FLIGHT SIMULATION ENVIRONMENTS APPLIED TO

AGENT-BASED AUTONOMOUS UAVS

Ricardo Gimenes

Polytechnic School - University of S˜ao Paulo, Av. Prof. Luciano Gualberto, trav. 3, n. 158

CEP 05508-010, S˜ao Paulo, Brazil

Daniel Castro Silva, Lu´ıs Paulo Reis, Eug´enio Oliveira

FEUP - DEI / LIACC, Rua Dr. Roberto Frias, s/n 4200-465, Porto, Portugal

Keywords:

Simulation Environments, Multi-Agent Systems, UAVs.

Abstract:

Developed countries have made signiﬁcant efforts to integrate Unmanned Aerial Vehicle (UAV) operations in

controlled aerial space due to a rising interest of using UAVs for civilian as well as military purposes. This

paper focuses on looking for reliable solutions and a way to validate an autonomous multiagent control system

through a different variety of Flight Simulations. This study has two main lines, the ﬁrst being the use of

multiagent systems in UAVs, aiming at a fully autonomous control. The second and focal line is a survey

about the variety of simulation systems dedicated to aerodynamics and aircrafts, comparing them and their

feasibility to validate the developed multiagent control system. One critical factor is hazard situations, like

emergency landing without runway or equipment failure, which should be predicted in an automated system.

Most ﬂight simulators are not realistic enough to validate a multiagent algorithm in hazard situations. At the

same time, it is impossible to predict every type of failure in real world. The boundaries of simulation should

be very well enclosured in order to present results using simulation.

1 INTRODUCTION

Ever since the construction of the ﬁrst vehicles, men

havedreamed of automating its operations. While this

autonomy has been restricted to limited functions, it

is desirable to elevate it to a full scale, allowing vehi-

cles to be fully autonomous. When full autonomy is

reached, vehicles will have to cooperate and coordi-

nate their actions with one another, in order to ensure

both security and an optimal use of resources. The

choice of an agent-oriented approach to control au-

tonomous vehicles is intended to save both time and

resources, enabling the vehicles to communicate, and

make their own mission planning and real-time deci-

sions.

In order to better develop and test such agent-

oriented approaches and coordination methods, a sim-

ulation environment must be used. Such environment

also saves time and resources, since it would be pro-

hibitory to test these systems using real vehicles.

This paper focuses on several existing simulation

environments that can be adapted to serve the objec-

tive at hand.

1.1 Unmanned Aerial Vehicles

UAV, acronym for Unmanned Aerial Vehicle, is con-

sidered, in this paper, as a vehicle able to ﬂy au-

tonomously without any external remote control.

There are already some examples of civilian UAV

usage, either operating or in developmental stage,

namely those related to surveillance or in need of ac-

cess to risky areas (Air & Space Europe, 1999). These

do not have signiﬁcant inﬂuence on the air trafﬁc sys-

tem, since the UAVs ﬂy over clearly deﬁned areas at

low altitude and speed.

The use of UAVs still involves a few uncertain-

ties and legal limitations as to where and how they

can be used. The majority of projects involving UAVs

are strictly military. Opposing to the military vision,

crewless civilian passenger aircraft raise a completely

diverse paradigm in terms of reliability. Because of

this, new hurdles come in the way of a future catego-

rization of passenger carrying UAV, as well as of non-

passenger carrying UAV sharing airspace with con-

ventional aircrafts.

Thus, operating UAV on international airspace

243

Gimenes R., Castro Silva D., Paulo Reis L. and Oliveira E. (2008).

FLIGHT SIMULATION ENVIRONMENTS APPLIED TO AGENT-BASED AUTONOMOUS UAVS.

In Proceedings of the Tenth International Conference on Enterprise Information Systems - SAIC, pages 243-246

DOI: 10.5220/0001710802430246

 SciTePress

must entail two closely related conditions (Allouche,

2000): the UAV must be safe and reliable enough to

ﬂy over densely populated areas and must be safely

operated through airspace.

1.2 Multiagent Systems

A multi-agent system (MAS) can be seen as a system

where entities are represented by independent agents,

which in turn communicate with each other, coordi-

nating their activities.

The UAV operation can surely be considered a

MAS in which every aircraft is an agent with its own

goals (destination, time frame of arrival, service stan-

dards, etc.), independent of the goals of other air-

craft. With this approach, it is possible to apply ne-

gotiation techniques, which will allow the aircraft to

cooperate in solving airspace conﬂicts. Answers in-

volving artiﬁcial intelligence software or MAS have

been researched on, in the search for decision-making

systems capable of avoiding aircraft encounters in a

given airspace (Lian and Deshmukh, 2006).

In section two of this paper, some of the require-

ments that were taken into account when evaluat-

ing the simulation environments are brieﬂy described.

Section three presents some of the analyzed simula-

tors, game engines, ﬂight simulators and middleware

engines. In the last section, some conclusions are

withdrawn and future work is outlined.

2 SIMULATION ENVIRONMENT

REQUIREMENTS

The choice of a simulation environment is dependent

on the objectives of the project. In this section, some

key features that may be desirable in the simulation

environment are brieﬂy described. These features can

be grouped into categories, such as graphical, phys-

ical or openness characteristics, among others, for a

better analysis.

The openness of the software is an important as-

pect. Openness can be deﬁned in terms of the possi-

bility to view and alter the source code, as well as ex-

pansion capabilities, such as the possibility to develop

add-ons, external modules and different agents that

can be linked to the environment, via deﬁned APIs,

interaction models or communication protocols, the

accessibility and format of the data that can be input to

the environment, and the output data format as well.

Accurate physical simulation is very important,

when dealing with robotic mobile agents, such as

UAVs. The following subsections explain in more de-

tail the importance of a correct physical modeling and

simulation of the aircraft and the environment.

2.1 Speciﬁc Aviation-Related

Requirements in Simulated

Environments

Realistic simulation of robotic sensors and actuators

in a complex, unstructured, dynamic environment,

such as ﬂuids, constitutes a research challenge. Most

simulators consider only the four basic vectors that

compose a ﬂight: lift, weight, drag and thrust. (Schiff,

1978). However, in real ﬂight, an aircraft has to deal

with numerous factors, not only pertaining to the air-

craft itself, but also external, environmental factors.

This way, to choose a ﬂight simulator system, it is

very important to deﬁne what is more relevant in the

research. Each simulator has its characteristics suit-

able (or not) with a speciﬁc kind of research. Con-

sidering developing an autonomous multiagent envi-

ronment as the main goal, the relevant parameters in

the ﬂight system can have three major lines: Accept-

able ﬂight model, considering the application under

investigation; Flexibility to interact with the agent

via programming interfaces; Possibility to have the

ﬂight model or simulated elements changed by exter-

nal software.

Agent algorithms which have to control an aircraft

will need a ﬂight simulator model as real as possi-

ble. As the number of simulated elements offered by

the ﬂight simulator increases, so does its adequacy to

serve as a suitable tool for simulated validation of the

agent-based UAVs.

2.2 Fault Injection in Flight Simulation

Models

The ﬁrm safety requirements related with aviation re-

quire the studies of any kind of ﬂight simulation to

keep in mind failure considerations. Fault injection is

a complex research line to reach a reliable and safe

ﬂight model in ﬂight simulation.

A fault injection method is essentially hardware

fault injection. In a ﬂight simulation, faults can be in-

jected into different inputs or outputs of the aircraft.

Considering the UAV completely controlled by soft-

ware, the range of failure possibilities determines the

way the faults have to be injected. The quantity of

injected faults is a quality parameter of any fault in-

jection method. Basically, the fault injection module

is a software module which interrupts the original in-

puts and outputs of the simulated aircraft.

A critical problem for fault injection using COTS

ICEIS 2008 - International Conference on Enterprise Information Systems

244

ﬂight simulators is how they are opened to interact

with their simulated hardware of an aircraft. Thus,

open source ﬂight simulators or open API from pro-

prietary software should be analyzed as decisive fac-

tors in a ﬂight environments choice.

2.3 UAV Flight Simulation

The way the ﬂight simulation is developed determines

its possibilities and, consequentially, the possibilities

and limitations the research has to deal with. Real-

time visual simulation is not necessarily a relevant

parameter. Simulations without visual interface of-

fer the possibility to produce a quantity of different

situations quicker than the visual simulation does.

The simulation of physics, kinematics and aerody-

namics is fundamental in a reliable ﬂight simulation.

In contrast, structural forces are not relevant in UAV

ﬂight model studies.

3 COMMON SIMULATION

ENVIRONMENTS

There are two main simulator categories: Game En-

gines and Flight Simulators.

In game engines, the most important aspect is an

appealing visualization. Flight Simulators have the

main focus in the ﬂight simulation itself. There is

some focus on the visual aspect as well, but the main

efforts are dedicated to aerodynamics and ﬂight fac-

tors present in real world.

3.1 Game Engines

First-person games offer a possibility to researchers

play, test, and evaluate their robots with a different

range of sensors. An accessible 3D environment with

physics and kinematics opens new robotic research

perspectives. Conversely, when aerodynamics is the

main focus, game engines normally do not attend the

requirements (Lewis and Jacobson, 2002).

3.2 COTS Flight Simulators

There is a great variety of COTS Flight Simulators.

The graphics in modern ﬂight simulators are very

realistic. However, scientiﬁc research has to worry

about the ﬂight dynamics of an aircraft. Therefore, an

evaluation on a few low cost COTS ﬂight simulators

is presented below.

3.2.1 Microsoft Flight Simulator

The current version of the Microsoft Flight Simula-

tor series (FSX) is probably the most realistic on the

market, in graphical terms, in part thanks to the use

of the new DirectX 10 technology. Its ﬂight model

is based on a set of tables, and is independent of the

visual model (Stock, 2007). The failure-modeling in-

cludes equipment failure, but without variation of the

manner in which the equipment will fail.

FSX presents the SimConnect API, an FSUIPC-

like access to functions and variables, which allows

developers to create new add-ons to add or replace

functionalities, as well as monitor activities (Mi-

crosoft Corporation, 2006).

3.2.2 X-Plane

X-Plane uses the geometric approach and its model is

deﬁned as structural. An engineering process called

”blade element theory” calculates the ﬂight model,

breaking the aircraft down into many small elements

and then ﬁnding the forces on each element, which are

then converted into accelerations, velocities and po-

sitions (Laminar Research, 2007). X-Plane also has

failure-modeling, but as in FSX, it is not allowed to

vary the way the equipments fail (McManus et al.,

2003).

A creditability factor about X-Plane is its profes-

sional use and approval by the FAA for training to-

wards airline transport certiﬁcate. However, Stock

(Stock, 2007) compared X-Plane and FSX, and the re-

sults show how some conceptual ﬂaws, and how they

have serious impressions in their models.

3.2.3 FlightGear

FlightGear is a free, open-source, multi-platform, co-

operative ﬂight simulator development project, with

the goal of creating a sophisticated ﬂight simulator

framework for use in research or academic environ-

ments, once it allows access to a very large number

of internal state variables (FlightGear, 2007). It al-

lows to remotely control FlightGear from an external

script and to use an external ﬂight dynamics module

(including hardware autopilot) (Aeronautical Devel-

opment Agency, 2001).

FlightGear allows the users to choose between

three primary Flight Dynamics Models. More than

this, it is possible to add new models or even inter-

face to external ﬂight dynamics models source, such

as from Matlab. The three available models are JS-

BSim, YASim and UIUC (based on LaRCsim origi-

nally written by NASA (Jackson, 1995)).

FLIGHT SIMULATION ENVIRONMENTS APPLIED TO AGENT-BASED AUTONOMOUS UAVS

245

3.2.4 Piccolo

Piccolo is a known auto-pilot system for small air-

craft, by CloudCap Technology (CloudCap Technol-

ogy, 2007). The main product is commercialized as

hardware components that can be mounted on a small

aircraft, allowing autonomous ﬂight. There is also

a software release of a simulation environment, for

software-in-the-loop tests. The graphical component

is minimal, but in terms of ﬂight simulation it is very

realistic, simulating many of the forces involved in a

ﬂight. It can also, by default, output its data to Mi-

crosoft Flight Simulator and FlightGear.

3.3 Dedicated Middleware Engines

There are hundreds of middleware engines which

could be used in scientiﬁc research. But, consider-

ing the UAV scenario controlled by multi-agents al-

gorithms, two are highlighted in this paper.

OpenFlight, MultiGen-Paradigm’s native 3D con-

tent, is the leading visual database standard in the

world and has become the standard format in the vi-

sual simulation industry (MultiGen-Paradigm, Inc.,

2007). AeroSim is a Matlab block library which pro-

vides components for development of nonlinear (with

six degrees of freedom) aircraft dynamic models (Un-

manned Dynamics, 2007).

4 CONCLUSIONS

Since the choice of an appropriate environment de-

pends on speciﬁc needs, no single environment is

elected, but the choices of a suitable environment

have been deﬁned concerning speciﬁc characteristics

chosen according to the objective in sight.

Some conclusions can be withdrawn from the

analysis made to the simulating environments in the

previous section, where characteristics like simulation

of graphical presentation, kinematics, physical inter-

pretation of the object, weather simulation and sim-

ulation cycle method were considered. In terms of

software openness, each presents its advantages, FSX

presenting for the ﬁrst time an open API, and other

products with their long-time known interfaces. This

ﬂexibility allows researchers to choose the combina-

tion of simulator modules that best suits their inter-

ests, thus being able to focus on the development or

improvement of either the graphical content, the air-

planes’ model, or, in our case, the creation of auto-

mated aircraft able to communicate with one another

and coordinate actions.

A variety of software able to simulate dynamics

and graphical interfaces challenges the researchers in

order to choose the best solution to their investiga-

tions. Further works need to better related and param-

eterize the solutions in deﬁned scientiﬁc parametersin

order to assist choices in new researches.

REFERENCES

Aeronautical Development Agency (2001). Pcs in ﬂight

simulation research the lca (navy) experience.

Air & Space Europe (1999). Civilian applications: the chal-

lenges facing the uav industry. Air & Space Europe,

1(5):63–66(4).

Allouche, M. (2000). The integration of uavs in airspace.

Air & Space Europe - Operations and Safety, 2:101–

104(4).

CloudCap Technology (2007). Piccolo system soft-

ware. Web Page. Available online at http://

www.cloudcaptech.com/resources autopilots.shtm.

FlightGear (2007). Flightgear ofﬁcial site. Web Page. Avail-

able online at http://www.ﬂightgear.org/.

Jackson, E. B. (1995). Manual for a workstation-based

generic ﬂight simulation program (larcsim), version

1.4. Technical Report NASA-95-tm110164, NASA.

Laminar Research (2007). X-plane ﬂight simulator. Web

Page. Available online at http://www.x-plane.com.

Lewis, M. and Jacobson, J. (2002). Game engines in scien-

tiﬁc research - introduction. Communications of the

ACM, 45(1):27–31.

Lian, Z. and Deshmukh, A. (2006). Performance

prediction of an unmanned airborne vehicle multi-

agent system. European Journal of Operational

Research, 127(2):680–695. Available online at

http://ideas.repec.org/a/eee/ejores/v172y2006i2p680-

695.html.

McManus, I. A., Greer, D. G., and Walker, R. A. (2003).

Uav avionics ’hardware in the loop’ simulator. In 10th

Australian International Aerospace Congress, Pro-

ceedings of, Brisbane, Queensland, Australia.

Microsoft Corporation (2006). Simconnect sdk reference.

Web Page. Available online at http://www.fs-seine-

75.com/SDK/Corennect.htm.

MultiGen-Paradigm, Inc. (2007). Openﬂight ofﬁcial site.

Web Page. Available online at http://www.multigen-

paradigm.com/products/standards/openﬂight/ in-

dex.shtml.

Schiff, B. (1978). Flying, A Golden Science Guide.

Stock, C. (2007). Flight dynamics: Fsx and x-plane battle

it out. Internet-based Magazine. Available online

at http://www.simpilotnet.com/index.php?option=

com content&task=view&id=20&Itemid=9.

Unmanned Dynamics (2007). Aerosim block-

set. Available online at http://www.u-

dynamics.com/aerosim/default.htm.

ICEIS 2008 - International Conference on Enterprise Information Systems

246