In: Civil Engineering
For this deliverable, you will evaluate software suitability, human systems integration, survivability,
and interoperability issues for the SWARM Unmanned Aerial Vehicle (UAV) system. Provide a report that summarizes your findings for this analysis.
5.1 Software suitability
For this task, in the Maier and Rechtin text, in other course material, and your general knowledge of software.
Identify three software suitability characteristics that you select as the highest priority for the UAV System. For each characteristic, justify why you selected it as a high priority for this system.
For each software suitability characteristics identied above, construct a well-written sample software suitability requirement that could be included in the system requirements for the UAV SWARM system.
For each of your software suitability requirements, discuss the test and evaluation activities needed to provide the basis for the final acceptance of the SWARM UAV system. Include estimates of the cost, duration, personnel, equipment, ranges, and other factors necessary for test, evaluation, and acceptance.
For each estimate, provide a rationale.
5.2 Usability, survivability, & interoperability
Identify one potential issue for each of the following ilities that applies specifically to the SWARM UAV system, and discuss a feasible solution to address the issue
usability
survivability
interoperability with an external system
5.3 Overall suitability
Clearly state you nal conclusion about the overall suitability of the UAV SWARM system as a research platform. Provide justication of your nal conclusion by using the key results from your previous analyses done for the system.
Controlling UAV swarms via human supervision is of
great interest to the US military. This paper reports
on a project that models control of a fleet of
unmanned aerial vehicles (UAVs) with minimal user
intervention via simulation. The project simulates a
situation where multiple UAVs must locate and track
multiple ground vehicles. Once located, a ground
vehicle must be “scanned” for a certain length of time
to simulate gathering information about the target. If
the target is lost and later reacquired, scanning may be
picked up where left off. Any UAV may contribute to
the scanning process. However, multiple UAVs
scanning a single target at the same time will not
speed up the scanning requirement. Once completely
scanned, the target is of no further significance and is
removed from consideration.
We have implemented a simulation of UAV swarms
in Java. Figure 1 shows a snapshot of our simulated
environment and its GUI control.
The system of UAV agents is essentially a hive mind
organization controlled by the main loop of the
program. However, individual decisions are controlled
by the individual UAVs. The main loop is a turn-
based system containing three main sub-phases. Each
UAV is instructed in turn to perform a sensor pulse to
locate nearby targets. Phase two consists of a
negotiation where UAVs determine their best targets
and settle on which UAV should be allowed to track
which target. In the third and final phase, the action phase, each
UAV is allowed in turn to perform its
chosen objective. UAVs are capable of high speed
and maneuverability relative to the target vehicles.
The exact maximum speed is parametric. Since power
consumption is not considered in this simulation, the
UAV generally fly at their maximum speed.
However, when nearing its chosen target, a UAV will
slow down to avoid overshooting the target position.
The UAV’s movement is constrained to two degrees
of freedom (the horizontal plane), but is quite
maneuverable within those parameters. It is able to
move in eight directions – North, South, East, West,
Northeast, Southeast, Southwest, and Northwest – and
can move in multiple directions each turn up to its
movement limit. The method used to move the UAVs
is rather simple. At the end of each turn, each UAV is
given the opportunity to move toward its intended
destination. The difference between the intended and
actual positions is computed. The UAV then moves to
minimize that difference. The UAV is only able to
sense ground targets. The range is specified by a
parameter and generates a circle of data around the
UAV. Any target not completely within the sensor
circle is not seen. The human operator is allowed to
give two types of orders directly to individual UAVs.
It can force a UAV to track a specific target. The
UAV will then travel to the last known location of that
target and attempt to acquire it. Second, the operator
can ban a UAV from tracking a target. Regardless of
its perceived appropriateness, the target cannot be
tracked by the UAV until the ban is lifted. These
commands are rarely issued. We found that the UAVs
are generally quite capable of locating targets without
any specific operator commands. The only times
where these commands were required were when
multiple UAVs were deadlocked in a bidding war or
when a UAV became “too dedicated” to reacquiring a
lost target. This latter situation is easily remedied via
an appropriate UAV personality parameter.
- Human operators will ultimately control UAV swarms
at a high level. Our work takes a modest step toward a
multi-tiered architecture that provides a high level
control using social and personality parameters. The
transition from strictly remote-controlled vehicles to
intelligent, goal-directed, cooperative agents requires
a well-defined hierarchy of autonomies.
The Commitment parameter (we also called it
dedication) could be further automated since it has a
fairly straightforward use. There may be
circumstances where the operator would want to
override the internal adjustments.
Second, Temperament (we also called it disposition)
appears to be a difficult parameter. Intuitively, one
would assume that a low disposition would be
appropriate in a setting where there is an
overabundance of UAVs. This “bad temper” would
allow the UAV to stick up for itself in a situation
where there would most likely be many other UAVs
that would try to take its target. However, we were
unable to determine any significant benefit from
adjusting the disposition values.
Finally, it would appear that Sociability and
Conformity allow a wide range of adjustments. We
have determined that positive sociability and negative
conformity can be detrimental in many situations. In
addition, fine-tuning the settings does not appear to
have a significant effect. We found the best settings
to be either relatively small or large positive values
(i.e., no middle value) for Conformity and negative
values for Sociability. Regarding reducing
argumentativeness, we were unable to produce enough
arguments to warrant user interdiction. With
dedication, conformity, and sociability characteristics,
UAVs perform a subset of behaviors that would
normally require exhaustive human supervision