26-05-2012, 03:43 PM
DEVELOPMENT OF AN EXPERIMENTAL PLATFORM FOR TESTING AUTONOMOUS UAV GUIDANCE AND CONTROL ALGORITHMS
DEVELOPMENT OF AN EXPERIMENTAL.pdf (Size: 1.22 MB / Downloads: 127)
Introduction
Motivation for Autonomous Cooperative Control of UAVs
Current Search and Destroy Mission
Since the end of the Cold War, the United States has found itself locked in urban
warfare and completing military missions other than war at a faster pace than ever before.
As a result, tactics once used in the open battlefield are no longer considered viable when
fighting against enemies without uniforms in large, mostly civilian, urban settings. One
current technology push to give the U.S. Armed Forces an advantage over their enemies
in this type of environment is the development of autonomous unmanned aerial vehicles
(UAV) and autonomous unmanned micro aerial vehicles (MAV). To best allocate these
invaluable resources in a battlefield setting, cooperative control of multiple UAVs &
MAVs is being explored at the Air Force Institute of Technology. Some benefits of using
cooperative UAV fleets include search redundancy, capability to search larger areas
quicker, multiple targets can be simultaneously tracked, and operators can be kept out of
the extreme danger of some of today’s urban war zones. Also, as suggested by three
researchers at Colorado State University (Richards, Whitley, and Beveridge, 2005), if the
UAV used for a particular mission is prone to failure, it might be cheaper to use multiple
inexpensive UAVs instead of one costly search system.
As mentioned above, the current enemies of the United States and its allies do not
follow established rules of war, and thus it is possible for almost any vehicle, building, or
person on the ground in a region of conflict to be a target. When terrorists use hospitals
or mosques as their hideouts or hide behind women and children, the line between
civilian infrastructure and legitimate targets, according to the rules of war, becomes
murky. To ensure collateral damage is minimized in this type of situation, UAVs must
be able to discern the actual targets from those entities that at first glance appear to be a
target, but are actually part of the civilian infrastructure being used illegally. It is this
point that makes the cooperative control aspect of UAV target searching critical to ensure
that a UAV has found a legitimate military target before it attempts to destroy it. As the
U.S. continues to fight in urban environments around the world, the need for this
technology will keep growing and the tolerance for error on the battlefield and in the
political arena will keep shrinking.
Full Scale Autonomous UAV Experimental Work
Even though this autonomous and cooperative technology is being heavily
researched and funded by the US Department of Defense, the UK Ministry of Defence is
also working to develop the same type of technology. As recently as 30 October 2006,
Qinetiq, a UK defence contractor, completed an in flight demonstration of the UAV
Command and Control Interface (UAVCCI) by using a BAC 1-11 1960’s era jetliner to
simulate a fighter pilot managing four UAVs as well as their own jet. To add realism to
the test and prove the functionality of the UAVCCI, the pilot in control of the BAC 1-11
sat in the back of jet where he controlled it as well as the UAVs.
The UAVCCI system is designed to allow for semiautonomous flight of the UAVs so
pilots can easily control their jet, without worrying about always giving commands to the
UAVs. When the UAVs do not get commands, they are programmed to fly straight and
level, but the pilot has the ability to direct them through a moving map and push buttons.
With these controls, the pilot can direct the UAVs to loiter, start a search, or attack. This
test showed that cooperative and autonomous control of UAVs can occur not only from a
ground station, but also from the cockpit of a military jet closer to the fight. The pilot
would then be able to use the displays as well as the real time battlefield environment to
give the UAVs specific commands (Marks, 2006). As previously noted, the remote or
autonomous control of military assets will help greatly in the Global War on Terrorism to
keep US and allied service members farther from their nameless and uniformless enemies
and their treacherous improvised explosive devices (IEDs). According to Icasualties.org,
a non military website that provides DoD verified information on Operation Iraqi
Freedom casualties, 1183 of the 3085 U.S. deaths through the end of January 2007
(roughly 38 percent) have been caused by IEDs (iCasualties.org, 2007). Development of
autonomous search vehicles will help mitigate the effects of this deadly tactic in the
future. In fact, the research in this thesis will help the Pentagon towards their goal of
having one third of their military assets “robotic or remotely controllable by 2015 (Marks
2006).”
While the physical integration of hardware and software of sensors into an
unmanned vehicle can be quite complex, the operational concept of the system is quite
straightforward. The system can be thought to be analogous to a self checkout area at a
grocery or retail store. With the self checkout process one operator monitors multiple
checkout stations and only intervenes if the customer at the station is having problems
that they cannot solve themselves. In the autonomous UAV search group concept one
operator will have the capability to monitor multiple UAVs to ensure that the group is
working towards its mission objectives, and only intervenes if there is a problem that one
or more of the UAVs cannot fix on their own.
Autonomous UAV Cost /Benefit Analysis
Many benefits come from operating UAVs in the autonomous regime. The
simplest advantage comes from the ability to allocate less personnel to operate more
UAVs. When UAVs are flown manually by an operator, there is at least one human for
each UAV and often several. If one operator can monitor 3-4 UAVs, then more UAVs
can be utilized with the same number of operators. This operator can also perform this
job from any ground station within communications range (radio, satellite, etc) of the
UAV fleet they are controlling, thus keeping them off of the battlefield. Other
advantages include being able to perform coordinated searches over larger areas than a
single UAV could search, and engaging multiple targets with multiple vehicles in the
same search.
Some challenges involved in fielding networked UAV systems include the
development of adaptable operational procedures, as well as planning and deconfliction
of assets. As these technologies progress, UAVs will be able to make better allocation
and targeting decisions on their own. However, autonomous UAVs will always have the
chance to make poor decisions because they are taking data acquired through real time
sensing and computing solutions based on human produced algorithms to make targeting
decisions that could result in a bad target selection as well as damage to or outright loss
of the air vehicle (Vachtsevanos, 2004). While some of these algorithms will possibly
involve multiple checks from other UAVs in the fleet before engaging targets, they will
never be foolproof instructions to ensure a wrong target is never hit. Because these
algorithms operate independent of human control, they must continually be updated,
refined, double checked, and monitored to keep up with the ever changing conditions on
the battlefields of the world.
Previous Applicable Research
The current state of the art in Unmanned Aerial Vehicle (UAV) targeting research
at the Air Force Institute of Technology (AFIT) has implemented analytical concepts into
robust multi-warhead and multi-vehicle Matlab/Simulink simulations. Since many AFIT
theses as well as a multiple dissertations have explored the autonomous UAV targeting
concepts and simulations, the next logical step in the process is to develop hardware to
prove it is possible for autonomous target recognition (ATR) systems to properly detect
and identify objects. This experimental validation of theoretical concepts will help the
Air Force move towards implementing robust targeting algorithms into operational
autonomous UAV fleets in the future.
Some of the topics of the wide area search research involve optimal path
planning, applying probability theory to the UAV fleet, conducting simulations using the
Multi-UAV simulation test bed (Rasmussen, Mitchell, Chandler, 2005), automatic target
recognition (ATR), performance under limited communication, non-linear control of
UAVs in close coupled formation, and most recently dynamic scaling of UAVs. Each
topic contributes greatly to cooperative control of autonomous UAVs, but only ATR and
dynamic scaling will be expounded in the present research. ATR theory will be used in
the development of a simple target identification algorithm that a ground based search
vehicle platform will use to identify targets and dynamic scaling will be used to ensure
that the vehicle has the proper dynamics to reasonably represent a flyable experimental
UAV system.
Autonomous Target Recognition
To better understand the logic behind cooperative UAV targeting algorithms, the
concept of a confusion matrix must first be introduced. It has been used in the work of
Dr. David Jacques and Dr. Meir Pachter (2003) to provide conditional probabilities for
each possible outcome when a search vehicle sweeps a given area and encounters an
object it determines is not part of the background. For simplicity, the concept will be
explained below using a single target scenario.