“Flight Map” Modelling Intellectual Geoinformation System for

Urban Areas Cargo Delivery by Unmanned Aerial Vehicle

Oleg Golovnin, Nikita Ostroglazov and Tatyana Mikheeva

Samara National Research University, Samara, Russia

Keywords: Intelligent Transport Systems, Unmanned Aerial Vehicles, Transport Network Model, Geoinformation

System, Genetic Algorithm, Branch and Bound Method.

Abstract: This article is dedicated to solving the problem of safe and secure cargo delivery in urban territory using

unmanned aerial vehicles (multicopters) by modelling “flight map”, i.e. a system based on intellectual

geoinformation system that develops an optimal and secure route for each UAV in the system and tracks all

of them on electronic map online. The paper describes the results of algorithms analysis and assumes the

algorithm showing higher test results.

1 INTRODUCTION

Today multicopters are highly sought and popular

remote control Unmanned Aerial Vehicles (UAVs).

Multi-rotor UAVs are equipped with special

electronic sensors like accelerometers and gyro-

scopes allowing the UAV to stay stable and maneuver

(Garcia et al., 2017). The multicopter can be used for

performing various tasks including cargo delivery.

UAV owners, logistics companies and carriers can

benefit from using UAVs for cargo delivery purposes

(Tkachev et al., 2015). Thereby, modelling “flight

map” for UAVs using geoinformation system is

absolutely necessary. “Flight map” modelling in this

case means developing a transport network model,

UAVs security control system (Bansal et al., 2017)

providing UAVs flights safety in predetermined

echelons and corridors, and online real-time UAVs

tracking system. The automation system is based on

geoinformation system, electronic vector map,

transport network model and transport infrastructure

objects database.

A step-by-step transition from autonomous route

guidance to dynamic is a promising direction of route

navigation development. In this case, the main task of

information processing is optimal route selection and

its dynamic correction during the flight (Iswanto et

al., 2017). Dynamic route management presumes,

first and foremost, the possibility to dynamically

correct (recalculate) at any time and at any point of

the route the optimal route taking into account any

changes in the transport network (Stepanov, 2015)

and also possible changes in route selection criteria

(Antoniades et al., 2017).

2 TRANSPORT NETWORK

MODEL

The paper provides graph-based transport network

model for UAVs route development consisting of

following entities:

area – piece of transport network represented

by a set of coordinate pairs, described as a set

of physical parameters. The area is encoded by

a polygon defined by its vertices coordinates;

node – point where traffic flows are separated.

The node is the vertex of the oriented graph,

channeling traffic flows. It lies at the junction

of two areas and indicates the possibility of

moving from one area to another in the

direction defined by the relevant arc. Two

adjacent areas are assigned with two nodes.

Node is encoded by a point;

arc – element of the oriented graph that defines

the direction of the traffic flow on the area. It

contains relevant characteristics including arc

length, traffic intensity, traffic density, etc.

Two nodes are assigned with one arc defining

the traffic direction, or two arcs defining the

direction reverse area. Several arcs (traffic

flows) can be included in one area but each arc

can be included in only one area. Arc is a linear

object.

Golovnin, O., Ostroglazov, N. and Mikheeva, T.

“Flight Map” Modelling Intellectual Geoinformation System for Urban Areas Cargo Delivery by Unmanned Aerial Vehicle.

DOI: 10.5220/0006670002770282

In Proceedings of the 4th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2018), pages 277-282

ISBN: 978-989-758-293-6

Copyright

c

2019 by SCITEPRESS – Science and Technology Publications, Lda. All rights reser ved

277

The “flight map” is meant for UAVs route

development (from one landing platform to another)

when performing cargo transportation from supplier

to a consumer taking into account normal and

abnormal situations. The “flight map” is

characterized by a corridor (a range of possible UAV

horizontal plane coordinates) and echelon (a

conditional height, established intervals value distant

from other heights).

This paper assumes that only 1 UAV is present in

the arc or node at the same time. That means UAVs

are canalizing in terms of time and space.

Landing Platform (LP) is designed for safe take-

off and landing of UAVs in urban areas. All the LPs

are deployed on the electronic map and included in

UAV “flight map”. The LPs can be divided into

groups according to its assignment: sources (take-off

platforms), outlets (landing points), charging points,

service stations, emergency landing platfoms.

3 ROUTE DEVELOPMENT

ALGORITHMS

3.1 Graph Model Algorithms

Route development algorithms are based on UAVs

“flight map” graph model and rely on Dijkstra’s

algorithm and branch and bound method for various

optimality criteria. In the context of dynamic route

management an additional criterion appears: optimal

path searching algorithm running time, which should

be minimized (Hayat et al., 2017).

In fact, route development is one of the traveling

salesman problem variations. All the optimal path

searching algorithms operate with graphs, all vertices

of which are included in the route (Vareldjan et al.,

2015).

3.2 Little’s Algorithm

An algorithm for the traveling salesman problem by

John D. C. Little is a particular case of the branch and

bound method. In a best-case scenario its usage

provides an opportunity to reduce the number of

operations.

The algorithm is used for an optimal route search

provided that an object (UAV) is returning to the

starting point. As a result, Little’s algorithm provides

a close loop (which may be not optimal) in less than

n steps. Calculation process complexity lies in the fact

that at each step it is necessary to analyze the elements

of the matrix and select zero elements (applicants for

branching and evaluation). With regard to algorithm

running time, with big n values the optimal path may

not be found at all due to the growth of the number of

branches and bounds. Therefore, it is required to

determine the optimal value for the algorithm.

3.3 Genetic Algorithm

Initialization, i.e. initial population formation is the

random selection of a predetermined number of

chromosomes represented by binary sequences of

fixed length. For UAV “flight map” modelling the id

number of the visited object is used as a gene. Route’s

weighting coefficient is assumed as a chromosome

fitness function (Silva Arantes et al., 2017).

3.4 Initial Data and Requirements

The algorithms were tested using the initial data

shown in table 1 for single UAV involving. New

graph is generated automatically after every test

cycle.

Table 1: Test cycles initial data.

№

№ of areas

№ of

arcs

Т-shaped Х-shaped I-shaped

1 5 20 3 275

2 10 40 6 550

3 20 80 12 1100

4 40 160 24 2200

5 80 320 48 4400

6 160 640 96 8800

7 320 1280 192 17600

8 640 2560 384 35200

9 1280 5120 768 70400

10 2560 10240 1536 140800

11 5120 20480 3072 281600

12 10240 40960 6144 563200

LP is an integral structure for UAV’s take-off and

landing providing safe and accurate landing in urban

areas. LP has to provide UAV’s wireless charging,

UAV’s status, options, cargo information and other

data transmission via WiFi / 4G / Ethernet networks.

LP’s normal functioning should be ensured for supply

voltage of 100-240 V, temperature of 5-45 C, wind

speed up to 5 m/s and light precipitation.

Weight of transported cargo should not exceed

5 kg. Cargo has to be packed in a special container for

transportation and should not be prohibited from

transportation by Government regulations.

The UAV should be supplied with GPS /

GLONASS navigation system, telemetry system,

VEHITS 2018 - 4th International Conference on Vehicle Technology and Intelligent Transport Systems

278

radio-beacon system, backup autonomous navigation

system, navigation light and 3-axis magnetometer.

The range of operating flight heights is from 300 m to

1200 m.

The analysis of each tested algorithm performance

is presented in “Test Results” section.

4 SYSTEM DEVELOPMENT

System was based on intellectual geoinformation

system ITSGIS. Figure 1 shows the UAV “flight

map” implemented in ITSGIS (Golovnin et al., 2017).

The system offers the possibility to add and edit

UAV “flight map” in multi-user mode, which

determines the use of multitier system architecture. It

is important to ensure data consistency and eliminate

emerging conflicts. The system is developed using a

three-tier architecture model (as shown in Figure 2).

Communication between clients and the application

server is based on Windows Communication

Foundation (WCF) technology.

Figure 1: UAV “Flight map”.

Figure 3 shows the diagram of system

components.

Fundamental system components are:

ITSGIS.exe – client’s executable file;

WCFServer.exe – server’s executable file;

DAO – database access layer;

Services – system services; located on the

server side and access the database through the

database access layer;

Managers – system managers; located on the

client’s side and access the services through the

network channel.

Figure 2: System architecture.

“Flight Map” Modelling Intellectual Geoinformation System for Urban Areas Cargo Delivery by Unmanned Aerial Vehicle

279

Figure 3: System components.

5 TEST RESULTS

Graphs of Figures 4 to 6 show the results of algorithm

tests. All algorithm tests were conducted under the

following testing conditions: CPU with 2 cores 2.8

GHz; 8 GB of RAM; Microsoft Windows 8;

Microsoft .NET 4 and using the same test equipment.

Figure 4: Branch and bound method and its adapted

analogue comparison.

Figure 4 shows the comparison of branch and

bound method and its adapted analogue according to

its running time parameter.

The point of adapted branch and bound method is

that it involves adding of an extra evaluation in order

to find the best node to branch. Thereby, evaluation

of every node in the tree is presented not only by the

distance travelled to the current node but also by the

distance left to travel to the final node.

In this case, the number of optimal path searching

iterations increases in proportion to the length of the

route.

Figure 5 shows the dependence of the optimal

route searching time on the number of vertices of the

graph containing about 35000 arcs.

Figure 5: Dependence of the optimal route searching time

on the number of vertices.

Figure 6 shows the comparison of branch and

bound method and Dijkstra’s algorithm. This test

results show that with the number of nodes up to

30 000, Dijkstra’s algorithm and branch and bound

method have roughly the same running time.

However, with the vertices number increasing,

Dijkstra’s algorithm is showing significantly lower

results.

Figure 6: Dependence of the optimal route searching time

on the model dimension.

Graph of Figure 7 shows the results of Little’s

algorithm running time analysis for various numbers

of points.

For routes with the number of points more than 60

algorithm’s running time increases significantly.

Thereby, the optimal number of points to use that

algorithm is ≈60.

VEHITS 2018 - 4th International Conference on Vehicle Technology and Intelligent Transport Systems

280

Figure 7: Little’s algorithm running time analysis.

Graphs of Figures 8 to 10 show the genetic

algorithm test results. An important parameter for

mutation operator usage is a mutation probability

р

тut.

. Graph of Figure 8 shows the dependence of the

route length on the algorithm running time for

different mutation probabilities (0.3, 0.15, 0.05). The

test was carried out for the 100 points graph.

Figure 8: Dependence of the route length on mutation

probability.

Algorithm’s performance for 100 points graph

various crossing over methods is shown in Figure 9.

Figure 9: Dependence of the route length on crossingover.

The graph shows that standard crossing over

method outputs a near-optimal result in less time than

“greedy crossover”. Convergence rate and accuracy

depends on the number of individuals specified

beforehand.

Figure 10 shows the genetic algorithm

convergence analysis for 10, 100, 250 and 500

individuals. The analysis shows that an increase in the

number of individuals leads to a more accurate result.

Figure 10: Dependence of the convergence rate on

individuals number.

6 CONCLUSIONS

This paper suggests the concept of “flight map”,

developed for providing safe and secure unmanned

aerial vehicles cargo delivery in urban areas.

The paper describes a graph-based transport

network model for UAVs route development;

requirements for unmanned aerial vehicles, landing

platforms and cargo parameters; system architecture

model and system components diagram.

The “Test Results” section presents the results of

various algorithms performance analysis and its

comparison depending on relevant parameters.

Proceeding from the results obtained the Little’s

algorithm is appropriate to use for optimal route

searching in the cases of 60-70 vertices present. When

performing the tests for the graphs of relatively small

dimensions, all the examined algorithms show equal

results on the average. For the cases with the mean

distances in a graph the adapted algorithm has an

advantage. And for the cases with maximum

distances in a graph (from one graph’s end to another)

the benefits of adapted algorithm are decreasing.

ACKNOWLEDGEMENTS

This research is financially supported by Funding

Institution: Foundation for Assistance to Small

Innovative Enterprises (FASIE) under the contract

№1773ГС1/25439.

The experimental research was conducted with

the use of equipment of the Science Production

Centre “Intelligent Transport Systems”.

“Flight Map” Modelling Intellectual Geoinformation System for Urban Areas Cargo Delivery by Unmanned Aerial Vehicle

281

REFERENCES

Garcia, E., Casbeer, D., 2017. Coordinated Threat

Assignments and Mission Management of Unmanned

Aerial Vehicles. In Cooperative Control of Multi-Agent

Systems: Theory and Applications. P. 141.

Tkachev, S.B., Krischenko, A.P., Kanatnikov, A.N., 2015.

Automatic generation of complex spatial UAV

trajectories and synthesis control. In Mathematics and

math modelling, 1(1), Pp 1-17. (In Russ.).

Bansal, S., Chen, M., Fisac, J. F., Tomlin, C. J., 2017. Safe

Sequential Path Planning of Multi-Vehicle Systems

Under Presence of Disturbances and Imperfect

Information. In American Control Conference.

Iswanto, Wahyunggoro, O., Cahyadi, A. I., 2017. Path

planning of decentralized multi-quadrotor based on

fuzzy-cell decomposition algorithm. In AIP Conference

Proceedings, 1831 (1). AIP Publishing.

Stepanov, D.N., 2015. Application of Gabor wavelets in the

problem of navigation of a UAV using a video camera.

In Fundamental research, 12(1). Pp 85-91. (In Russ.).

Antoniades, K., Ishizaka, A., Ries, J., 2017. A Multi-

Criteria Support System for UAV Maritime

Surveillance. In Highly Dynamic Environment.

Hayat, S., Yanmaz, E., Brown, T. X., Bettstetter, C., 2017.

Multi-objective UAV path planning for search and

rescue. In Robotics and Automation (ICRA), IEEE

International Conference. Pp. 5569-5574.

Vareldjan. K.S., Paramonov, A.I., Kirichek, R.V., 2015.

Optimization of UAV trajectory in flying sensory

networks. In Electrosvyaz, Pp. 20-25. (In Russ.).

Silva Arantes, J. D., Silva Arantes, M. D., Motta Toledo, C.

F., Júnior, O. T., Williams, B. C., 2017. Heuristic and

genetic algorithm approaches for UAV path planning

under critical situation. In International Journal on

Artificial Intelligence Tools, 26(01).

Golovnin, O., Mikheeva, T., Fedoseev, A., 2017. Intelligent

geographic information platform for transport process

analysis. In CEUR Workshop Proceedings, Vol. 1901.

Pp 78-85.

VEHITS 2018 - 4th International Conference on Vehicle Technology and Intelligent Transport Systems

282