Вы находитесь на странице: 1из 6

Phased Array Radar Resource Management:

A Comparison of Scheduling Algorithms


S.L.C.Miranda, C.J.Baker, K.Woodbridge and H.D.Griffiths
University College London
Torrington Place
London, UK
WClE 7JE

Abstract- Scheduling is an important sub-problem of radar management as there is a strong correlation between how tasks
resource management as there is a strong correlation between should be carried out and the time available to carry them out.
how tasks should be carried out and the time available to perform
them. In this paper, two scheduling algorithms presented in the There have been a number of reports in the literature
literature are compared to investigate whether there are examining the design of efficient scheduling algorithms by the
significant differences in their performance related to the use of neural network, artificial intelligence, Operations
allocation of radar time resources. We developed a radar model Research (OR) theory, fuzzy logic and others [3, 4, 5. 9, IO].
applying a modular architecture to use the same operating and Although using different techniques, most of these approaches
environment conditions in the analysis. The results suggest that aimed to optimise the number of tasks the multifunction radar
both algorithms provide similar performance, apart from minor
differences explained here. can undertake within a given time interval.
The purpose of this paper is to address a comparison of two
of the differing approaches described in the literature,
I. INTRODUCTION evaluating their behaviour as a function of the radar timeline.
In recent years phased array antenna technology has been Initially, Section I1 describes the architecture of the
rapidly maturing and this form of transduction is set to become simulation used in the comparison. Next, Section 111 explains
the nomi in complex and advanced radar systems. Whilst the scheduling algorithms used in this analysis [3, 51. Finally,
being able to instantaneously and adaptively position and Section IV presents the comparison of the algorithms.
control the beam has clear advantages, it also brings about a
new set of challenges. The radar is now able to adapt its
parameters according to how it perceives its operating 11. ARCHITECTURE
environment. Optimisation of this adaptivity is necessary if The architecture used in the comparison is presented in fig.
phased array radar systems are to realise their full potential. A 1. It provides an environment in which any scheduling
simple example might be the need to increase the dwell for algorithm can be represented and examined against a given
fading targets in order to maintain a given tracking accuracy. scenario. This study considered the behaviour of one face of a
However, a multifunction radar has many tasks to carry out multi-face phased array antenna radar system. For
and must therefore make a decision as to how long a dwell can simplification, the analysis included both radar functions of
be allowed whilst still meeting its other requirements. The surveillance and tracking. A modular approach was used in
process of making these decisions and determining their developing the simulation model.
allocation as a function of time is known as radar resource
management. In this paper we concentrate on the allocation of
radar resources and this is usually referred to as “scheduling”.
In particular, we examine and compare the performance of
two very different approaches to scheduling that have been
reported in the literature. Scheduling has to consider a number
of constraints such as time and energy in order to maximise the
number of the tasks that are able to meet their desired
deadline. Moreover, other aspects must be considered such as
a control mechanism to ensure urgent tasks are executed
promptly. Additionally, some time critical tasks have to be
performed within specified windows of opportunity although
the duration of the tasks can vary. Thus, it is seen that
scheduling is an important sub-problem of radar resource Fig. 1. Block diagram of the simulation architecture used in the
comparison

0-7803-8234-X/04/$17.000 2004 IEEE 79


This modularity allows the comparison of the two tasks. Hence, a better use of the radar time could be achieved.
scheduling algorithms by using the same initial conditions and The flow diagram of this algorithm is presented in fig. 2.
tactical characteristics with respect to the targets and
environment. Thus, it was possible to evaluate their behaviour
in several situations under identical circumstances.
In the system used here, the user defines the boundaries of
the volume of coverage of the antenna face. Based on this
information, the surveillance function estimates the number of
radar beams that will be necessary to survey that volume. A list
of task requests is generated considering the desired
surveillance performance of the radar system. The surveillance
manager is fed by the task list, maintains a queue of inactive
tasks (not scheduled yet) and provides the scheduler with a
smaller queue of requests that are close to their due time of
execution.
Sinularly, the track function calculates the update times of
the targets under track and feeds the track manager with a
queue of tasks to be scheduled. The track manager also
maintains a list of inactive track tasks, which are sent to the
scheduler when they are close to their due time of execution.
Both surveillance and track manager select from the
waveform database the parameters to be used in the
transmission of the radar pulses associated to each radar job.
The scheduler is fed by the queues of track, plot
confirmation and surveillance task requests. It creates the list
of scheduled tasks based upon the criteria defined by its
algorithm, considering task priorities and time constraints. A
feedback loop is provided between the output of the scheduler
and the surveillance and track functions, so the next update
times related to those tasks can be calculated.
The block named "Operator Csr Strategy" accounts for the
overall preferences related to the mission and to the decisions
based upon evaluation of the tactical scenario.

111.DESCRPTION
OF THE SCHEDULING ALGORITHMS

A. Oman type scheduler


The first of the scheduling algorithms is that proposed by
Orman et al. [5], and is centred on coupled-tasks. A coupled-
task is a job consisting of two different operations separated in
time by a specified interval. Thus, each coupled-task can be
represented by the processing time of the first task
(transmission of radar pulses), the separation time between the
tasks and the processing time of the second task (reception of
radar echoes). The scheduler can organise a queue of tasks to
be executed in any way provided that two tasks do not overlap
in the radar timeline. The separation time (or idle time) and the
reception time can be estimated according to the radar function
considered and additional information such as range of target,
boundaries of the surveillance region and the transmission time
of tasks.
Once the duration of these is determined, interleaving
algorithms may use the idle time for scheduling additional Fig. 2. Flow diagram of the Oman type Scheduler

80
B. Butler type scheduler IV.COMPARISON
The second algorithm was described by Butler [3],
The comparison of the algorithms was performed
considering some improvements on a previous algorithm
considering identical radar task prioritisation list and regions
proposed in [SI.A timebalance scheme controls the scheduling
of surveillance. Thus, the performance of both schedulers
process of the requested tasks. There is a timebalance related
to every radar job, indicating to the scheduler how much time could be compared using the same simulation conditions.
is owed by the radar to that job. A job consists of several tasks Both algorithms used the following prioritisation order:
and is usually associated with surveillance of a region of
1. Track maintenance (highest priority).
coverage or with keeping a target under track.
2. Plot confirmation.
A task is a group of activities that are non-coherently
integrated to give detection. A task can be divided into looks, 3. Track initiation.
consisting of one or more activities. A modified version was
4. Track update.
used in this work, in which the timebalance is computed in
seconds, and represents the earliness or lateness of a task. 5. Surveillance (lowest priority).
When a task is late, its timebalance is positive. On the other
hand, an early task has a negative timebalance. Finally, a task The surveillance performance was determined either by the
due to be undertaken at the exact moment has a timebalance radar operator or by mission requirements and three regions of
equal to zero. If the job table receives a new task request that surveillance were considered in this work. For this
should not be scheduled for n s, the timebalance associated to comparison, a configuration composed of three sectors of
that task should be set to - n s. The timebalances of all tasks surveillance spanning a 90" azimuth sector was considered.
Region 1 spanned 0-10" in elevation, region 2 spanned in 10-
are incremented as time elapses. A new job is always inserted
in the job table before its due time of execution and hence, 20" in elevation, and region 3 spanned in 20-50" in elevation.
with negative timebalance. The flow diagram of this An example of the surveillance parameters used for these
regions is presented in Table I. In this case, the overall
scheduling algorithm is presented in fig. 3.
surveillance performance was set to loo%. According to the
desired surveillance performance considered in the analysis,
the surveillance function time was adjusted, changing the
occupancy in each of the regions of surveillance.

TABLEI
SCHEDULING PARAMETERS OFTHREE REGIONS
OF SURVEILLANCE

Broadside Number Function

. Region

1
Dwell Time

1 nis
kupancy

20%
oPBeans

108
Time

0.143 s
2 2 nis 30% 135 0.359 s

A. General Aspects
Both algorithms only allocated radar time to tasks or
functions of lower lever of priority if there was available time.
Therefore, in a situation where several targets were consuming
a large amount of radar resources, the detection performance
of surveillance regions was degraded. All the tracking tasks
were scheduled and the resulting available radar time was
shared amongst the surveillance jobs. As a result, if all radar
lookinthe
task? resources were used for tracking, no surveillance would be
performed.
The Orman type scheduling algorithm attempted to
maintain the specified surveillance performance even if there
were available resources that could be used for surveillance
Fig. 3. Flow diagram of the Butler type algorithm and improving detection (Fig. 4). Then, it was possible to
perceive idle times in the radar time line because no tasks were
scheduled early in order to make use of all the available time
resources. On the other hand, the Butler type scheduler did not

81
leave any idle time on the radar time line. When the desired B. Stressing Conditions
surveillance occupancy was a, less than 100%, and the radar The schedulers were also compared in situations where
load associated to all other functions was less than 1-a, there were not enough radar resources to perform all the
surveillance tasks were scheduled earlier and the surveillance requested radar tasks. In this case, the desired surveillance
occupancy was increased until !.he total radar time line was occupancy was set to loo%, which meant that the entire radar
utilized (Fig. 5). Thus, spare capacity was used to improve the time line should be used to perform that function. As targets
surveillance performance. were detected, it was expected that the surveillance
performance should be degraded to free the radar resources
However, when the tracking load increased due to a great
and requested tracking tasks be performed. The use of such
number of detected targets, the surveillance jobs were
desired surveillance occupancy was to overcome the drawback
progressively delayed and the detection performance
related to idle times when using the Ornian type scheduler, as
degraded.
mentioned in the previous subsection.
The results of the simulations are presented in figs. 6 and 7.
0 Region 1 These figures show that, as new targets were detected, the
0 Region2
1- c. Region3 occupancies of the surveillance regions decreased, causing a
x Trackmg
degraded detection performance. Under these conditions the
performance of the two approaches was broadly similar
However, the Orman algorithm scheduled all the tracking tasks
as close as possible to their due time of execution and the
allowed earliness and lateness related to these tasks was not
enough to prevent the insertion of idle times in the radar
timeline. As a result, the overall occupancy was slightly
smaller than 100%.This effect was not seen in the results from
the Butler type scheduler that utilized the total amount of the
radar time.

0 2 4 6 8 10 12 14 16 18 20
Time (sec)

Fig. 4. Results froiii the Orman type scheduling of surveillance and 0 Region1
tracking tasks in three regions of coverage. Required Surveillance (3 Rqpon3
occupancy= 80%
0 Total Occu n

0 Region 1
3 Regon2
0 Regon3

A Total Occu

5 10 15 20 5
Time (sec)

Fig. 6. Results from the Onnan type scheduling of surveillance and


tracking tasks in three regions of coverage. Required Surveillance
0 at o o o o o 0 0 0 0 occupancy= 100%

Additionally, an analysis was performed in the ability of


both algorithms to schedule radar tasks close to their desired
times of execution when the multifunction radar is operating
under stressing conditions.
Like the situation analysed before, the achieved
surveillance performance was degraded as new targets were
detected and put under track. It is shown in fig. 8 that the
Orman algorithm was able to schedule both plot confirmation
and tracking tasks using very small earliness and lateness.
However, this could be at the expense of reduced surveillance
-
perforniance.
70t
70

60 -
% 50-
OI

8B
E- 4 0 -
0 RBgion1

4
Tracking
Total mcu
4 30-

20 -

10 -

I
-O0015
-0°015 -0.k R50!-
-Oil 0 0.005
0005 0.01
001 0.015
0015 0.02
002 0.025
0025 003 < 35

Fig. 9. Number of Tracking Tasks x Earlinesshteness when using the


Butler type scheduler

-c-. '
5 10 15 20
Time (sec1

Fig. 7. Results from the Butler type scheduling of surveillance and V. CONCLUSION
tracking tasks in three regions of coverage. Required Sundlance
occupancy= 1008 The scheduler is an important component of the resource
manager in multifunction array radars. In this study, two
different approaches previously presented in the literature were
developed and compared. The modularity used when creating
the radar model allowed the use of the same tactical and
operational conditions in the comparison with respect to
targets and environment.
From the results presented in this paper, it is seen that both
schedulers had similar performance under the conditions used
in the simulations, especially when analysing their ability to
schedule the tracking tasks as close as possible of their desired
time of execution. The difference in their performance was
related to how the radar timeline would be used by the
scheduler. The Omian type scheduler left some idle times in
the radar timeline even when operating in a stressing situation.
- 0
As a result, the radar resources were not fully allocated to
-0015 -001 -0005 0 0005 001 0015 002 0025 003 5
Earlinesaaleness -Time (sec) perform radar functions. In contrast, the Butler type scheduler
made use of all the available radar time resources by fulfilling
Fig. 8. Number of Tracking Tasks x Earlinesshteness, when using the entire radar timeline. By scheduling some tasks earlier or
the Onnan type scheduler
later than their due time of execution, this algorithm provided
a better utilization of the radar time. A problem could he
Conversely, fig. 9 shows that the Butler type scheduler indicated when using this kind of scheduler in low and medium
allowed not only greater earliness and lateness when power radars. In this case, the surveillance tasks can be very
scheduling the tasks, but also planned a greater number of long (about tenths or hundreds of milliseconds) and then the
early and late tasks. Nonetheless, these aspects were not plot confirmation and track maintenance tasks could be too
relevant as the great majority of tasks were scheduled with a late. To overcome this problem, it was proposed in the
*
displacement of only about 5 milliseconds of their due time description of the scheduler that in such situations the long
of execution. This did not represent a significant degradation dwell times should be divided into smaller slices of about 5
in the tracking performance. milliseconds. Thus, when a plot confirmation was requested
the lateness related to the scheduling process could be
nunimized.

03
REFERENCES
Barton, D.K., Modern Rudur Sysfeins Anulysi.~,Artech House, 1988.
Billetter, D.R.. Mu/tifuncfiortArray Rudur, Artech House. 1989.
Butler, J. M., Mu/fi-FuncIiori Rudur Trucking mid Control, University
College London. PhD Thesis, 1998.
Izquierdo-Fuente, A.; Casar-Corredera, J.R.. “Optimal radar pulse
scheduling using a neural network“, IEEE Infeniufionul Conference 0 1 7
NeurulNefworks.vo1.7, pp. 4588 - 4591. June-July 1994.
Oman, A.J. , Potts, C.N., Shahani, A.K. & Moore, A.R., “Scheduling
for a Multifunction Array Radar System”, European Ji~uniu/ of
Operurionul Reseurch, No. 90, pp. 13-25, 1996.
Onnan, A.J.. Shahani. A.K. and Moore, A.R., ”Modelling for the
Control of a Complex Radar System”. Compufer.s arid Operations
Reseurcli. Vol. 25, No. 3, pp. 239-249. 1998.
Sabatini. S . and Tarantino. M.. Mu/rifuncfiori Arruj Rudur - Sjsfent
Design arid Anulysis. Artech House. 1994.
Stafford, W.K., ”Real-Time Control of a Multifunction Electronically
Scanned Adaptive Radar (MESAR)”. IEE Colloquium 011 R e d Tinie
MUJiUgeJJ7enfofAdupfivcRudur Sjsfenw. pp. 711-715, June 1990.
Stromberg D.; Grahn P., ”Scheduling of Tasks in Phased Array Radar”,
IEEE I~~fenintiortulSviiposiunt on Pltosed Arruy Sjsferns arid
Tcchnologj. pp. 3 18-32 1. October 1996.
[lo] Vine. M. T.. “Fuzzy Logic in Radar Resource Management”.
Mu/fifurictioriRudur cmd Soriur Sensor Muriugeritmf Techniques (Ref.
NO. 2001/173). IEE. pp. 5/1-514. 2001.

84

Вам также может понравиться