![Well-clear threshold [11, 12].](https://www.researchgate.net/profile/Roberto-Sabatini-3/publication/321084002/figure/fig1/AS:562755653640204@1511182811830/Well-clear-threshold-11-12_Q320.jpg)
Content uploaded by Roberto Sabatini
Author content
All content in this area was uploaded by Roberto Sabatini on Nov 20, 2017
Content may be subject to copyright.
This is the author uncorrected pre-publication version. This paper does not include the changes arising from the revision,
formatting and publishing process. The final paper that should be used
(available at http://dx.doi.org/10.1007/s10846-017-0661-z) is:
S. Ramasamy, R. Sabatini and A. Gardi, “A Unified Analytical Framework for Aircraft Separation Assurance and UAS Sense-
and-Avoid”, Journal of Intelligent and Robotic Systems, pp. 1-20. DOI: 10.1007/s10846-017-0661-z
A Unified Analytical Framework for Aircraft Separation Assurance
and UAS Sense-and-Avoid
Subramanian Ramasamy· Roberto Sabatini· Alessandro Gardi
School of Engineering – Aerospace Engineering and Aviation Discipline, Melbourne, VIC 3000, Australia
roberto.sabatini@rmit.edu.au
Keywords: Unmanned Aircraft Systems · Collision avoidance · Trusted autonomous operations · Unified approach ·
Cooperative systems · Non-cooperative sensors · Sense-and-Avoid · CNS systems
Abstract
A novel unified analytical framework for aircraft
separation assurance and Unmanned Aircraft System
(UAS) Sense-and-Avoid (SAA) is presented. A brief
review of the state-of-the-art/state of research in
Separation Assurance and SAA (SA&SAA)
technologies is included to highlight the benefits
offered by the unified approach. In this approach, the
employment of Adaptive Boolean Decision Logics
(BDL) allows automated selection of sensors/systems
including passive and active Forward Looking
Sensors (FLS), Traffic Collision Avoidance System
(TCAS) and Automatic Dependent Surveillance –
Broadcast (ADS-B). This system performance based
selection approach supports trusted autonomous
operations during all flight phases. After describing a
SA&SAA reference architecture, the mathematical
models employed in the unified approach to compute
the overall uncertainty volume in the airspace
surrounding an intruder/obstacle are described. The
algorithms support the translation of navigation errors
affecting the host aircraft platform and tracking errors
affecting the intruder sensor measurements to unified
range and bearing uncertainty descriptors. Simulation
case studies are presented to evaluate the performance
of the unified approach on a representative UAS host
platform and a number of intruder platforms in both
cooperative and non-cooperative scenarios. The
results confirm the validity of the proposed unified
methodology providing a pathway for certification of
SA&SAA systems. The significance of this approach
is also discussed in the Communication, Navigation
and Surveillance/Air Traffic Management and
Avionics (CNS+A) context, with a focus on the
evolving UAS Traffic Management (UTM)
requirements.
1. Introduction
In order to support unrestricted access of Unmanned
Aircraft Systems (UAS) to all classes of airspace, it is
essential to develop innovative Decision Support
Systems (DSS) based on emerging Communication,
Navigation, Surveillance and Air Traffic
Management and Avionics (CNS+A) technologies.
The introduction of novel CNS systems, for manned
and Unmanned Aircraft System (UAS) operations, is
driven by emerging Air Traffic Management (ATM)
and UAS Traffic Management (UTM) requirements.
Safety, efficiency and airspace capacity requirements
are defined in the ATM modernisation programmes
including Single European Sky ATM Research
(SESAR) and Next Generation Air Transportation
System (NextGen) [1]. On the other hand, UTM
system requirements are defined by various
programmes worldwide to support a wide variety of
J Intell Robot Syst
unmanned aircraft from those equipped with
minimalistic avionics to autonomous UAS [2]. The
UTM system will need persistent CNS coverage to
ensure and monitoring conformance to the
constraints, that can be provided by a combination of
low-altitude radar, cell, satellite, and other means.
Technological challenges exist in the UAS domain, as
effective integration of these aerial robots in non-
segregated airspace relies upon the introduction of a
certifiable Separation Assurance and Sense-and-
Avoid (SA&SAA) capability [3 - 5]. Higher levels of
on board autonomy are also required to mitigate the
risks arising in connection to possible failures to the
Command and Control (C2) loop involving the
ground pilot.
UAS collision avoidance, referred to as SAA can be
defined as the automatic detection of possible
conflicts by the Unmanned Aerial Vehicle (UAV) and
the resolution of any existing collision threats by
accomplishing safe avoidance manoeuvres. The
maturity of SAA techniques and enabling
technologies is considered very limited when viewed
in the perspective of civil airworthiness regulations
for manned aircraft, raising concerns to certification
authorities and airspace users [4].
The aviation research community is working on a
number of SA&SAA algorithms to replicate and even
to exceed the see-and-avoid capability of humans.
Despite a huge number of efforts that have been
devoted to the integration of UAS in non-segregated
airspaces, the standards as well as a certification
framework for SA&SAA is yet to be established.
Some existing approaches include defining dynamic
separation thresholds for small UAS, aggregated
collision cone approach, 3D obstacle avoidance
strategies for UAS planning and re-planning, sensor
resource management, including uncertainties in the
framework and well-clear boundary models. A
unified approach to SA&SAA for manned and
unmanned aircraft has the potential to address the
challenges of UAS integration and also to provide a
pathway to certification. One of the key technology
enablers to achieve this goal is the implementation of
suitable hardware components and data fusion
techniques for cooperative and non-cooperative
SA&SAA tasks. Such functions will provide manned
and unmanned aircraft the capability to consistently
and reliably perform equally or even to exceed the
see-and-avoid performance while allowing a seamless
integration of unmanned aircraft in the ATM network
and UTM system.
2. SA&SAA Requirements and Technologies
The successive steps for integration of UAS into the
commercial airspace and then into the aerodrome
areas are identified in the Aviation System Block
Upgrades (ASBU) by the International Civil Aviation
Organization (ICAO) [5]. Some recommendations
towards addressing operational and certification
issues for civil UAS were provided by the Joint
Aviation Authorities (JAA) CNS/ATM Steering
Group [7]. Additionally, a number of special groups
and committees such as ASTM F38, EUROCAE
WG-73, ICAO UASSG and RTCA SC-228 are
working on requirements, design, performance,
quality acceptance tests and certification of UAS and
its supporting systems including avionics,
communication links and ground control station
elements [7-9]. Current advances in state-of-the-art
avionics technologies (sensors and multi-sensor data
fusion software) have led to a number of innovative
non-cooperative and cooperative SAA solutions.
Such techniques have been predominantly developed
either for non-cooperative or cooperative scenarios. A
number of global and regional programs are
investigating such implementations. In 2009, the
European Defense Agency commenced a project
named Mid Air Collision Avoidance System
(MIDCAS) to develop an experimental SAA system
based on electro-optical, infrared and radar sensors
[10]. Another such development was the
establishment of the SAA Science and Research
Panel (SARP) in the year 2011 by the Office of the
Under Secretary of Defense (OUSD) for acquisition,
technology, and logistics in order to provide solutions
for collision avoidance [11]. Although these projects
are addressing the requirements of SAA systems, a
solid mathematical framework that can be used by the
certification authorities is yet to be developed.
Some requirements for designing and developing an
effective SAA system can be derived from the current
regulations in place for see-and-avoid [10-13]. In case
of see-and-avoid, the main roles and responsibilities
of pilots are stated in FAA AC 90-48C and FAR
91.113 and they are described in terms of regulations
on maintaining vigilance, regardless of whether the
operation is conducted under Instrument Flight Rules
(IFR) or Visual Flight Rules (VFR). One of the
fundamental limitations for certification authorities to
fully certify SAA is to evaluate the ability of the
current and future UAS to be able to replicate the
human see-and-avoid capability, at a comparable or
superior level upon replacing the on-board pilot. This
is applicable both for the Ground Control Station
(GCS) remote pilot and UAV platform when operated
in a fully autonomous mode.
In manned flight, see and avoid, visual sighting,
separation standards, radar, proven technologies and
procedures, and well-defined pilot behaviors combine
to ensure safe operation. ICAO provides
recommendations for defining separation between
aircraft (manned) and is categorised as light, medium
and heavy based on wake vortices generated by the
aircraft. Some differences that affect UAS
interoperability with the ATM system and in defining
separation are [14]:
En Route: Current UAS are not able to meet the
requirements to fly in Reduced Vertical Separation
Minima (RVSM) airspace. UAVs do not fly
traditional trajectory-based flight paths and require
non-traditional handling in emergency situations.
Terminal: Current UAS cannot comply with Air
Traffic Control (ATC) visual separation clearances
and cannot execute published instrument approach
procedures.
Airports: The introduction of UAS at existing
airports represents a complex operational
challenge. For the near term, it is expected that
UAS will require segregation from mainstream air
traffic, possibly accommodated with UAS launch
windows, special airports, or off-airport locations
where UAS can easily launch and recover.
In the US, FAA regulations are highly prohibitive for
UAS operations, expecting non-cooperative sensors
to be on board all UAV platforms, but this approach
is changing [15]. In this perspective, a number of
non-cooperative sensors have been employed for
detecting and tracking other traffic in the surrounding
airspace. A lightweight radar sensor and a
computationally efficient method for determining
collision avoidance manoeuvres proves efficient for
avoiding any identified collision. Vision-based
sensors have been explored for a number of years
considering the criticality of these systems in UAS
applications. These systems are efficient in detecting
real-time collision conflicts at distances that are safe
for performing an avoidance manoeuvre. The key
issues associated with these sensors are the cost, high
computational complexity and the need for high
resolution sensors. Nevertheless, several research
activities have focused on the development of low-
cost vision-based sensors and the development of
efficient algorithms to deal with the computational
cost drawbacks associated with high-resolution
optical sensors. LIDAR has emerged as a promising
technology for obstacle detection and tracking. The
main advantages of this sensor are the higher levels of
accuracy that can be achieved and the narrow FOV
that it offers. Furthermore, these systems are typically
complemented with other non-cooperative
technologies.
The currently available SA&SAA technologies do not
completely meet the targeted levels of safety with the
practical Size, Weight and Power (SWaP) criteria of
UAVs for Line-Of-Sight (LOS) and Beyond Line-Of-
Sight (BLOS) operations. The proposed detection
range and Field of View (FOV) have to be adequate
to ensure separation from intruders to prevent a
probable mid-air collision. In case of cooperative
scenarios, Automatic Dependent Surveillance–
Broadcast (ADS-B) systems, Portable Collision
Avoidance System (PCAS), FLight AlaRM
(FLARM) and different classes of Traffic Collision
Avoidance System (TCAS) are employed for sensing
and tracking intruders. In recent years, a number of
manned and unmanned aircraft are equipped with
ADS-B transponders to locate, identify and
communicate with neighbouring traffic. ADS-B,
although might currently have lower levels of
integrity, plays a crucial role, specifically for SAA, in
order to support collision avoidance as well as
separation maintenance. There have been studies
performed on achieving SAA using a cooperative
approach with ADS-B, superseding traditional
practices using only non-cooperative sensors such as
vision based sensors.
A summary of airborne surveillance systems
currently available is provided in Table 1.
J Intell Robot Syst
Table 1 Airborne surveillance systems.
Airspace
Conventional
Systems
State-of-the-art
Systems
Oceanic
continental
En-route
airspace
with
low/high-
density
traffic
Primary
radar/Secondary
Surveillance
Radar (SSR)
Very High
Frequency
(VHF) voice
position reports
OMEGA/ Long
Range
Navigation
Customized
Navigation
Database
(LORAN-
CNDB)
Automatic
Dependent
Surveillance
(ADS)
Continental
airspace
with high-
density
traffic
Primary radar
SSR Mode A
(ATC
Transponder
Mode
signifying
aircraft call
sign)
SSR Mode C
(ATC
Transponder
Mode
signifying
aircraft call sign
and altitude)
SSR Mode A
SSR Mode C
SSR Mode S
(ATC
Transponder
Mode signifying
call sign, altitude
and additional
aircraft data)
ADS
Terminal
areas with
high-density
Primary radar
SSR Mode A/C
SSR [Mode
A/C/S]
ADS
Typically, the FOV has to be equivalent or superior to
that of a pilot in the cockpit and it corresponds to a
primary FOV of 60˚ vertically/70˚ horizontally and a
secondary FOV of 100˚ vertically and 120˚
horizontally. To satisfy this requirement, typically a
suite of sensors including optical, thermal, LIDAR,
MMW radar, Synthetic Aperture RADAR (SAR) and
acoustic sensors are employed. LIDAR has been used
for detecting, tracking and avoiding obstacles in low-
level flight [16 - 18]. The adoption of a multi-sensory
approach to CA (employing passive and active MMW
radar, Forward Looking Infra-Red (FLIR), LIDAR
and an Electronic Surveillance Module (ESM) for
obstacle detection) has resulted in adequate
performance especially in low- to medium-dynamics
platform applications. Acoustic sensors are
specifically used in small UAVs and they provide
effective intruder detection a 360º FOV, that can be
used for performing quick-reaction avoidance
manoeuvres [19]. More recently, cooperative systems
including TCAS and ADS-B systems are used in
conjunction with a non-cooperative sensor suite.
Since a variety of information is available, effective
multi-sensor data fusion techniques and novel Human
Machine Interface (HMI) designs are required. Such
considerations are addressed by researchers in the US
Air Force Common Airborne Sense and Avoid
(C-ABSAA) program and similar programs
worldwide.
3. Existing Approaches to SA&CA
An avoidance volume is fundamentally a virtual 3D
volume defined in the airspace. The avoidance
volume is generated from real-time navigation
measurements, tracking observables, relative
dynamics between platforms, collision distance, time
of collision and manoeuvrability of the platform. The
inclusion of navigation error of the host aircraft and
the tracking errors of all other traffic in the airspace
provides a practical buffer. It also maximises the
ability to predict collisions and provides the host
aircraft with adequate time to generate a re-optimised
trajectory and execute the required steering
commands. A buffer is also added accounting for the
propagation of errors through the aircraft dynamics.
Different approaches to SA&CA were proposed
recently. Some of the state-of-research techniques
are:
SA&SAA using Variable Surveillance Envelope:
The size and shape of a safety volume was
monitored by onboard sensors and the volume was
modified according to the speed and motion
vectors of the aircraft or other traffic, so as to
maximize efficient use of sensor capabilities and
minimize the size, cost and power requirements of
the system [20]. This method provided a strategy
only for effective sensor utilisation.
UAS Collision Encounter Modelling and
Avoidance Algorithm: Detection of other traffic
was performed by defining a geometric sensor
model and a collision cone approach. Avoidance
algorithms consisted of proportional navigation
guidance that uses information from the collision
potential between the unmanned aircraft and other
traffic [21].
Dynamic Separation Thresholds: In this method,
the distance-based thresholds are replaced with
time-based thresholds that account for intruder
performance using turning flight geometry to
implement a computationally inexpensive
solution [22].
UAS Collision Avoidance Algorithm Minimizing
Impact on Route Surveillance: An aggregated
collision cone approach was introduced to detect
and avoid collision with two or more aircraft
simultaneously (geometric collision avoidance
system) [23].
Europe and US Standards Perspective: The SC-
203 and WG-73 definitions are considered, which
define SAA to encompass two high level
functions: Self Separation (SS) and collision
avoidance. SS is intended to resolve any conflict
early, so that a UAS remains “well clear” of other
aircraft and avoids the need for last-minute
collision avoidance manoeuvres [24].
3D Obstacle Avoidance Strategies for UAS
Mission Planning and Re-Planning: The focus was
on mission planning tasks. A planning algorithm is
described, which allows the vehicle to
autonomously and rapidly calculate 3D routes
[25]. The calculation of routes was based on
obstacle avoidance strategies and on specific
vehicle performance constraints.
Development of a Mobile Information Display
System for UAS Operations: The focus was on
estimating the current aircraft position uncertainty
volume. Overlap of an intruder and host aircraft
uncertainties determines the current risk [26].
Sensor Resource Management to Support UAS
Integration into the National Airspace System:
The aim was to provide resource allocation
strategies and ensures aircraft adhere to the
minimum separation requirement. An evolutionary
algorithm is impleaded and Kalman filter's
covariance matrices are used to determine
positional uncertainty to predict if a separation
requirement is violated [27].
Coordination of Multiple UAS for Tracking under
Uncertainty: Partially Observable Markov
Decision Processes (POMDPs) were used for
controlling fleets of UAS under uncertainty [28].
Self-Separation Support for UAS: Criteria were
identified for defining requirements for a future
SA and CA system Concept of conflict probing
and the associated stability for the available data,
interfaces and display are identified [29].
Analysis of Well-Clear Boundary Models for the
Integration of UAS in the NAS: A definition of
well clear was provided to the SA concept for the
integration of UAS into civil airspace. This
research by NASA presents a family of well-clear
boundary models based on the TCAS II resolution
advisory logic. Analytical techniques were used to
study the properties and relationships satisfied by
the models. Some of these properties are
numerically quantified using statistical methods
[30].
Fig. 1 shows the well-clear threshold. Based on the
current ATM operations, a conflict is defined when an
aircraft encounter happens within 3.5 NM of one
another horizontally and within 2000 ft above an
altitude level of 29,000 ft and 1000 feet below the
29,000 ft level. A self-separation volume is defined
much larger than the collision volume but it may vary
in size with operational area and airspace class [30].
In this case, a conflict is defined to occur when other
traffic enters the self-separation volume. The self-
separation threshold is then defined as a boundary at
which the host aircraft performs a manoeuvre to
prevent other traffic from penetrating the self-
separation volume. Hence, it can be inferred that the
addition of the self-separation volume provides a
performance goal that is analogous to the collision
volume. The encounter geometry is then evaluated in
the relative coordinate frame supporting the study of
relative dynamics between the host aircraft and
intruders in the airspace. All the approaches discussed
above have attempted in defining a methodology for
UAS SA/CA, which are limited in its applicability.
That is, given a scenario or specific onbaord
sensor/system, these methods can provide
information for SA&CA. The unified approach
overcomes the shortcomings of these methods by
allowing quantification of total uncertainty volume in
the airspace surrounding the intruder tracks in real-
time for both cooperative and non-cooperative
scenarios.
J Intell Robot Syst
Well Clear Threshold
Other Traffic
4000’
700’
100’
Host Aircraft
Fig. 1 Well-clear threshold [11, 12].
4. Unified Analytical Framework
The unified analytical framework for SA&CA
supports the quantification of the total uncertainty
volume surrounding other traffic in the airspace in
real-time for both cooperative and non-cooperative
scenarios. The unified analytical framework targets
successful certification by Civil Aviation Safety
Authority (CASA), Federal Aviation Administration
(FAA), Civil Aviation Authority (CAA), European
Aviation Safety Agency (EASA) and other prominent
aviation safety organisations worldwide. The unified
approach allows carrying out such assessment and
supports the case for certification. Both real-time and
off-line determination of the safe-to-fly envelope
based on the installed avionics sensors and on the
own/intruder platform dynamics or, alternatively,
identifying the sensors required for the platform to
safely fly a certain pre-defined envelope (in the
presence of intruders with specified vehicle
dynamics).
The currently available remotely piloted aircraft
standards provided by CASA (Australia) supports a
weight based classification (CASR Part 101) and are
given by [31]:
very small (100g<2kg)
small (2-25kg) (where required with 7kg
restriction)
medium (25-150kg)
large (>150kg)
The unified approach can be applied to all classes of
UAS defined by CASR Part 101 and other regulations
worldwide. Additionally, the proposed approach can
be extended to the operation of single UAS, multi
UAS, mixed fleet and mixed equipage aircraft.
In the unified method to cooperative and non-
cooperative SA&SAA, both navigation error of the
host UAV platform and tracking error of other traffic
are combined in order to obtain an overall avoidance
volume. Therefore, the navigation and tracking errors
are expressed in range and bearing uncertainty
descriptors. In order to estimate navigation and
tracking errors, sensor error modelling is performed.
The variation in the UAV state vector, is expressed
as:
(1)
where p is the position of the UAV and t is the time
of measurement. Let , and be the range,
azimuth and elevation obtained from a SA&CA non-
cooperative sensor/cooperative system. Let ,
and be the nominal range, azimuth and elevation
values. Consider , and as standard deviations
of the error in range, azimuth and elevation
respectively. Hence, the error ellipsoids are given as:
(2)
In order to develop a unified approach to cooperative
and non-cooperative SA&CA, the error ellipsoids are
typically subjected to two transforms: rotation, R and
translation, T that is defined as a projection along the
LOS vector of the UAV. The inverse transformation
applied to one of the two ellipsoids with respect to
another, L is thus expressed as:
(3)
The intruder position vector, translated from host
body frame to Earth Centred Earth Fixed (ECEF)
reference frame with respect to is given by:
(4)
The intruder position vector uncertainty in ECEF
frame () is expressed as:
(5)
where is the error in the position vector of other
traffic in the host body frame and is the rotation
(angular) error matrix. The rotation matrix in terms of
azimuth and elevation angles is given by:
(6)
where c and s represent cosine and since of azimuth
and elevation angles. Therefore the position vector
after rotation is expressed as:
(7)
The angular error matrix is given by:
(8)
and the error in position is expressed as:
(9)
In a static non-cooperative case, the errors in range,
azimuth and elevation are given by:
(10)
(11)
(12)
where , , are the nominal range, azimuth and
elevation measurements. {, } are the
parameterization factors required for reduced
information transfer between air and ground systems.
The transformation of {, , } to {x, y, z} is given
by: (13)
(14)
(15)
A study on correlation between the navigation and
tracking sensor error measurements is essential to
determine the overall uncertainty volume. As a result,
uncorrelated, covariant and contravariant cases are
possible. As an example, considering navigation
measurements from the host UAV and tracking
observables of the intruders, the dependences of
errors in {x, y, z} on the correlation between the
sensor measurements are given by:
) (16)
) (17)
) (18)
where {, } is the position of the intruder
obtained from the airborne surveillance system, {,
} is the position of the host UAV and
{, , } define the
correlation between the system measurements. An
example of the two combined navigation and tracking
error ellipsoids assuming error in range only, and the
resulting uncertainty volumes for uncorrelated and
correlated (covariant and contravariant) sensor error
measurements (3 out of a total of 27 possibilities) are
illustrated in Fig. 2.
All 27 combinations for generating the PUV are
shown in Fig. 3. Data size is a key attribute and larger
the data size, the more is the confidence on
determining the correlation coefficient. A kinematic
analysis is performed to characterise the relative
position and displacement factors. The uncertainty
volume at different time epochs is obtained based on
a confidence region and is governed by the errors in
{x, y, z}, and is given by:
) (19)
) (20)
) (21)
where v is the velocity of the host UAV in {x, y, z}
and t is the time epoch. When an error exists in the
measurement of elevation and azimuth angles, the
resultant cone obtained at the estimated range is
illustrated in Fig. 4a. In a timevarying case, the
variation of PUV due to errors in range and bearing
measurements can be conveniently represented by the
unified approach. An example of the PUV obtained at
the estimated range due to tracking errors is shown in
Fig. 4b. The inflation of the PUV due to an increase
in the navigation and tracking errors is shown in Fig.
4. In order to generate the Collision Avoidance
Volume (CAV), a further inflation is applied to the
PUV by considering the traffic relative dynamics. For
each platform in a certain airspace region, the CAV is
generated at discrete time intervals. An optimal
avoidance trajectory is computed when a possible
collision is predicted (i.e., when the nominal host
aircraft trajectory intercepts a CAV). Thus, in a
multiplatform UTM system implementation, an
autonomous deconfliction strategy can be
implemented by generating dynamic geo-fences for
all conflicting trajectories and CAVs. The uncertainty
volume varies at different time epochs and is
dependent on the relative dynamics between the host
aircraft and the intruder.
J Intell Robot Syst
Fig. 2 Uncertainty volume for range only errors (uncorrelated).
Fig. 3 Range only error uncertainty volumes.
After the states of the intruder are observed (by the
host aircraft) in a specific time interval, the inflation
of the CAV due to relative dynamics is dependent on:
the maximum observed velocity of the intruder in the
observation period; and the maximum projected
velocity of the intruder. This projection can be made
using the aircraft dynamics model. If the aircraft type
and/or its dynamics model are not known, a
conservative approach can be employed to estimate
velocity (i.e., introducing worst case assumptions). In
the most conservative case, the relative dynamics
inflation is given by:
(22)
Where is the relative dynamics inflation in the
x, y, z directions, is the maximum observed
velocity of the intruder (a scalar) in the observation
time interval , is the maximum
projected velocity of the intruder in the projection
time interval (i.e., from the end of the
observation period to the estimated time of conflict).
a.
b.
Fig. 4 a) Uncertainty volume due to error in bearing
measurements b) Inflation of the PUV
In addition to the uncertainty given by navigation,
tracking and relative dynamics, further consideration
should be given to the current separation
requirements applicable to particular airspace regions.
The subsequent step involves the selection of the
optimal trajectory from the generated set of safe
trajectories, which is then provided in the form of
steering commands to the automatic flight control
system. The implemented decision logics are based
on minimisation of the following cost function:
(23)
where, given TT as the time-to-threat and TM as the
avoidance manoeuvre time, is the time at which
the safe avoidance condition is successfully attained,
defined as: (24)
and
is specific fuel consumption, is
thrust profile and the coefficients are the
weights attributed to time, fuel and distance
respectively. The term dm (t) is given by:
(25)
and corresponds to the minimum distance from the
uncertainty volume (ground and aerial), where ,
and are the coordinates of the bounding
surfaces of the volume.
5. Detection Sensor/System Models
The detection equipment in a SAA system involves a
combination of non-cooperative sensors, including
active/passive Forward-Looking Sensors (FLS) and
acoustic sensors, as well as cooperative systems,
including Automatic Dependent Surveillance
Broadcast (ADS-B) and Traffic Collision Avoidance
System (TCAS). Global Navigation Satellite Systems
(GNSS), Inertial Measurement Unit (IMU) and
Vision Based Navigation (VBN) sensor
measurements (and possible augmentation from
ADM) are used for navigation computations. A
conceptual representation is shown in Fig. 5.
Navigation
Sensors
Non-cooperative Sensors and
Cooperative Systems
Intruder Tracks
Overall Avoidance Volume
Safe Avoidance Trajectory
Integrity Flags
SA&SAA
Algorithms
GNSS IMU Passive/
Active FLS
ADS-B TCAS
Acoustic
Vision-
based
Fig. 5. Navigation and surveillance sensors/systems.
J Intell Robot Syst
State-of-the-art SA&SAA technologies are listed in
Table 2 representing C for cooperative and NC for
non-cooperative (both active and passive) sensors.
Table 2 SA&SAA technologies.
Sensor/
System
Type
Information
Trajectory
Visual
camera
NC,
Passive
Azimuth,
Elevation
Extracted
Thermal
camera
NC,
Passive
Azimuth,
Elevation
Extracted
LIDAR
NC,
Active
Range
Extracted
Millimeter
wave radar
(MMW
radar)
NC,
Active
Range,
Bearing
Extracted
Synthetic
Aperture
Radar
(SAR)
NC,
Active
Range,
Bearing
Extracted
Acoustic
NC,
Active
Azimuth,
Elevation
Extracted
Automatic
Dependent
Surveillance
(ADS)
C
Position,
Altitude,
Velocity
and
Identity
State
Vectors
Provided
Traffic
Collision
Avoidance
System
(TCAS)
I/II/IV/
Airborne
Collision
Avoidance
System
(ACAS)
I/II/III/X
C
Range,
Altitude
Extracted
A low-cost navigation and guidance system is
adopted for position estimates, which includes Global
Navigation Satellite System (GNSS), Micro-
Electromechanical System (MEMS) Inertial
Measurement Unit (IMU) and Vision Based
Navigation (VBN) sensors [32]. When the set
threshold is exceeded and the detection is continuous,
high level tracking detection is performed by using a
Kalman Filter. The predicted state, at time is
given by:
(26)
where is the position in the and directions
respectively as a function of time, . is the
velocity in the and direction respectively,
is the acceleration and is the prediction Gaussian
noise.
The Kalman filter equations are given by:
(27)
(28)
where: (29)
(30)
where represents the design matrix and is
the measurement noise covariance matrix and is the
sample time. Sensor fusion in case of cooperative
systems can be achieved by employing a variety of
algorithms including Track-To-Track (T3) and others.
The primary advantage of adopting T3 method is to
combine the estimates instead of combining the
observations from different sensors [33]. The track
fusion algorithm is defined as the weighted average
variance of all the tracks and is given by:
(31)
(32)
Once the tracks are fused and the states are estimated,
the imminent trajectory is predicted. The errors in
predicted trajectory can be derived from the quality of
the measurements, reflected in the prediction error,
which are expressed as:
(33)
where is the exhibited (modelled) trajectory
and is the predicted optimal trajectory at
sample time . For trajectory prediction, the
obstacle centre of mass, the target orientation and the
geometric shape of the uncertainty volume are
determined. Once the trajectory is predicted, the Risk
of Collision (ROC) is determined by calculating the
probability of a near mid-air event for the predicted
trajectory over the time horizon by employing Monte
Carlo approximations.
For trajectory prediction, the obstacle centre of mass,
the target orientation and the geometric shape of the
uncertainty volume are determined. Once the
trajectory is predicted, the Risk of Collision (ROC) is
determined by calculating the probability of a Near
Mid-Air Collision (NMAC) event for the predicted
trajectory over the time horizon by employing Monte
Carlo approximations given by:
(34)
where is the number of samples, is defined as
a future horizon up to where it is desired to predict
the trajectory, is the minimum distance required
to avoid the obstacle, is the time horizon defined
for collision and the subscript is added to
distinguish it from the number of runs in a Monte
Carlo simulation. The accuracy of the approximation
is entirely based on the number of samples. ADS-B
system is used to obtain the state of the intruders. The
future position of the intruders is projected based on
the estimate of the current state vector and the flight
profile. The ADS-B measurement model adopted for
intruder position and velocity estimates in and
cardinal directions is given as:
(35)
Assuming that the velocity components, , ,
and are affected only by Gaussian noise
with zero mean, the standard deviation is defined by
the covariance matrix given by:
(36)
where represents the mean. Data fusion
algorithms including Interacting Multiple Model
(IMM) can be adopted specifically for the cooperative
case. The IMM model is a state-of-the-art tracking
algorithm suitable when multiple kinematic behaviour
are to be considered [10]. Using this model, the state
vector of the intruders is determined and this is
propagated to predict the future trajectories using a
probabilistic model. After computing the mixing
probability, the combination of the state estimate is
given by:
(37)
where is the mode probability update. For
conflict detection, the resultant covariance matrix,
after transformation is defined as:
(38)
where S is the diagonal covariance matrix and R
represents the transformation matrix between the
heading aligned frame to that of the RPAS host
platform frame. The probability of conflict is defined
as the volume below the surface of the probability
density function, representing the conflict
zone. The conflict probability, is expressed as:
(39)
where represents the conflict separation
distance and correspond to the rows of the
conflict boundary matrix. The conflict probability is
simplified as:
- (40)
6. UAS Avoidance Trajectory Optimisation
In the context of conflict identification and resolution
perspective, trajectory optimisation is characterised
by the identification of the most suitable 3D/4D
avoidance trajectory from the time of detection to a
point where the avoidance trajectory re-joins with the
nominal one. In this optimisation problem,
dynamics/airspace constraints, user preferences,
intruder trajectory, as well as meteorological and
traffic conditions are considered. Hence, the adoption
of computational algorithms required for trajectory
optimisation in achieving SAA represents a
substantial evolution from the conventional safe-
steering methodologies adopted in current CNS/ATM
systems. Current research efforts are addressing
practical implementations of advanced multi-model
and multi-objective 3D/4D trajectory optimisation
algorithms in novel ground-based and airborne
J Intell Robot Syst
CNS+A systems [34-36]. Both direct and indirect
methods are employed for computing an efficient
avoidance trajectory. Most computationally efficient
trajectory optimisation algorithms adopted for UAS
applications belong to the family of direct methods.
Safety-critical applications of trajectory optimisation
algorithms are actively investigated for airborne
emergency Decision Support Systems (DSS), also
known as safety-nets. These safety-critical CNS+A
applications impose real-time requirements on the
trajectory generation algorithm. Additionally, all
generated trajectories must necessarily fulfil each and
every set constraint, as the obstacle avoidance and the
manoeuvring envelope are formulated as constraints.
As a result, these requirements limit considerably the
choice of solution methods and multi-objective
optimality decision logics [37]. Direct shooting
methods involving the transcription into finite-
dimensional NLP problem can be either performed by
introducing a control parameterisation based on
arbitrarily chosen analytical functions, as in
transcription methods, or by adopting a generalised
piecewise approximation of both control and state
variables based on a polynomial sequence of arbitrary
degree, as in collocation methods. In both cases the
transcribed dynamical system is integrated along the
time interval between an initial and final time.
The search of the optimal set of discretisation
parameters is formulated as a NLP problem, which is
solved computationally by exploiting efficient
numerical NLP algorithms. In the direct shooting and
multiple direct shooting, the parameterisation is
performed on the controls only. The dynamic
constraints are integrated with traditional numerical
methods including Runge-Kutta appraoch, and the
Lagrange term is approximated by a quadrature
approximation. In case of multiple shooting, the
analysed time interval is partitioned into
subintervals, and the direct shooting method is
applied to each divided subinterval. Parallel
implementations of direct shooting methods involve
the simultaneous integration of a family of
trajectories. The solution is based on different control
parametrisation profiles and takes advantage of
increasingly common multi-thread/multi-core
hardware architectures. The optimal solution is
determined a posteriori, both in the case of single
objective and multi objective implementations. In the
unified analytical framework, the following set of
Differential Algebraic Equations (DAE) introducing a
variable mass 3-DoF model was employed and are
given as:
(41)
where the UAV state vector consists of the following
variables: v is longitudinal velocity (scalar) [m s-1]; γ
is flight path angle (scalar) [rad]; χ is track angle
(scalar) [rad]; is geographic latitude [rad]; λ is
geographic longitude [rad]; z is flight altitude [m];
is thrust angle of attack [rad] and m is aircraft mass
[kg]; and the variables forming the control vector are:
T is thrust force [N]; N is load factor [ ] and μ is bank
angle [rad]. Other variables and parameters include:
D is aerodynamic drag [N]; is wind velocity, in its
three scalar components [m s-1]; g is the gravitational
acceleration [m s-2]; is radius of the Earth [m] and
FF is fuel flow [kg s-1]. Adopting a multi-phase
trajectory optimisation formulation, the selection of
the optimal avoidance trajectory in the safe steering
phase is typically based on minimising a cost function
of the following form [35]:
(42)
where is the slant distance of the host platform
along the avoidance trajectory from the avoidance
volume associated with other traffic and
is the time at which the safe
avoidance condition is successfully attained, where
TTT is the time-to-threat and AMT is the avoidance
manoeuvre time. is the host platform’s mass and
are the positive weightings
attributed to time, distance, integral distance and fuel
respectively. In time-critical avoidance applications
(i.e., closing-up obstacles with high relative
velocities), appropriate higher weightings are used for
the time and distance cost elements. SA also has to be
achieved in the vicinity of airports and in the
Terminal Maneuvering Area (TMA). In this case a
runway capacity model is considered taking into
account the time of separation between host UAV
platform and other traffic and is given by:
When :
(43)
When :
(44)
where is the time of separation, and are the
velocities of adjacent aircraft, is the distance of
separation, is the required separation and
represents the order of the separation required.
5. Avionics System Implementation
An efficient Flight Management System (FMS) that
enhances safety should address effectively both
collision avoidance and separation maintenance tasks.
In order to implement a common FMS functionality
for manned and unmanned aircraft, key CA tasks
including Tracking, Decision-making and Avoidance
(TDA) must be addressed considering the sequential
steps depicted in Fig. 6 (adapted from [38]).
Data Fusion
Management
Control
Analysis
Intruder
Tracking and
Trajectory
Determination
(Track)
Criticality
Analysis
(Evaluate
Prioritise)
Self
Separation
(Declare
Determine)
Collision
Avoidance
(Declare
Determine)
Sensor and Information Management
SA&CA
Technologies
(Detect)
SA&CA
Management
(Avoid)
Platform
Dynamics
(Execute)
UAS Pilot
(Command
Supervise)
Flight Management System
(Command and Avoid)
Fig. 6 SA&CA system process.
The state vector of the tracked obstacles are obtained
by employing multi-sensor data fusion algorithms
including Extended Kalman Filter (EKF), Unscented
Kalman Filter (UKF), Particle Filter (PF) and other
knowledge-based techniques including learning based
mechanisms in order to predict the intruder trajectory
in a given time horizon. Currently, on-board
trajectory re-planning with dynamically updated
constraints based on intruder and the host UAV
platform dynamics is used to generate obstacle
avoidance trajectories [39].
After obtaining the trajectory information, criticality
analysis is performed to prioritize (i.e. to determine if
a collision risk threshold is exceeded for all tracked
intruders) and to determine the steering commands
required for executing an avoidance action. If
possibility of a collision exists, the SAA algorithms
generate and optimise an avoidance trajectory
according to a cost function that is based on minimum
distance, fuel, time and closure rate criteria with the
J Intell Robot Syst
aid of differential geometry or pseudo-spectral
optimisation techniques to generate a smooth and
flyable trajectory. The Airborne Separation
Assurance Function (ASAF) is implemented as a
FMS component (Fig. 7).
Host
Aircraft
Information
Trajectory
Planning,
Optimisation
and Guidance
Conflict
Detection
Cooperative
System R’r
Conflict
Resolution
Track Generation
& Maintenance
Cooperative
System T’r
CDTI
Cautions
and
Warnings
Received
Traffic
NG-FMS
Detected
Traffic
Non-
Cooperative
System
Separation
Maintenace
CNS/ATM
Network
Fig. 7 SA&CA functions.
The ASAF gradually will transfer from the current
ATCO controlled modes to distributed modes. The
distributed modes provide robustness in terms of
reliability. The FMS thus in addition to providing
enhanced path planning, navigation, guidance and
aircraft performance, provides automated separation
assurance capabilities. The Cockpit Display of Traffic
Information (CDTI) is used for displaying traffic
information. In the CNS+A context, the ASAF
equipage required is summarised in Table 3.
The design and implementation of Next Generation
Flight Management Systems (NG-FMS) algorithms is
aimed at satisfying the CNS performances namely
Required Communication Performance (RCP,
Required Navigation Performance (RNP) and
Required Surveillance Performance (RSP).
Additionally, higher levels of automation are required
to support the dynamic adaptation of decision logics
required to enable single pilot and UAS operations.
The NG-FMS software is based on multi-objective
and multi-model 4DT optimisation algorithms for
strategic, tactical and emergency scenarios. The
NG-FMS architecture is shown in Fig. 8.
Table 3 Equipage and tasks summary.
Function
Equipment/Task
Communication
Telecommunications Datalinks,
Controller Pilot Data Link Control
(CPDLC), Voice Communications
Navigation
Navigation sensors including
GNSS, INS, etc.
providing 3D/4D navigation
Surveillance
Cooperative Systems
(TCAS, ACAS, etc.)
Non-cooperative Sensors (FLS)
Situational
Awareness
CDTI Display
Autonomous
Decision Making
Strategic, Tactical and Emergency
Flight Planning
Intelligent Conflict Detection,
Resolution and Avoidance
Weather/ Terrain/ Contrails
Avoidance
The following software modules are introduced in the
NG-FMS:
4D trajectory planning and optimisation – to
perform 4D trajectory planning and optimisation
functions for strategic (offline and online), tactical
(offline and online) and emergency tasks. A
number of performance criteria and cost functions
are used for optimisation including minimisation
of fuel consumption, flight time, operative cost,
noise impact, emissions and contrails.
4D trajectory monitoring and correction – to
perform state estimation, to calculate of deviations
between the 4D trajectory intents and the
estimated/predicted aircraft states and to provide
steering commands to the automatic flight control
system.
Automated Separation Assurance and Collision
Avoidance (SA/CA) – to support cooperative and
non-cooperative separation maintenance as well as
collision avoidance tasks.
4D trajectory negotiation and validation – to carry
out the process of negotiation that can be initiated
by the pilot via the NG-FMS, making use of the
information available on board, or by the air traffic
controller via the 4-PNV system.
Performance manager – to monitor active 4D
trajectory intents for errors and to address RCP,
RNP and RSP requirements in all flight phases.
Integrity manager – to generate integrity C/N/S
caution (predictive) and warning (reactive) flags.
Inputs from a number of sensors/systems and
predefined decision logics are used to provide
annunciations, which are then used to perform
preventive/corrective actions. A typical example is
in which the main causes of GNSS signal outage
and degradation in flight including antenna
obscuration, multipath, fading due to adverse
geometry and Doppler shift are modelled in an
Avionics-Based Integrity Augmentation (ABIA)
system. This increases the levels of integrity and
accuracy (as well as continuity in multi-sensor
data fusion architectures) of GNSS in a variety of
mission- and safety-critical applications [40].
NG-FMS with SA&CA
Surveillance System
Vehicle Data
Management
System
Navigation
Subsystem
Integrated
Vehicle Health
Management
System
Separation
Maintenance
Vehicle
Dynamics &
Performance
Subsystem
Integrity
Management
Subsystem
4D Trajectory
Negotiation &
Validation
4D Trajectory
Planning and
Optimisation
Subsystem
Guidance
Subsystem
LOS and BLOS
Communication System
Flight Control
Unit
Collision
Avoidance
Subsystem
Ground
Control
Station NG-ATM
System
Navigation
Sensors
Cooperative
Systems/Non-
cooperative
Sensors
Autopilot
Flight Control
System
Fig. 8 NG-FMS architecture.
7. Simulation Case Studies
A typical case is that of multiple traffic performing
cooperative and/or non-cooperative surveillance and
communicating with the ground ATM systems [41].
The uncertainty volume is generated in real-time after
evaluating the risk of collision at the collision point
(Fig. 9). An avoidance trajectory is generated (based
on the platform dynamics) to maintain the required
separation maintenance (more than 500 m) and also
to prevent any mid-air collisions at all of the
predicted time epochs. Close-encounters are typically
evaluated as part of an intermediate step for pruning
the full set of potential conflicts. Such 4D close-
encounters are assumed to occur when the relative
distance (i.e., the norm of the 3D relative position
J Intell Robot Syst
vector) between the nominal positions of a pair of
traffic at a certain time is below a specified threshold
[41]. For all identified close-encounters, the
uncertainty volume associated with host and intruder
platforms are determined. After identification and
review of the state-of-the-art SAA technologies,
Adaptive Boolean Decision Logics (ABDL) are
employed, allowing a dynamic reconfiguration of the
SA&SAA architecture, based on the current error
estimates of navigation and tracking sensors/systems.
The system reference architecture is illustrated in Fig.
9. The architecture allows automated selection of
sensors/systems including passive and active Forward
Looking Sensors (FLS), TCAS I/II/II or ACAS
I/II/III/X and ADS-B system is performed to support
trusted autonomous operations during all flight
phases. A SAA approach that is dynamically
reconfigurable based on the current performances of
sensors/systems and satisfying the total required
system performance will support the CNS+A
framework. A typical example of selection is to
prioritise TCAS data over ADS-B information, given
the higher levels of integrity provided by the TCAS.
The ground surveillance network information
consisting of ATM radar tracks and Air Traffic
Controller (ATC0) instructions in digital format is
uplinked to the UAV platform and shared with the
ground control station remote pilot. Implementations
involving Boolean logics are generally hard wired
and cannot be reconfigured and this limits the scope
of cooperative and non-cooperative unified
framework in terms of automatic decision-making
capability. Therefore adaptive Boolean decision
logics, which are based on real-time monitoring of the
SA&SAA sensors/systems performances, are
presented in the CNS+A context.
A hierarchy for selecting sensors/systems is defined
based on their current error estimates. Such
implementations are feasible by employing field
programmable gate arrays that can provide effective
selecting and sorting mechanisms realised by an array
of dedicated programmable logic blocks.
Fault Tree Analysis (FTA) and Failure Modes Effects
and Criticality Analysis (FMECA) are required to
identify the probability of failure associated with the
sensor and system suite. As a result, the
sensor/system, which provides the best state estimate
of other traffic, is automatically selected. The
presented approach thus provides trusted autonomy
and robustness in all flight phases. The method lays
foundations for the development of an airworthy
SA&SAA capability and a pathway for
manned/unmanned aircraft coexistence in all classes
of airspace. Navigation and tracking performances
are also be improved, thanks to the robustness
introduced into the system. The sensor/system, which
provides the best estimate, is selected automatically.
The presented approach thus provides autonomy and
robustness in all flight phases, and supports all-
weather and all-time operations. Instead of
implementing hardwired decision logics (given by a
pre-defined set of instructions), a dynamic
reconfiguration of decision logics based on CNS
systems integrity augmentation is adopted [39].
Sensors/systems providing the most reliable SAA
solution are automatically selected, providing
robustness in all flight phases and supporting all-
weather operations. This is achieved by performing a
top-down and then bottom-up approaches for
determining the probability of failure associated with
each sensor/ system and ATM links. The probability
of failures dictates the selection of sensor/system for
strategic, tactical and emergency scenarios.
The decision logics are based on identifying the
primary cooperative system and non-cooperative
sensor. In the architecture defined above, the primary
cooperative system is ADS-B and the key non-
cooperative counterparts are vision-based sensors and
LIDAR/MMW radar. An example case would be the
prioritisation of ADS-B measurements when both
ADS-B system and TCAS are present on-board the
UAV.
Considering ADS-B measurements as A, TCAS data
as B and the output for cooperative SAA function as
O1, the expression for the Boolean logics
implementation is derived as follows:
O1 = A + (A T) (45)
O1 = A + (A T’+ T A’) (46)
O1 = A (1 + T’) + T A’ (47)
O1 = A + T A’ (48)
Optional Non-
cooperative
sensors
Cooperative
systems
Non-cooperative
sensors
Acoustic
MMW
RADAR
TCAS I/II/III
ACAS I/II/
III/ X
ADS-B
SAR
Ground
Surveillance
Network
LIDAR Visual
Camera
ATM RADAR
Tracks
ATCO
Instructions
in Digital
Format
Tracked Intruder/Obstacle
State Information
Fig. 9 SAA sensors/systems test-bed for simulation.
Similarly, a Boolean logics decision tree is also
employed for the non-cooperative case. Denoting
data from visual camera as V, thermal camera as I,
LIDAR as L, MMW radar as M, acoustic as A and
the output for non-cooperative SAA function as O2,
the expression for the Boolean logics
implementation is derived as follows:
O2 = A · M · L · (V + (V I)) (49)
O2 = A · M · L · (V + (V I’+ I V’)) (50)
O2 = A · M · L · (V + I V’) (51)
A combined decision tree is adopted for
accommodating both non-cooperative sensors and
cooperative system information. This is described
by the overall output O as:
O = (O1 O2) + O2 (52)
Due to bandwidth limitations existing in current
communication systems, a compact and versatile
parameterisation of the uncertainty volume is highly
desirable to extrapolate its actual shape and size at
close encounter points with minimal data link and
computational burden. In the second case, both host
unmanned aircraft and intruder (Airbus A320
aircraft) platforms are assumed to have on board
ADS-B systems. The host unmanned platform
computes an avoidance trajectory as per the rules of
flight while the other traffic performs a step
descend phase to avoid the mid-air collision. The
cost function use to obtain these avoidance
trajectories are the same as defined in equation 22.
Simulation case studies were performed using the
AEROSONDE UAV as the host platform. In all the
cases, the host UAV was presumed to be equipped
with both non-cooperative sensors (vision-based
camera and LIDAR) and cooperative system
(ADS-B). Other traffic including manned and
J Intell Robot Syst
unmanned aircraft platforms (AEROSONDE UAV
and Airbus A320 aircraft) were considered in the
simulation case studies. In the first case, it is
assumed that no cooperative systems are on board
the intruder (AEROSONDE UAV). In this scenario,
potential conflicts are defined as close encounters in
the 4D space-time domain. The simulations were
performed on a Windows 7 Professional
workstation (64-bit OS) supported by an Intel Core
i7-4510 CPU with clock speed 2.6 GHz and 8.0 GB
RAM. The execution time for uncertainty volume
determination and avoidance trajectory optimisation
algorithms was in the order of 8 sec. Such an
implementation makes it possible to perform real-
time separation maintenance tasks as well as
avoidance of any identified collisions (emergency
scenarios). In order to fulfil safety requirements for
SA&SAA system certification, performance
monitoring and augmentation algorithms (including
integrity) have to be implemented encompassing the
entire CNS sensors/systems chains and the
associated navigation and tracking loops. In the
CNS+A context, this means that either a specified
level of performance is available (with a specified
maximum probability of failure) or, if not, a usable
integrity flag is generated within a specified
maximum Time-To-Alert (TTA). Using suitable
data link and signal processing technologies on the
ground, a certified SA&SAA capability can thus
become a core element of future network-centric
ATM operations.
The algorithms thus support the generation of
appropriate dynamic geo-fences, whose
characteristics are dictated by the obstacle
classification and intruder dynamics, to allow
computation of the optimal avoidance flight
trajectories (Fig. 10). In this approach, separation
(both vertical and lateral) between aircraft is
dictated by the computation of avoidance volumes,
resulting as a combination of navigation error of the
host aircraft and tracking error of the detected
intruders.
Fig. 10 Avoidance of a ground obstacle (non-cooperative case).
Fig. 11 Simulation sceanario.
Fig. 12 Horizontal resolution – intruder 1.
Fig. 13 Horizontal resolution – intruder 2.
Fig. 14 Horizontal resolution – intruder 3.
In case of moving targets, a suite of non-
cooperative sensors and cooperative systems can be
employed to detect and predict the intruders’
trajectories respectively. The Risk of Collision
(RoC) is evaluated and based on this assessment, if
there is a possibility of a collision, an uncertainty
volume is computed according to the models
described in Section 4. A simulation case study for
UAS SA&SAA in a multi-platform coordination
scenario was also performed. In this scenario, it is
assumed that the host aircraft is equipped with NG-
FMS and both cooperative (ADS-B) and non-
cooperative (FLS) means of intruder detection are
present. Intruder 1 is equipped with neither a
cooperative system nor a non-cooperative sensor.
Intruders 2 and 3 are equipped with ADS-B.
Intruder 4 employs a NG-FMS with SA&SAA
algorithms. The simulation scenario is shown in
Fig. 12.
The uncertainties in navigation and tracking
associated to each platform (as seen by all other
conflicting platforms) are combined to generate
avoidance volumes surrounding each traffic. The
avoidance volumes are computed at discrete time
intervals as a function of traffic relative dynamics.
Figures 12, 13 and 14 show the horizontal
resolution obtained between the host platform and
the first three intruders. After obtaining all the
avoidance volumes in the given airspace sectors, the
largest of the four avoidance volumes can be
selected as a reference to perform dynamic capacity
demand balancing (intruder 4 in this case). With
J Intell Robot Syst
respect to intruder 4, since the generated avoidance
trajectory deviates from the planned trajectory, a re-
join trajectory command was performed to ensure
that the generated intent leads back to the original
trajectory.
The avoidance trajectories are generated with
respect to the cost function defined earlier in
equation (22). After the overall avoidance volume is
computed, the avoidance trajectory and a
subsequent re-join trajectory are generated by the
NG-FMS (Fig. 15).
Fig. 15 Re-optimised host platform trajectory.
8. Conclusions and Future Work
A unified analytical framework for Separation
Assurance and Collision Avoidance (SA&CA) was
presented, which can support an evolution of the
current certification framework, hence allowing the
safe operation of unmanned platforms in non-
segregated airspace. The SA&CA unified
framework is based on mathematical algorithms
that quantify in realtime the total volume of
airspace surrounding the intruder track that has to
be avoided. In particular, all navigation (host
aircraft) and tracking (intruders) errors affecting the
state measurements are transformed into unified
range and bearing uncertainty descriptors
facilitating a Communication, Navigation,
Surveillance/Air Traffic Management and Avionics
(CNS+A) system implementation. State-ofthe-art
non-cooperative and cooperative technologies for
Sense-and-Avoid (SAA) were identified and a
reference system architecture was presented.
Furthermore, the advantages offered by the unified
approach when compared with other state-of-the-
art/state of research methods have been included.
The algorithms employed to achieve effective self-
separation and collision avoidance functionalities
were described. Simulation case studies were
presented to corroborate the effectiveness of the
proposed approach. This method provides a
pathway to certification of next generation collision
avoidance (and self-separation) systems for both
manned and unmanned aircraft. Future work will
focus on experimental activities to verify the correct
implementation of the SA&CA functionalities
including ground functional tests and flight test
activities. Further evolutions will address the
inclusion of wake turbulence and meteorological
factors in the unified approach allowing the
application of the same framework to strategic and
tactical ATM and airspace management tasks.
References
[1] R. Sabatini, A. Gardi, S. Ramasamy, T. Kistan and M.
Marino, “Modern Avionics and ATM Systems for Green
Operations”, Encyclopedia of Aerospace Engineering, eds.
R. Blockley and W. Shyy, John Wiley: Chichester, 2015.
DOI: 10.1002/9780470686652.eae1064
[2] P.H. Kopardekar, “Unmanned Aerial System (UAS) Traffic
Management (UTM): Enabling Low-Altitude Airspace and
UAS Operations,” NASA Technical Report, 2014.
[3] S.P. Cook A.R. Lacher, D.R. Maroney, and A.D. Zeitlin,
“UAS Sense and Avoid Development - The Challenges of
Technology, Standards, and Certification”, 50th AIAA
Aerospace Sciences Meeting Including the New Horizons
Forum and Aerospace Exposition, Nashville, TN, USA,
2012. DOI: abs/10.2514/6.2012-959
[4] K. Dalamagkidis, K.P. Valavanis, and L.A. Piegl, “On
unmanned Aircraft Systems Issues, Challenges and
Operational Restrictions Preventing Integration into the
National Airspace System”, Progress in Aerospace
Sciences, vol. 44, issue 7, pp. 503-519, 2008. DOI:
10.1016/j.paerosci.2008.08.001
[5] The International Civil Aviation Organization (ICAO),
Working Document for the Aviation System Block
Upgrades - The Framework for Global Harmonization,
Montreal, Canada, 2013.
[6] G. Amato, EUROCAE WG-73 on Unmanned Aircraft
Systems, ed: EUROCAE.
[7] JAA, UAV TASK-FORCE, Final Report - A Concept for
European Regulations for Civil Unmanned Aerial Vehicles
(UAVs), 2004.
[8] A. Oztekinmet and R. Wever, “Development of a
Regulatory Safety Baseline for UAS Sense and Avoid”,
Handbook of Unmanned Aerial Vehicles, pp. 1817-1839,
Springer Verilag, Netherlands, 2015.
[9] X. Yu and Y. Zhang, “Sense and Avoid Technologies with
Applications to Unmanned Aircraft Systems: Review and
Prospects”, Progress in Aerospace Sciences, vol. 74, pp.
152-166, 2015.
[10] J. Pellebergs and SAAB Aeronautics, The MIDCAS
Project, SAAB Aeronautics, 2012.
[11] K.P. Valavanis and J.V. George, “UAV Sense, Detect and
Avoid: Introduction”, Handbook of Unmanned Aerial
Vehicles, Springer Netherlands, pp. 1813-1816, 2014.
[12] Federal Aviation Administration, Pilot's Role in Collision
Avoidance, AC90-48C, Washington DC, USA, 1983.
[13] DOT/FAA/CT-96-1, Human Factors Design Guide for
Acquisition of Commercial-Off-the-Shelf subsystems,
Non-Developmental Items, and Developmental Systems-
Final Report and Guide, January 1996.
[14] FAA, Integration of Civil Unmanned Aircraft Systems
(UAS) in the National Airspace System (NAS) Roadmap,
US Depatment of Transportation, Federal Aviation
Administration, 2013.
[15] M. Strohmeier, M. Schäfer, V. Lenders, and I. Martinovic,
“Realities and Challenges of NextGen Air Traffic
Management: The Case of ADS-B”, IEEE
Communications Magazine, vol. 52, issue 5, pp. 111-118,
2014. DOI: 10.1109/MCOM.2014.6815901
[16] R. Melnyk, et al. “Sense and Avoid Requirements for
Unmanned Aircraft Systems Using a Target Level of
Safety Approach”, Risk Analysis, vol. 34, issue 10, pp.
1894-1906, 2014.
[17] L. Mejias, J. Lai, and T. Bruggemann, “Sensors for
Missions”, Handbook of Unmanned Aerial Vehicles, pp.
385-399, 2015.
[18] A. Moses, M.J. Rutherford, and K.P. Valavanis, “Scalable
RADAR-Based Sense-and-Avoid System for Unmanned
Aircraft”, Handbook of Unmanned Aerial Vehicles, pp.
1895-1953, Springer, 2014.
[19] A. Finn and S. Franklin, “Acoustic Sense & Avoid for
UAV's”, Intelligent Sensors, Sensor Networks and
Information Processing (ISSNIP), IEEE Seventh
International Conference, 2011.
[20] D. Wood, "Collision Avoidance System and method
Utilizing Variable Surveillance Envelope", U.S. Patent
6,804,607, 2004.
[21] A. Smith, D.M. Coulter and C.S. Jones, “UAS Collision
Encounter Modeling and Avoidance Algorithm
Development for Simulating Collision Avoidance”, AIAA
Modeling and Simulation Technologies Conference, 2008.
[22] M. Mullins, M. Holman, K. Foerster, N. Kaabouch and W.
Semke, “Dynamic Separation Thresholds for a Small
Airborne Sense And Avoid System”, AIAA Infotech@
Aerospace Conference, Boston, MA, USA, pp. 19-22,
2013.
[23] A.L. Smith and F.G. Harmon, “UAS Collision Avoidance
Algorithm Minimizing Impact on Route Surveillance,
AIAA Guidance, Navigation, and Control Conferenc, pp.
1-20, 2009.
[24] E.A. Euteneuer and G. Papageorgiou, “UAS Insertion into
Commercial Airspace: Europe and US Standards
Perspective”, 30th AIAA/IEEE Digital Avionics Systems
Conference (DASC), pp. 5C5-1, 2011.
[25] F. De Crescenzio, G. Miranda, F. Persiani and T.
Bombardi, “3D Obstacle Avoidance Strategies for UAS
(Uninhabited Aerial System) Mission Planning and
Replanning”, Proceedings of the 26th Congress of the
International Council of the Aeronautical Sciences, 2008.
[26] R. Marsh, K. Ogaard, M. Kary, J. Nordlie and C. Theisen,
“Development of a mobile information display system for
UAS operations in North Dakota”, International Journal of
Computer Information Systems and Industrial Management
Applications, Vol. 3, pp.435-443, 2011.
[27] N. Hanlon, K. Cohen and E. Kivelevitch, “Sensor Resource
Management to Support UAS Integration into the National
Airspace System”, AIAA SciTech 2015 Conference, 2015.
[28] J. Capitán, L. Merino and A. Ollero, “Coordination of
Multiple UAS for Tracking under Uncertainty”, In
Proceedings of the 1st Workshop on Research, Education
and Development on Unmanned Aerial Systems, RED-
UAS, 2011.
[29] J. Tadema, E. Theunissen and K.M Kirk, “Self Separation
Support for UAS”, American Institute of Aeronautics and
Astronautics (AIAA), 2010.
[30] J. Upchurch, C. Munoz, A. Narkawicz, J. Chamberlain,
"Analysis of Well-Clear Boundary Models for the
Integration of UAS in the NAS”, Langley Research Center,
Hampton, Virginia, TM–2014–218280, NASA, 2014.
[31] CASA, “CASR Part 101 - Unmanned aircraft and rockets”,
CASA Regulations, Civil Aviation Safety Authority,
Canberra, Australia, 2016.
[32] F. Cappello, S. Ramasamy and R. Sabatini, “A Low-Cost
and High Performance Navigation System for Small RPAS
Applications”, Aerospace Science and Technology, vol. 58,
pp. 529–545, 2016. DOI: 10.1016/j.ast.2016.09.002
[33] L. Rodriguez, R. Sabatini, S. Ramasamy and A. Gardi, “A
Novel System for Non-Cooperative UAV Sense-And-
Avoid”, European Navigation Conference (ENC 2013),
Vienna, Austria, 2013.
[34] J. T. Betts, Survey of Numerical Methods for Trajectory
Optimization, Journal of Guidance, Control and Dynamics,
vol. 21, pp. 193-207, 1998.
[35] C.K. Lai, M. Lone, P. Thomas, J. Whidborne, and A.
Cooke, “On-Board Trajectory Generation for Collision
Avoidance in Unmanned Aerial Vehicles”, Proceedings of
the IEEE Aerospace Conference, pp. 1-14, 2011. DOI:
10.1109/AERO.2011.5747526
[36] T. Yu, J. Tang, L. Bai and S. Lao, “Collision Avoidance for
Cooperative UAVs with Rolling Optimization Algorithm
Based on Predictive State Space”, Applied Sciences, vol. 7,
issue 4, no. 329, 2017.
J Intell Robot Syst
[37] A. Gardi, R. Sabatini and S. Ramasamy, “Multi-objective
Optimisation of Aircraft Flight Trajectories in the ATM
and Avionics Context”, Progress in Aerospace Sciences,
vol. 83, pp. 1-36, 2016. DOI:
http://dx.doi.org/10.1016/j.paerosci.2015.11.006
[38] C.K. Lai, “A Novel Collision Avoidance Logic for
Unmanned Aerial Vehicles using Real- Time Trajectory
Planning”, PhD Thesis, Cranfield Univeristy, UK, 2014.
[39] S. Ramasamy, R. Sabatini and A. Gardi, “LIDAR Obstacle
Warning and Avoidance System for Unmanned Aerial
Vehicle Sense-and-Avoid”, Aerospace Science and
Technology, vol. 55, pp. 344–358, 2016. DOI:
10.1016/j.ast.2016.05.020
[40] R. Sabatini, T. Moore and C. Hill, “Avionics Based GNSS
Integrity Augmentation for Mission- and Safety-Critical
Applications”, Paper presented at 25th International
Technical Meeting of the Satellite Division of the Institute
of Navigation: ION GNSS-2012, Nashville (Tennessee),
September 2012.
[41] A. Gardi, S. Ramasamy, R. Sabatini and T. Kistan,
“Terminal Area Operations: Challenges and
Opportunities”, Encyclopedia of Aerospace - UAS, eds R.
Blockley and W. Shyy, John Wiley& Sons: Chichester,
2016. DOI: 10.1002/9780470686652.eae1141
A preview of this full-text is provided by Springer Nature.
Content available from Journal of Intelligent & Robotic Systems
This content is subject to copyright. Terms and conditions apply.
