continues its route.
Murray and Chu’s model introduces operational constraints such as UAV endurance,
launch and recovery times, and service time at customer nodes. Their work demonstrated
the practical benefits of using UAVs for last-mile deliveries, particularly in urban areas
where traffic congestion can slow down traditional delivery trucks Murray & Chu (2019)

7
2.3 Variants of TSP with Depots and UAVs
Variants of the TSP, such as the Multiple Depot Vehicle Routing Problem (MD-
VRP), address the challenges of multi-depot logistics, where trucks and UAVs start and
end at different depots. This flexibility allows for more efficient route planning, espe-
cially in scenarios where logistics centers are distributed across a city or region. However,
adding UAVs to this framework introduces additional complexity due to UAV endurance
constraints and the need for synchronization between truck and UAV operations.
Studies like those by Laporte Laporte (1992) and Murray and Chu Murray & Chu
(2019) illustrate the growing complexity of these problems as new variables are intro-
duced. Heuristic approaches such as local search, metaheuristics, and dynamic pro-
gramming are increasingly necessary to handle the exponential growth in the number of
possible routes and schedules.

8
Chapter 3

Mathematical Formulation

This section presents the mathematical formulation of the mFSTSP with Two Depots.
The objective is to minimize the total delivery time, or makespan, while ensuring each
customer is served exactly once by either the truck or a UAV.

3.1 Sets and Indices

• D = {d1 , d2 }: Set of depots, where d1 is the starting depot and d2 is the ending
depot.
• C = {1, 2, . . . , n}: Set of customer nodes.
• V = {v1 , v2 , . . . , vk }: Set of UAVs.
• N = C ∪ D: Set of all nodes (customers and depots).

3.2 Parameters
• τij : Time taken by the truck to travel between nodes i and j.
• τvij
′
: Time taken by UAV v to travel from node i to node j.
• ev : Endurance (battery capacity) of UAV v.
• sLv,i : Launch time for UAV v at node i.
• sR
v,i : Recovery time for UAV v at node i.

• σi : Service time for the truck at node i.

3.3 Decision Variables

• xij ∈ {0, 1}: 1 if the truck travels from node i to node j, 0 otherwise.
• yvijk ∈ {0, 1}: 1 if UAV v flies from node i to deliver to customer j and returns to
node k, 0 otherwise.
• t̂i ≥ 0: Time when the truck arrives at node i.
• t̂′vj ≥ 0: Time when UAV v arrives at customer j.

9
3.4 Objective Function
Minimize the makespan, i.e., the total time to complete all deliveries and return to depot
d2 :
min t̂d2

3.5 Constraints
3.5.1 Customer Coverage
Each customer must be visited exactly once by either the truck or a UAV:
X X X X
xij + yvijk = 1, ∀j ∈ C
i∈D∪C v∈V i∈D∪C k∈D∪C

3.5.2 Truck Flow

The truck must start at depot d1 and end at depot d2 :
X X
xd1 j = 1, xid2 = 1
j∈N i∈N

The truck must visit each customer exactly once, maintaining flow conservation:
X X
xij = xjk , ∀j ∈ C
j∈N k∈N

3.5.3 UAV Sortie and Endurance Constraints

A UAV can perform at most one delivery per sortie. The total travel time of a UAV
sortie must not exceed its endurance:
′ ′
τvij + τvjk ≤ ev , ∀v ∈ V, i ∈ D ∪ C, j ∈ C, k ∈ D ∪ C

3.5.4 Time Synchronization

The truck and UAVs must coordinate their routes and timings to minimize delays:
t̂j ≥ t̂i + τij , ∀i, j ∈ N
UAV sorties must respect launch and recovery times:
t̂′vj ≥ t̂i + sLv,i + τvij
′
, ∀v ∈ V, j ∈ C
After completing the delivery, UAVs return to the truck at a recovery node:
t̂k ≥ t̂′vj + sR ′
v,k + τvjk , ∀v ∈ V, j ∈ C, k ∈ D ∪ C

3.5.5 Subtour Elimination

Subtour elimination constraints ensure that the truck’s route is valid:
uj ≥ ui + 1 − M (1 − xij ), ∀i, j ∈ C

10
Chapter 4

Methodology

Due to the complexity of solving large instances of the two-depot mFSTSP, a heuristic
approach is implemented. The methodology follows a three-phase process:

1. Phase 1: Initial Customer Assignment

Customers are initially divided between the truck and UAVs based on their prox-
imity to the depots. A nearest-neighbour heuristic is used to assign customers to
either depot based on their distance and potential delivery time reductions.

2. Phase 2: UAV Sortie Generation

Feasible UAV sorties are generated based on the endurance constraints and the
proximity of customers. The launch and recovery points for each UAV are optimized
to minimize waiting times at recovery nodes.

3. Phase 3: Scheduling and Route Optimization

A Mixed Integer Linear Programming (MILP) model is used to optimize the
scheduling of both truck and UAV operations. The objective is to minimize the
makespan by fine-tuning the routes and timings of each vehicle. A local search
algorithm is applied to refine the solution further.

The heuristic approach allows the problem to be solved efficiently for larger instances,
where direct MILP solvers would struggle due to the computational complexity.

11
Chapter 5

Results and Discussion

5.1 Numerical Experiments

The proposed model was tested on various problem instances with different numbers
of customers, UAVs, and depot locations. The two-depot model was compared to the
single-depot configuration to evaluate the impact on delivery time.

5.1.1 Comparison with Single-Depot Model

The introduction of a second depot provides additional flexibility for both the truck and
UAVs, leading to significant reductions in the makespan. On average, the two-depot
configuration reduced the overall delivery time significantly compared to the single-depot
model. This improvement was more pronounced in instances where customers were evenly
distributed between the two depots.

5.1.2 Impact of UAV Endurance

The endurance limits of UAVs played a critical role in determining the effectiveness of
UAV operations. Instances with higher UAV endurance saw greater utilization of UAVs,
leading to reduced makespan. However, when UAV endurance was constrained, the truck
had to cover a more significant portion of the route, diminishing the benefits of UAVs.

5.2 Discussion
The results demonstrate that the two-depot configuration provides significant operational
advantages in reducing delivery times and increasing flexibility in route planning. The
heuristic approach developed in this thesis was able to solve large instances efficiently,
providing near-optimal solutions in a fraction of the time required by exact MILP solvers.

12
Chapter 6

Conclusion and Future Work

The Multiple Flying Sidekicks Traveling Salesman Problem (mFSTSP) with two depots
offers a novel and practical extension to the classic mFSTSP. Introducing two depots
allows for more flexible route planning and significant reductions in the total delivery
time. The heuristic solution approach developed in this thesis is effective for solving
large problem instances in a reasonable amount of time.
Future research could explore the following areas:

• Integration of multiple trucks and UAVs to handle even larger delivery networks.

• Development of more sophisticated endurance models for UAVs that consider real-
world factors such as weather conditions and battery degradation.

• Investigation of dynamic customer requests, where customers may place orders after
the delivery route has already been initiated.

13
Bibliography

Laporte, G. (1992), ‘The traveling salesman problem: An overview of exact and approx-
imate algorithms’, European Journal of Operational Research 59(2), 231–247.

Murray, C. & Chu, A. (2019), ‘The flying sidekick traveling salesman problem: Par-
cel delivery with multiple drones’, Transportation Research Part B: Methodological
116, 86–110.

14