Skip to main content

Full text of "A Multi-Operator Simulation for Investigation of Distributed Air Traffic Management Concepts"

See other formats


A Multi-Operator Simulation for Investigation of 
Distributed Air Traffic Management Concepts 

Mark E. Peters,* Mark G. BallmJ John S. Sakosky* 
Hampton, VA 23666 



Abstract 

This paper discusses the current development of 
an air traffic operations simulation that supports 
feasibility research for advanced air traffic 
management concepts. The Air Traffic 
Operations Simulation (ATOS) supports the 
research of future concepts that provide a much 
greater role for the flight crew in traffic 
management decision-making. ATOS provides 
representations of the future communications, 
navigation, and surveillance (CNS) 

infrastructure, a future flight deck systems 
architecture, and advanced crew interfaces. 
ATOS also provides a platform for the 
development of advanced flight guidance and 
decision support systems that may be required 
for autonomous operations. 



Introduction 

The Air Traffic Operations Simulation (ATOS) 
is a mid-fidelity simulation under development 
to support research of air traffic operations 
within future airspace environments. Hosted by 
the Air Traffic Operations Laboratory at the 
NASA Langley Research Center, ATOS is a 
workstation-based human-in-the-loop simulation 
that serves as a test bed for investigations of 
future distributed air/ground traffic management 
concepts and their associated decision support 
tools. This paper describes the ATOS capability, 
its design, and its application to advanced traffic 
management research. The paper first presents a 
brief background of the future concepts under 
study and describes the simulation capabilities 

Copyright © 2002 by the American Institute of Aeronautics 
and Astronautics, hic. No copyright is asserted in the United 
States under Title 17, U.S. Code. The U.S. Government has a 
royalty-free license to exercise all rights under the copyright 
claimed herein for Government purposes. All other rights are 
reserved by the copyright owner. 

* Systems Engineer, Seagull Technology, Member, AIAA 
' Research Engineer, NASA Langley Research Center, 

Associate Fellow, AIAA 

* Senior Systems Engineer, SAIC, Senior Member, AIAA 



needed for their development and assessment. 
An overview of the simulation architecture is 
then provided with explanations of the design 
choices that were made. Elements of the future 
airspace system represented in the initial build of 
ATOS are then described. These elements 
include the aircraft components and their future 
avionics capabilities, the airborne and ground- 
based communications, navigation, and 
surveillance / air traffic management 
(CNS/ATM) infrastructure, and future system 



Background 

The Distributed Air Ground Traffic Management 
PAG-TM) concept, proposed by NASA, 
describes a set of future traffic operations that 
redefines the roles of flight crews, air traffic 
service providers, and aeronautical operational 
control organizations.^ It is based on the premise 
that the sharing of information between these 
system participants and the delegation of 
decision authority to the most appropriate 
decision maker will result in large improvements 
to airspace system capacity and robustness. ^ 
Distributed decision making authority may also 
be a key to enabling large increases in airspace 
user efficiency and flexibility, and in minimizing 
human workload bottlenecks that limit system 
capacity. 

DAG-TM builds upon the concept of 
"revolutionary evolution" advocated by the 
Future Air Navigation System (FANS), which 
cleared the way for a distributed air-, ground-, 
and space-based CNS/ATM infrastructure. 
DAG-TM also utilizes the idea of user-optimal 
routing efficiencies as advanced by the Free 
Flight paradigm, in which airspace users are free 
to select their path and speed in real-time. To do 
this, advanced airborne systems are needed to 
enable autonomous airborne flight planning in 
the presence of traffic, terrain, and weather 
hazards. These systems build upon the concepts 
of Traffic Collision Avoidance System (TCAS), 
Terrain Awareness and Warning System 



1 
American Institute of Aeronautics and Astronautics 



(TAWS), and WindshearAVeather radar. 
Through secure multi-layered data management 
supported by systems such as Active Dependent 
Surveillance-Broadcast (ADS-B), Controller 
Pilot Data Link Communications (CPDLC), 
Flight Information Service (FIS), and Traffic 
Information Service (TIS), an increased 
awareness of hazards and constraints should 
allow crews to have greater authority and 
responsibility in managing their flights. 

The simulation of airspace and the operations of 
aircraft within it is a useful and common 
approach for the study of new air traffic 
management concepts, decision support 
automation, and procedures. For instance, the 
Pseudo Aircraft System was developed to 
facilitate the development of the 
Center/TRACON Automation System.^ 

Additionally, the Federal Aviation 

Administration (FAA) uses its Target Generation 
Facility to study operational air traffic control 
problems,'' which has recently been enhanced to 
include a higher fidelity dynamics model' For 
the study of future concepts such as DAG-TM, a 
traffic operations simulation is needed to study 
the interactions of distributed human decision 
makers, especially regarding the transfer of 
responsibility between airborne and ground- 
based human operators. It also facilitates the 
study of supporting systems, such as decision 
support tools and future CNS infrastructure, and 
their impacts on concept feasibility. As 
feasibility requirements and bounds are 
established, traffic operations simulation is 
needed to support detailed design and evaluation 
of air and ground-based decision support 
automation, computer/human interfaces, and 
operational procedures. When these activities are 
sufficiently mature, traffic operations simulation 
can facilitate the assessment of concept benefits 
and economic viability. 

Simulation Requirements for Research 

There are 6 main requirements that a simulation 
must meet to be useful as a tool for (CNS/ATM) 
research and development: 



Multiple Human Operators 

The simulation must support multiple human 
operators, each of which is provided a level of 
sophistication and fidelity that enables them to 
make traffic management decisions as they 
would in the actual environment. 



System Flexibility with Variable Fidelity 

The simulation must be flexible enough to 
accommodate the different levels of fidelity and 
different levels of equipage. Workstation 
simulations are usually adequate to simulate 
multi-aircraft scenarios with human operators in 
an affordable and manageable way, however, 
high-fidelity simulations are occasionally desired 
to duplicate a full-crew and full-workload 
environment, and to aid in the development of 
procedures. The CNS/ATM infrastructure 
representation must also be flexible to compare 
future CNS concepts with today's system. Since 
the future CNS infrastructure will also be 
required to accommodate many legacy systems, 
a range of present and future CNS components 
and aircraft equipage levels must be 
accommodated simultaneously. 

Simulation Repeatability 

For proper experimental control, the simulation 
should produce repeatable results. Test subjects 
must be provided with identical situations, so 
that only test subject actions produce differing 
results. Additionally, some experiments will test 
algorithms, such as traffic conflict resolution, 
which can produce large response differences to 
small input differences. Simulation repeatability 
for functions such as these is very important for 
the verification and validation of algorithms. 

Modes Of Operation 

The simulation must employ two modes of 
operation. These are: human in the loop (HITL) 
and human operator modeled batch. HITL is 
defined as a mode in which subject pilots, 
subject controllers, and/or subject dispatchers 
interact with systems or simulators that allow 
them to make flight-management or traffic - 
management decisions as they would in a 
deployed system. Human operator modeled 
batch is a mode in which all system operations 
are simulated with full repeatability, using 
automated human operator models to represent 
humans in the simulated system. 

Component Error Modeling 

Success of DAG-TM and other future concepts 
hinges upon the quality of information available 
to the decision maker. Required Navigation 
Performance (RNP) compares actual aircraft 
position against required position. Required 
Communication Performance (RCP) states 
operational performance in terms of 
communication process time, integrity. 



American Institute of Aeronautics and Astronautics 



availability, and continuity of function. Required 
Surveillance Performance (RSP), yet to be 
defined, will standardize requirements for 
surveillance accuracy. Parametric control of 
RNP, RCP, and RSP levels as well as the 
accuracy of system predictions is needed to 
understand the impacts of the associated errors 
on concept feasibility. The simulation must also 
independently generate and track information 
that represents actual (truth) values from that 
which represents values available to simulation 
components and participants. 



Simulation Execution Control 

Simulation execution control is normally used in 
flight simulators that utilize a real-time 
simulation environment. To be compatible with 
such simulations, several simulation mode 
control functions must be accommodated. These 
are: initialize, trim, hold, operate, and reset. 



A new version of the simulation is under 
development to provide higher fidelity 
representations of system elements and improved 
utility to researchers and technology developers. 
Referred to as ATOS Build 1, it resolves several 
limitations of the prototype simulation. Among 
the limitations was the reliance on target 
generators for all aircraft other than the test 
subject stations. Target generators were 
developed to enable many simulated aircraft to 
be controlled by pilot operators (referred to as 
"pseudo-pilots"), who in turn receive voice 
clearances and instructions from ground-based 
human controllers. However, the aircraft 
represented by these simulations normally have 
no capability to avoid conflicts, and it is 
impractical to expand their capability. These 
simulations, while appropriate to support 
ground-based air traffic control research, do not 
adequately support DAG-TM research 
requirements for large numbers of aircraft and 
their systems that enable autonomous operations. 



Simulation Evolution 

The development of ATOS, formerly known as 
Free Flight Simulation,' was initiated after an 
investigation of existing traffic management 
simulations revealed that none met the above 
goals. Existing real-time simulations were not 
flexible enough to model future as well as 
current operational environments and they lacked 
the mode control needed for rigorous and formal 
research. Existing batch assessment tools were 
not suitable because they do not represent human 
decision-making. They also lack the detailed 
flight deck and service provider representations 
required for design of decision support 
automation tools. 

Initial goals for the simulation also included an 
aggressive development schedule that would 
provide research support as soon as possible. 
Toward this goal, a prototype simulation was 
developed that made much use of existing 
capabilities. It allows several workstation-based, 
real-time aircraft simulators to interact with each 
other and with target aircraft generated by a 
central airspace simulation. Existing software 
codes developed by NASA and National 
Aerospace Laboratory of the Netherlands (NLR) 
were modified to create the initial prototype,'^' ^ 
which has been used to investigate the use of 
traffic intent information in constrained air- 
traffic operations^. 



Design Overview 

In addition to the identified research 
requirements, which focus on the research utility 
of the simulation, several software design goals 
were set to facilitate scalability and ease of 
maintenance: 

Maximize simulation scalability, so that 
large numbers of aircraft can be 
simulated without exceeding 

computational speed limitations of any 
process 

Develop a simulation architecture that 
avoids the development and 
maintenance of duplicate functions 
Include a capability to duplicate and 
distribute simulation components 
among several facilities, via dedicated 
connections and possibly via the 
internet 

Require few or no changes to flight 
deck decision support technology to 
interface with ATOS, Langley high- 
fidelity simulations, and Langley 
research aircraft such as the NASA 
Airborne Research Integrated 

Experiments System (ARIES), a 
specially instrumented Boeing 757 that 
is used to support in-flight validation 
activities. 



American Institute of Aeronautics and Astronautics 



ATOS Build 1 is an extensible distributed 
system consisting of many independent 
processes that operate and communicate with 
each other through the network. The Build 1 
system components are illustrated in Figure 1. 
The system architecture is flexible to allow for 
experimental investigations of competing 
systems developed for autonomous operations. 



Target 
Generator (TG) 

- Background 
aircraft 



Raw ADS-B, 
Truth aircr^t state 
► / 



Simulationicontrol, 
truth wind4 



\ ASTORpseudo- 
\ pilot commands 

\ 1z^ 

\ ASTOR states, 
Emulation control 



Pseudo-Pilot 

Interface 

(PPI) 



Flight 
Information 
Service (FIS) 

- Polygon weather 

- Winds/temps aloft 
-SUA 

- D-ATIS 



Truth winds, 
weath^, 
simulathn coi 



Simulation 
Manager (SM) 

- Scenario initialization 
-Mode control 

- Timing control 



Monitd[ (heartbeat(s), 
errors)\ truth aircraft 

- Simulation Monitoring sfafe 

- Heafth Monitoring 

- Truth winds 



Raw ADS- B 
Truth aircraft 



Simulation control 



Autonomous Aircraft 

(pilot, pseudo-pilot, or pilot 
model controlled) 



Data Link Simulator 
(DLS) 

- Mode S Model 

- UAT Model 



FflS-B, 
Truth \pinds, 
pseudp-pilot 
comriiands, 
simulation hontrol 



Truth winds, ^Distribute! 
""""' \ Network' 



\ / 



^ ads-b" 



ASTOR 

Sensor Package 
Displays 
Control Panels 
Aircraft Model 
Pilot Model 



E3 



a: 



Advanced Flight Deck 
Decision Support Tools! 



Figure 1. Illustration of the distributed system, 
which enables flexibility and extensibility 

An independent aircraft simulation, referred to as 
the Aircraft Simulation for Traffic Operations 
Research (ASTOR), is employed for each 
aircraft represented within ATOS Build 1 . These 
workstation-based, HITL aircraft simulations 
have 6 degree of freedom dynamic models 
supported by actual aircraft aerodynamic data. 
The dynamics model can simulate jet and piston 
aircraft, and the simulation is equipped with 
representative cockpit displays and equipage 
levels for the different types of aircraft. 
Furthermore, the models are extensible to 
different aircraft types by simple 
aerodynamic/propulsion parameter changes. 

The aircraft simulations can be executed with 
pilot test subjects in HITL operation, or pseudo- 
pilots can drive them as background traffic. This 
form of background traffic is much more capable 
than target generated traffic and can be expected 
to have the same behavior as the HITL aircraft. 
If necessary, additional traffic can still be 
provided by target generation software for large 
simulations, since the currently existing target 



generation capability is not compromised. With 
the addition of a pilot model for each aircraft 
simulation, background traffic can be 
represented without the need for pseudo-pilots, 
or the entire simulation can be executed in a 
batch mode to support Monte-Carlo analysis. 

ASTOR also models a generic future flight deck 
architecture that contains air/air and air/ground 
datalinks, and advanced crew interfaces, and it 
hosts the flight management systems and 
decision support systems required for 
autonomous operations. These processes are 
independent from the aircraft simulation code 
itself, to provide accurate representations of 
aircraft equipage. The simultaneous use of 
various versions of ASTOR enables the 
investigation of operations involving various 
aircraft types with differing equipage levels as 
well as widely varying performance and 
operating characteristics. 



Simulation Architecture 

To support such generalized simulations 
involving a large number of independent 
systems, a sophisticated software network is 
required. 



HLA (High Level Architecture) 

The ATOS infrastructure is based on the 
Department of Defense's (DoD) High Level 
Architecture (HLA)'. The use of HLA to 
implement the interfaces between the ATOS 
components enables a widely distributed 
simulation limited only by the available 
hardware resources on the network. Each ATOS 
component, known as a Federate in the HLA 
paradigm, communicates through the distributed 
network, eliminating the need for centralized 
control of data flow that has in the past limited 
the extensibility of the simulation. Furthermore, 
the use of HLA enables the inclusion of other 
simulation assets such as full cockpit simulations 
that require remote access. 

Federation Components 

Each ATOS component is a separate software 
process that usually operates on hardware 
dedicated for that process. These processes 
include aircraft simulations, airport models, and 
target generators. The simulation manager is a 
federate that is used to manage the simulation. 
The only non-federate is the Run Time 



American Institute of Aeronautics and Astronautics 



Infrastructure (RTI) exec, an HLA process which 
manages HLA communications and other HLA 
specific tasks. Figure 2 iUustrates the 
components of the Federation. 



Manages 

federates and 

monitors federate 

health 



SIM MANAGER RTI EXEC 

^^ h , 



Supports HLA 
message traffic 
within the federation 



1 Standard Network 



n — L 



FEDERATE 1 FEDERATE 2 




FEDERATE n 



t v 



Hardware resources for running multiple fee _ 



Figure 2. lUustration of a Federation, the 
simulation architecture when using HLA 



particular federate through its real-time scripting 
capability. The scenario and 



simulation time 
display 



Buttons for pop-up 
windows 

File narrie 
entry 

Real/Fast 
Time and 
batch control 

Mode control 

Federate mode 
status and health 
display 



Simulation 

message 

display 



Snapshot 
control 




»»Ste«J 



RTI Executive and Interactions 

The Real Time Interface (RTI) executive, which 
is part of HLA, monitors the joining of the 
federation by federates and manages the various 
messages that can be sent within the federation. 
Within the HLA paradigm, these messages are 
referred to as interactions. The number and types 
of interactions are defined at the beginning of the 
simulation by the RTI exec and its associated 
data files. Each federate subscribes to a subset of 
the total available interactions when it is 
initialized. The federate also agrees to publish 
certain types of interactions. Examples of these 
interactions include aircraft state information and 
simulation state modes. The end result is that the 
any federate can send or receive information 
to /from any other federate directly. 



Simulation Manager 

The simulation manager is the primary 
simulation control and monitoring station. The 
simulation control window, illustrated in Figure 
3, controls the simulation state and monitors the 
health status of all the federates during the 
simulation. It can command the federates into the 
various states such as reset, hold, and operate. 
The second window, the scenario and event 
window, can record important events, collect 
data, and send specialized instructions to a 



Figure 3. Detailed explanation of the simulation 
control window 



event window is pictured in Figure 4. The 
scenario manager also has a planview display 
which allows the operator to monitor the airspace 
being modeled. 



Simulation Modes 

The simulation manager controls transition to 
and from various system modes. Figure 5 shows 
the simulation modes and the possible transitions 
between them. The modes include: Configure, 
Reset, Trim, Hold, Operate, and Terminate. 

Configure 

The CONFIGURE mode is the mode that the 
federate application enters after being started. 
The CONFIGURE mode cannot be transitioned 
to from any other simulation modes. Configuring 



American Institute of Aeronautics and Astronautics 



:::TftR^CT:FEDER»T:B:i:i: 



iiiitRWJiaMissfGifliMGifliiraRi 



aeiecTFeE'ep^Ts 




!»W»WSi»S» 



^^^mmm 




Figure 4. Illustration of the Scenario and Event 
window 



consists of initializing federates and all non- 
scenario related data. 



CONFIGURE 



Reset 



Trim 



[Reset 



Fre eze I ^j^ m 



Snapshot ^i 
Take '^ 



OP 



Hold 

=1* 



OP 



m^ 



TERMINATE 



Sliutdown 



Figure 5. Simulation Modes and Transitions 



Reset 

The RESET mode applies or reapplies initial 
conditions. 



Trim 

The TRIM mode allows federates to reach a 
steady state before running the current scenario. 
For aircraft federates, the TRIM mode may 
involve determining the control surface 
deflections and power settings required to reach 
equilibrium. For other federates that do not need 
to achieve a steady-state condition, TRIM may 
simply be a "do nothing" mode. 

Hold 

The HOLD mode is one where the federates 
have been initialized and trimmed and is ready to 
run the scenario. The HOLD mode also occurs 
when an active simulation has been temporarily 
suspended. In the HOLD mode, the simulation 
clock does not advance, however, user interface 
displays are allowed to change 

Operate 

The OPERATE mode is where the federates are 
running the simulation scenario. 

Snapshot 

The SNAPSHOT feature saves the current state 
of the system to disk so that it can be used as the 
starting point of a future scenario. Each 
component is required to identify and save its 
own mode information needed to meet this 
objective. 

Terminate 

The TERMINATE mode shuts down the 
federates. Once the component has exited, it will 
need to be re-launched in order to be used again. 



Health Monitoring 

The simulation manager maintains a record of 
the system health. To do this, each federate sends 
a periodic message indicating its current state 
and its current health status. There are three 
health statuses: Normal, Warning and Failed. 
The federate status is presented to the operator 
on the simulation control window in green, 
yellow, or red indicator lights. If a federate 
completely fails, and is unable to send any 
messages back to the simulation manager, the 
simulation manager flags the federate as failed 
and posts a red indication after a specified 
number of missed messages. 



Figure 6 illustrates a failure of one of the 
federates. In this case, the overall system status 



American Institute of Aeronautics and Astronautics 



row indicates a failed federate somewhere within 
the simulation. The operator is then alerted to 
determine which federate has failed by scrolling 
through the federate list, or by reading failure 
messages within the message box. 




i i iii?^^iiiiii l ii in??;;?iiiiiil iii??;??^^mii J 




Figure 6. Illustration of the simulation manager 
indicating a failure on ASTOR 2 



Real Time Scripting 

Under certain conditions, the operator may wish 
to send a specific set of instructions to a federate 
while the system is running. Examples might 
include sending a new route to an aircraft 
simulation federate, or a different data file name 
to a FIS federate. This task is accomplished 
using the top portion of the scenario and event 
window as illustrated in Figure 7. The operator 
can choose a federate to receive the message 
from the federate list. Then a string is typed in 
the message box. When the string is sent to the 
appropriate federate, the simulation manager 
responds with two confirmations. These are: 1. 
The message was received, and 2. The message 
was understood and complied with. 



Event Marking 

The scenario and event marker also enables the 
operator to mark interesting or noteworthy 
events. The operator marks the event and then 



types a short message indicating the significance 
of the event. This feature enables the researcher 
to easily find the spot during data analysis. 



Federate which will Monitors the success Used to select the target 
receive the script of the transmission federate, the federate which 
message .: will receive the command 



•|i«i?8iii8»«l 



: TMpiierFECiff-vre 



SELECT FED£P"TE 



IAST0R3 

TFA|J^Wi;:-l'.'IJM','NlTCiF 




^STORI 
AST0R2 
(\ST0R3 
'\ST0R4 
AST0R5 
iftSTORS 



CQM.WNDJS 

PECftLi m 



New Route MOLV38RNLV16 



Monitors whether or not the 
federate could execute 
the command 



Enables the user to page through 
old commands and use them as 
tht basis for new commands 



Contains the command 
sent to the federate 



Sends the command to the 
federate 



Figure 7. Illustration of the real time scripting 
feature in the scenario and event window 



Note Posting 

The operator can also take notes with a small text 
editor that is positioned in the middle of the 
Scenario and Event window. These notes are 
posted to the data file when they are submitted 
and to not pertain to any particular event. 

Data Collection 

Each federate is responsible for collecting its 
own data. The simulation manager is responsible 
for instructing the federates as to whether data is 
to be collected. The simulation manager collects 
its own data which consists of all messages 
received from the federates, all notes, and posted 
events. 

Operating Modes 

The ATOS software supports an emulated real- 
time and a batch mode of operation. 

Emulated Real Time 

The emulated real-time mode is equivalent to 
real-time (or synchronous) operation, except that 
a real-time operating system is not used. In 
emulated real-time mode, a frame overrun is not 
an error. Components can attempt to "catch up" 
by immediately re-executing the task instead of 



American Institute of Aeronautics and Astronautics 



waiting for the next clock tick. The emulated 
real-time operating mode has two sub-modes: 
human in the loop (HITL), as illustrated in 
Figure 8, and fast time. In HITL mode, the 
simulation time always advances at a normal 
real-time rate, i.e. at the same rate as a wall 
clock. The fast-time mode is implemented as the 
"speed times N function" defined in ARINC 
610A^°. 



Subject Pilot 



Pseudo Pilot Station 




Pseudo Pilot Station 



Figure 8. Illustration of HITL mode consisting 

of subject pilot workstations with high fidelity 

background traffic provided by pseudopilots 

Batch Mode 

In batch mode, the executions of periodic tasks 
are not synchronized to a clock. Periodic tasks 
run at the fastest rate that the underlying 
hardware and operating system allows, or the 
clock tick is substituted for an asynchronous 
event. Unlike emulated real-time operation, 
event-driven tasks may be allowed to run 
without preemption by the periodic tasks. Batch 
mode is illustrated in Figure 9. 



Batch Pilot Station 



Batch Pilot Station 




BatchPiiotStatior 



Figure 9. Illustration of batch mode 
functionality 

While the simulation architecture supports batch 
and fast time, not all federates may be able to do 
this. The use of these tools is dependent on the 
nature of the individual federates that comprise 
the simulation. 



Time Synchronization 

The simulation manager controls the master 
clock of the simulation and periodically sends 
out a time-stamped message to all of the 
federates. This message is used by the federates 
to insure they are synchronized with the master 
clock. 

Aircraft Simulations 

While any aircraft simulation that complies with 
the basic federate infrastructure of ATOS can be 
used, the primary aircraft simulation is ASTOR. 
ASTOR supports four objectives: 

• Advanced mid-fidelity aircraft simulation 
with enough sophistication to study DAG- 
TM concept feasibility while remaining 
simple enough to execute on a workstation 
platform. 

• Modular architecture, which facilitates 
simulation of various aircraft types, each 
with unique airframe/engine and 
autopilot/autothrottle models, flight deck 
displays/controls, and various avionics 
equipage levels. 

• Advanced avionics modeling 
representative of the future CNS/ATM 
environment. 

• Capable of operating in several 
configurations to support research requiring 
piloted aircraft, pseudo-piloted aircraft, and 
aircraft with modeled pilots (batch mode). 

Mid-Fidelity Workstation Based Simulation 

ASTOR was conceived as an instance of the 
Langley Standard Real-Time Simulation 
(LaSRS) framework^ \ LaSRS has successfully 
been applied to a spectrum of aircraft simulations 
ranging from mid-fidelity workstation-based 
general aviation simulations to high-fidelity full- 
motion cockpit air transport category 
simulations. As such the flexibility of this 
framework is proven. 

The DAG-TM concept may rely on advanced 
trajectory management functionality provided by 
the autonomous operations planner (AOP),^^ 
which continuously provides a conflict free route 
to the crew, and real time flight management 
(FMS) subsystems. The FMS guidance function 
may need to provide provides continuous closed 
loop (temporal and spatial, i.e., 4D) guidance 
commands to the auto -flight /auto -throttle 
systems. ASTOR must host these advanced 
functions for each simulated aircraft that is 
equipped for autonomous operations. 



American Institute of Aeronautics and Astronautics 



Matched performance across the aero / mass / 
propulsion, auto-flight / auto-throttle and 
AOP/FMS models is essential for viable DAG- 
TM research. This need for matched and 
appropriate levels of representation for traffic 
management research in turn drives the need for 
the accurate modeling of climb, cruise, idle- 
thrust descent, and final approach phases of 
flight; dynamic aircraft mass as a function of fuel 
bum, which affects climb and descent 
performance; and throttle/thrust lever position, 
altitude, and Mach effects on engine dynamics. 



properties, propulsion, flight controls, displays, 
and avionics have been abstracted from the 

generic six degree of freedom ASTOR 
simulation framework. Models for a large array 
of aircraft types with widely varying 
performance and equipage levels are in 
development. 

Figure 10 through Figure 12 show the ASTOR 
Flight Deck displays, which are representations 
of currently available equipment enhanced for 
autonomous operations. 



Modular Architecture 

To meet the me ta- simulation requirements 
associated with DAG-TM research and airborne 
technology development, all aircraft type- 
specific features such as aerodynamic data, mass 






£5 jisaaij 







Figure 10. MDl 1 ASTOR Pilot Station 




Figure 11. Boeing 777 ASTOR Pilot Station 



American Institute of Aeronautics and Astronautics 




■ill ^EJ i:3 Js^ 




tfi^^i ii ^^ii-"^-.^ 



Figure 12. Lancair C300 ASTOR Pilot Station 



Advanced Avionics Modeling 

The ASTOR design focuses on a modular 
avionics architecture which achieves a high level 
of functional integration in the spirit of state of 
the art integrated modular architectures for 
avionics as employed in the commercial Boeing 
777 and military F22 designs. ^^ 

This additional level of fidelity supports a 
fundamental research goal to enable ASTOR to 
not only assist in concept level development, but 
also provide insight in to the design of the actual 
hardware that would be required in a real flight 
deck. 

ASTOR bridges the gap between system-level 
design / assessment tools and detailed subsystem 
designs, which are needed to support 
implementation of future concepts, but often can 
not easily be tested in a distributed environment. 
ASTOR provides an appropriate level of 
modeling in terms of information transfer, and 
information quality (RNP, RCP, RSP). These 
attributes enable the prototyping of various flight 
deck concepts with a level of operational realism 
that can address feasibility, robustness, and 
certification concerns as necessary for the 
validation of the resultant research data. This 
type of interaction between subsystems is critical 
to understanding the detailed flight deck design 
requirements. 



The flexible nature of the avionics architecture 
enables ASTOR to represent a spectrum of actual 
aircraft equipage levels. Fundamental design 
components (e.g. GPS, IRS, FCC, TMC, and 
FMS) can be individually selected for a 
particular equipage level. These interfaces 
between components are also explicitly modeled 
and represent realistic avionics architectures at a 
high level. The flexibility of ASTOR enables 
researchers to consider/understand/model both 
legacy systems and industry modernization plans 
such as ARINC660A. 



Integrated CNS Avionics Architecture 

The advanced autonomous flight deck features 
embodied in the AOP and FMS subsystems are 
integrated into a CNS/ATM avionics architecture 
compatible with ARINC660A. The ARINC 
660A CNS/ATM standard provides key design 
guidance for CNS functional partitioning. The 
ARINC 429 BITS data bus is simulated for the 
widest range of reverse and forward 
compatibility. 

Figure 1 3 illustrates the integration of AOP and 
FMS elements of the Future Autonomous Flight 
Deck into an ARINC 660A CNS/ATM avionics 
architecture. 



10 
American Institute of Aeronautics and Astronautics 



Kasv SsttKiM 



lilCAS 




i'W rNK/ATM Avinmcs F^MMisoEial Archifecfia'e 



Figure 13. ASTOR Architecture of the Future Autonomous Fhght Deck 



Avionics Bus Fault Modehng 

FuU-featured error and failure modes such as 
total failure, signal bias failure, intermittent 
signal failure, and random noise failure are 
modeled within the Avionics Bus package. 



ADS-B Datalink Simulations 

The ADS-B subsystem allows the transmission 
of on board data to air or ground based users via 
a data link (e.g. Mode S, VDL-4, UAT) using a 
broadcast mode. ADS-B equipped aircraft and 
vehicles automatically broadcast important 
information — latitude and longitude, velocity, 
altitude, heading, identification and, optionally, 
intent — as determined by the avionics on board. 
The ATOS simulation provides several datalink 
simulation (DLS) options. These simulations are 
based on a mathematical datalink model and 



implemented using a distributed software 
architecture. 

Centralized and Distributed Solutions 

There are trade-offs between a centralized DLS 
model and a distributed DLS model. The DLS 
architecture can be either centralized, where a 
single process collects all messages and then 
sends a subset of the messages to each federate 
aircraft, or it can be distributed, where each 
federate must simulate the DLS for itself Figure 
14 and Figure 15 illustrate the centralized and 
distributed solutions respectively. 



11 
American Institute of Aeronautics and Astronautics 



ASTOR 
1 



Raw 
ADS-B 




Simulated 
Datalink 
ADS-B 



•J-D y 



Centralized 
DLS 



ASTOR 
n 



Raw 
ADS-B 



Figure 14. Illustration of a centralized DLS 



Computational Comparison 

When simulating the DLS in a centralized 
architecture, all messages must be gathered at a 
central location and then rebroadcast according 
to the algorithms within DLS model. The 
centralized approach is simple because only one 
process is needed, but it can become a simulation 
bottleneck that requires very high computational 
speed if the number of aircraft becomes large. 
The design limits the scalability of the system. 

The distributed approach does not reduce the 
amount of computation required, but rather it 
distributes the computational load to all the 
aircraft. This eliminates a computational 
bottleneck on a single process but 

Decentralized 
Network 



ASK 
1 


3R 


Raw 
ADS-B 


i\ 


Raw 
ADS-B 


AS' 


FOR 








^ 








Decentralized 
DLS 


< 

Raw 


— ► 
Raw 


Decentralized 
DLS 














ADS 


\ ADS 


-B 







Figure 15. Illustration of a distributed DLS 



creates a more complex software architecture 
since each aircraft now must incorporate a copy 
of the DLS model within its own process. 



Network Load Comparison 

To determine the network traffic of distributed 
and centralized systems, a series of experiments 
were performed between test processes designed 
to send and receive ADS-B messages. Two 
configurations were created. These were: 

• 30 test aircraft and 500 background 
aircraft (generated with a target 
generator) 

• 60 test aircraft and 500 background 
aircraft 

An simplified data link representation was used 
(every aircraft sees every other aircraft) with an 
evenly distributed network usage (no spikes). In 
the first test, all messages were passed through a 
centralized processor. In the second test, 
messages were sent through a distributed 
network. The results are shown in Table 1 . When 
using HLA in conjunction with Transmission 
Control Protocol (TCP), the centralized and 
distributed systems had nearly the same load 
with the centralized load being a bit more 
efficient. When User Datagram Protocol (UDP) 
is used however, the distributed system used an 
order of magnitude less bandwidth and required 
only a fractional increase to handle the 60 
aircraft case over the 30 aircraft case. Therefore, 
maximum extensibility is achieved using the 
distributed design with a UDP transmission. 

Table 1. Results from network load evaluation 



Design and IPC 


Usage (bits/sec) 




30 

aircraft 


60 
aircraft 


Centralized with TCP 


4.25E+7 


8.73E+7 


Distributed with TCP 


4.80E+7 


l.OOE+8 


Distributed with UDP 


2.64E+6 


2.74E+6 



Distributed UDP Approach 

From both a computational and network traffic 
point of view, the distributed solution using UDP 
is the most effective. Each ADS-B message is 
broadcast once (by each transmitting aircraft) 
and then received by every other aircraft 
federate. Then each federate is responsible for 
choosing its own subset of ADS-B messages. 
The advantage of the distributed UDP system is 
that the messages are each only broadcast once 
resulting in an nxl solution. Furthermore, delays 
that might be incurred by double transmission of 
the same message are eliminated. However, 



12 
American Institute of Aeronautics and Astronautics 



since message reception is not guaranteed, this 
approach may not guarantee the repeatabihty that 
is required. Therefore, while the UDP solution 
may provide the ultimate scalability 
performance, reliability is compromised. 



Distributed TCP Approach 

To maximize performance and reliability a 
distributed TCP approach still has the advantage 
of computational distribution while maintaining 
the message reliability of the TCP. Since 
repeatability of message transfer is critical to the 
success of the simulation, this approach is 
ultimately the one chosen. 

Flight Information Service 

Flight Information Services - Broadcast (FIS-B) 
is defined as the non-control information needed 
by pilots to operate in the National Airspace 
System (NAS) and internationally. Pilots, 
dispatchers, schedulers, and controllers all need 
accurate, timely FIS-B data to plan (or re-plan) 
and assess the execution of flight operations. 
Figure 16 illustrates the FIS-B system. FIS-B 
broadcasts weather information periodically 
through a VHF data link to the cockpit. FIS-B 
provides coverage over the continental US from 
5000 ft. AGL to 17,500 MSL, except m areas 
unreachable due to mountainous terrain. 




VDL Mode 2 



Air Traffic 
Service Provider 



FIS Provider 



Figure 16. The FIS-B System 

The actual FIS-B contains many assorted 
products. The ATOS FIS-B model contains a 
subset of these products that is relevant to the 
research to be conducted. Initially, the FIS-B 
products are the following: 

• Hazardous Weather Advisory (SIGMETs) 



• Special Use Airspace (SUA) 

• Winds and Temperatures Aloft 

Area Hazards 

Both SIGMETs and SUAs are area hazards to 
be avoided by aircraft. To represent these 
advisories, a vector-based data format is used for 
the FIS-B area hazard message. The vector 
format allows for a series of connected line 
segments or a constraint polygon region to be 
identified using connected line segments 
described with a series of constraint vertices 
(defined by Ci, C2, ..., Cn). An illustration of a 
hazard is shown in Figure 17. 




Figure 17. Example SIGMET polyhedrons 
modeled for ATOS. 



Final Remarks 

With the introduction of satellite navigation, 
ADS-B, and new CNS/ATM technologies, the 
airborne elements of the airspace system will 
have a much greater role in traffic management 
decision-making. As this role continues to 
increase in importance, ATOS will become 
increasingly more important in concept 
development and assessment, CNS requirements 
development, and the development of decision 
support technologies that support the system's 
human operators. ATOS Build 1 provides the 
foundation for an extensible and scalable multi- 
operator traffic operations simulation capability 
for investigation of distributed air/ground traffic 
management concepts of operation. Build 2 will 
introduce ground-based components and links to 
high-fidelity simulators, and Build 3 will provide 
links to research aircraft. 



Acknowledgments 

The ATOS was developed to support objectives 
of the NASA Advanced Air Transportation 
Technologies Project, a part of the Aviation 



13 
American Institute of Aeronautics and Astronautics 



Systems Capacity Program. The Small Aircraft 
Transportation System Program and the Aviation 
Safety Program have also provided support for 
development of the simulation. 



^ Concept Definition for Distributed Air/Ground 
Traffic Management pAG-TM), Version 1.0. 
Advanced Air Transportation Technologies 
Project, Aviation System Capacity Program, 
National Aeronautics and Space Administration, 
September 30, 1999. 

^ Ballm, M.G., Wmg, D.J., Hughes, M.F. 
Conway, S.R. : Airborne Separation Assurance 
and Traffic Management: Research of Concepts 
and Technology, AIAA-99-3 989, 1999. 

^ Weske, R.A., and G.L. Danek,: Pseudo Aircraft 
Systems: A Multi-Aircraft Simulation System 
for Air Traffic Control Research, Proceedings of 
the AIAA Flight Simulation Technologies. 
Conference, Monterey, CA, August 1993. 

'' Target Generation Facility (TGF) Overview, 
Aviation Simulation and Human Factors 
Division, FAA William J. Hughes Technical 
Center, Atlantic City, NJ, September 1996. 

^Peters, M. E., The Engineering Analysis and 
Design of the Aircraft IDynamics Model For the 
FAA Target Generation Facility. Tech. Report 
99162-01, SeaguU Technology, Los Gatos, CA, 
August 1999. 

^ Miles, E. S., Davis, P. D., and Wmg, D. J.: 
Development of a Free-Flight Simulation 
Infrastructure. AIAA-99-4193, July 1999. 

^ Mondolom, S., Palmer, M.T., and Wmg, D. J.: 
Development of a Prototype Airborne Conflict 
Detection and Resolution Capability. AIAA- 
2002-4446, August 2002. 

^ Wing, D. J., Adams, R. J., Barmore, B. E., and 
Moses D. : Airborne Use of Traffic Intent 
Information in a Distributed Air/Ground Traffic 
Management Concept: Experiment Design and 
Preliminary Results. 4 USA/Europe Air Traffic 
Management R&D Seminar, Santa Fe, December 
2001. 



Guidance for use of Avionics Equipment and 
Software in Simulators. ARINC Report 610A, 
February 1, 1994. 

^^ Leslie, R.; Geyer, D.; Cunningham, K.; 
Madden, M.; Kenney, P.; and Glaab, P.: 
LaSRS++: An Object-Oriented Framework for 
Real-Time Simulation of Aircraft, AIAA-98- 
4529, Modeling and Simulation Technology 
Conference, Boston, MA, August, 1998. 

^^ Ballm, M.G., Sharma, V., Vivona, R, 
Johnson, E.J., Ramiscal, E., A Flight Deck 
Decision Support Tool For Autonomous 
Airborne Operations, AIAA-02, July, 2002. 

^^Contract Task Order 4, Preliminary Design 
Review. Advanced Air Transportation 
Technologies /Air Traffic Management System 
Development & Integration, Contract No. 
NAS2-00015, NASA Langley Research Center, 
Hampton, VA, November, 200 1 . 



RTI 1.3 -Next Generation Programmer's Guide, 
Version 4. Defense Modeling and Simulation 
Office, Alexandria, VA, June, 2001. 



14 
American Institute of Aeronautics and Astronautics