TimeStep based Distributed Simulation Time Synchronization Method for Moving Objects

1
Time-Step based Distributed Simulation Time
Synchronization Method for Moving Objects
Atsuo Ozaki, Masashi Shiraishi, Shusuke
Watanabe, Minoru Miyazawa, Masakazu Furuichi,
and Hiroyuki Sato Mitsubishi Electric
Corporation
2
Outline

• Background, Motivation and Objectives
• Issues of Moving object (MO) simulation
• DTSSevent-aware Dynamic Time Step
• Synchronization Method
• Discussion of the best HLA time management
• for implementing DTSS
• Performance evaluation
• Considerations
• Conclusion and future works
• Work in progress

3
Background

• High performance and high fidelity computer
  simulation
• for a large number of moving objects (MOs) is
  eagerly
• desired by the defense and disaster prevention
  community
• MO e.g. aircraft, vehicles, ships,
  people and so on.
• The standardization of simulators is also eagerly
  desired to
• enhance the reuse of simulators for preventing
  rising costs,
• or to enhance the interoperability, the
  integration of the
• different types of simulators by different
  organizations.

4
Motivation Objectives

• Utilize the distributed simulation technology
  with
• a multiple computer environment to speed up
  the
• MO simulation.
• Employ the HLA, the standard for distributed
  simulation
• in order to standardize the simulators.

5
Issues of MO simulation
It is difficult to predict events and their
logical time in a large-scale MO simulation,
since MOs interact with one another in a
complicated manner.
This kind of simulation does not in general
employ an event-based method, but a time step
method
6
Issues of MO simulation - Time Step Method -
In general, each MO has a different speed, hence
employing an identical ?t and simulating every
MO by this ?t is inefficient.
MASF (Method for More Efficiently Achieving a
Simulation Fidelity)
With MASF, MOs can avoid GTAs (Go Through
Accident) even when MOs with different speeds are
approaching each other
Event Queue
Events
000 001 002 003 004 005 006
Low speed MO
Schdular
s
s
e.g s 1 nautical mile
Simulation Engine
Simulation space
s Unit of distance (simulation fidelity)
7
Issues of MO simulation - MASF -
In the case of occurring causality error.
t1 MOa detects MOb and, sends an
interaction message as attacking onto MOb. t2
(Normally, MOb must be destroyed by executing the
message from MOa, though it is not
destroyed at this time.) t3 MOa detects MOb
and, sends an interaction message to MOb again.
8
Issues of MO simulation - MASF -
Methodology for solving the causality error.
t1 MOa detects MOb and, sends an
interaction message as attacking onto MOb.
t2 MOb is destroyed by executing the message
from MOa. t3 MOa confirms the destruction
of MOb.

• It is difficult to predict a beginning time of
  SoD (Scene of Detection)

9
DTSS event-aware Dynamic Time Step

Synchronization Method
Concept
DTSS does not calculate exactly the beginning
time of SoD, but calculates the fastest
possible beginning time (FPBT) of SoD
10
DTSS Algorithm for calculating Dt
11
DTSS Advantage of employing base?t.
12
Discussion the best HLA time management
for implementing DTSS
(1) HLA-TimeStep(TS)
(2)HLA-EventBased(EB)
13
Discussion Implementation of DTSS
by HLA-TS adjusting Dt
Example process flow of MOs entering CC
CC (Critical Condition)the simulation time of
MO is FPBT of SoD or greater.

• Calculated distance (D) and beginning time
• of CC will be different for MOi and MOj
• Accuracy of the beginning time of CC is
• improved when MO calculates this time
• later than other MOs.

14
Discussion Implementation of DTSS
by HLA-TS adjusting Dt
Example process flow of MOs in CC

• There is no advantage in sending ahead
• an interaction message and equality
• can be maintained between MOs since
• they use the same ?t after entering CC.

15
Performance evaluation - Scenario -

• Basic scenario of a war-game simulation
• Two teams compete in tactics and strategy using
  information
• gained through the simulation
• FPBT of SoD, which is assumed as the battle area,
  is identical for all MOs

16
Performance evaluation - Parameters -
Table 1. Parameters for simulation times
Table 2. Computation and communication loads
17
Performance evaluation - Machine environment -
Fast-Ethernet(100Mbps)
PC5RTI
Table 3. Performance and environment of the
machines
18
Performance evaluation - Results (1/2) -

• DTSS can be executed with little extra
  computational cost.
• The ideal plus approximately its 1.

19
Performance evaluation - Results (1/2) -
Execution time when the number of time steps of
SoD from FPBT is 20( small), 50( medium),
80( large).
20
Considerations
Conventional method
DTSS
MO-x
MO-x
Detection
Movement
Calculate Dt
Dt1
Communication
Simulation Time
Dt1
Dt ? larger
Detection
Movement
Calculate Dt
Communication
21
Conclusions

• Proposed DTSS can speed up simulation by
  increasing ?t
• without introducing causality errors.
• Discussed a suitable HLA based time management
  method to
• implement DTSS. gt HLA-TS adjusting ?t by
  introducing
• the
  concept of the event base is the best.
• DTSS can be executed within the ideal time plus
  its 1 over
• -cost on this HLA distributed computing
  environment when
• a basic scenario of war-game simulation is
  employed.
• gt The criterion for determining when DTSS is
  superior to
• the conventional method has been
  introduced.

22
Future works

• Every distance value (D) between MOs does not
  need to be
• calculated when the distance between MOs is
  beyond the range
• of detection.
• gt The evaluation of improved DTSS based on
  this concept
• There are many kinds of objects besides MOs in
  the real world.
• (We cleared HLA-EB is utilized efficiently as
  long as the objects send no messages.)
• gt Systematizing these synchronization methods
  is a great future
• work.

23
Work in progress
Apply DTSS to RoboCupRescue Simulation

• MOs
• Civilian72
• Ambulance team5
• Fire brigade10
• Police force10
• Simulation Time
• 300(min)
• (300 60(sec) steps)
• Visual range
• 10m

Ratio of SoD
Viewer of RoboCupRescue Simulation System
24
Work in progress
Tendency for the number of acting MOs and MOs in
SoD
Calm down the rescue
Busy with the rescue
25
Work in progress
Execution time can be expected to reduce up to
37.4 by applying DTSS. We continue to
work on evaluating DTSS