UTSAF: A Simulation Bridge between OneSAF and the Unreal Game Engine

1
UTSAF A Simulation Bridge between OneSAF and the
Unreal Game Engine

• J. Manojlovich, P. Prasithsangaree, J. Chen, and
  M. Lewis

This project is supported by AFOSR contract
F49640-01-1-0542.
2
Agenda

• Introduction and Motivation
• Some Background
• Challenges in Bridging Distributed Simulation
• Our Proposed UTSAF Architecture
• Snapshots from our testbed
• Conclusions

3
Background

• Modular Semi-Automated Force (ModSAF) and OneSAF
• SIMNET, DIS and PDU
• Unreal Tournament (UT) Game Engine
• RETSINA

4
ModSAF

• Military simulation environment
• Provide a platform within which all services
  (Army, Air Force, Navy, etc.) could expand the
  synthetic battle space
• Provide an architecture to support a community of
  SAF developers who would expand the battle space
  through the integration of individual code
  modules

5
(No Transcript)
6
ModSAF Communication

• IP multicast
• ModSAF uses the Distributed Interactive
  Simulation protocol (DIS, IEEE1278 1993).
• Protocol Data Unit is basic packet of information
• There are roughly 100 different classes of PDU,
  but generally only the Entity State PDU is
  interesting for simulation purposes

7
Distributed Interactive Simulation

• The purpose of DIS is to create and maintain an
  interactive virtual environment for the following
  purposes
• Training
• Replace and supplement field training
• Testing
• Planning
• Standard for communicating a ground truth
  database among hosts

8
Distributed Interactive Simulation (cont.)

• Basic unit of DIS is the entity
• These are typically vehicles in the simulation,
  such as tanks, trucks, aircraft, or missiles
• DIS also specifies how to transmit terrain
  information, and electromagnetic emissions like
  radio and radar

9
What is Unreal Tournament?

• UT is a multiplayer video game
• Produced by Epic games (http//www.epicgames.com)
• Runs on Windows, Linux, Macintosh, PlayStation,
  and Xbox
• http//www.unrealtournament2003.com

10
(No Transcript)
11
Why Use UT?

• By using UT as a base, we can avoid needless
  reinventing while benefiting from a proven
  development platform
• UT is designed for programming, and many tools
  and resources are available to assist outside
  developers
• UT runs well on commodity hardware and software,
  such as Windows and Linux

12
Why Use UT? (cont.)

• UT has numerous user interface enhancements over
  ModSAF
• Three dimensional, instead of two
• Ability to attach view to entity
• Much easier to dynamically change the field of
  view in UT
• Can use other projects that are built on UT

13
Problems

• When the virtual simulators are not exactly the
  same, it gets harder to pass entity information
  between these simulators
• UT was not designed for full external control, so
  the built-in functionality must be overridden
  many times

14
Attribute Transfer Between Virtual Domains

• What is absolutely necessary?
• Entity identification
• Position in virtual world
• Coordinate systems
• What else is useful?
• Entity velocity, appearance, orientation

15
Entity Conversion

• Naturally one would like to know what each object
  looks like in the original virtual environment,
  so that the corresponding objects in the other
  environment resemble them
• Appearance includes such things like color and
  texture

16
Entity Conversion (cont.)

• ModSAF does not have 3D models for the entities
  in the simulations
• For UTSAF, we had to both acquire from public
  sources, and create from scratch, our 3D models
• 3D Studio Max
• Milkshape3D

17
Terrain Conversion

• Once you have the entities moved into the other
  simulator, you naturally want the static portions
  of the simulation as well
• Walls, floors, ground, sky
• Two approaches
• Convert what's in the original simulation
• Create new terrains from scratch, hand tune to
  look like original simulation

18
Terrain Conversion (cont.)

• We chose to convert the ModSAF terrains directly
  to the Unreal format
• Multi-step conversion pipeline
• ModSAF ? OpenFlight ? 3D Studio ? Unreal
• TerraVista DART
• NuGraf
• 3ds2unr
• Unreal levels are restricted in size, so ModSAF
  terrains are converted piecemeal

19
UTSAF Testbed
CAVE Theater
OneSAF Testbed DIS Network
RETSINA Agents and Server
Unreal Game Engine and Gamebots
20
RETSINA Agent Architecture

• Reusable Environment for Task-Structured
  Intelligent Networked Agents
• http//www.cs.cmu.edu/softagents
• Multi-agent system (MAS) that supports
  communities of heterogeneous agents
• Implements distributed infrastructural services
  that facilitate the interactions between agents,
  as opposed to managing them

21
RETSINA in Multi-Agent Environments
22
Why use Agents?

• Weve found agents to be effective glue for
  tying together disparate proprietary software
  that cannot be linked by conventional means
• Allows for loose coupling of different pieces of
  software, and sharing of information among
  software that normally cannot do so

23
Problems with Agents

• Communication lag
• Network lag
• Operating system latency
• Architecture complexity
• The more components in a system the more
  susceptible it is to failure
• Agent programming being more loosely coupled than
  other types, is more immune to this type of
  failure however

24
Unreal Tournament Agents

• The most important of the remaining agents are
  the ones that talk to UT
• Since this currently is a one way street, these
  are not referred to as brokers, but they will
  eventually become them
• These agents listen for the DIS information from
  the broker, and pass the updated entity
  information to UT

25
Gamebots Network Extension
Network
Gamebots Server
Gamebots Client
Port 1
Gamebots Module
Port 2
Gamebots Client
. . .
Port n
Unreal Tournament Server
Gamebots Client
Gamebots is an extension module for the Unreal
Tournament multiplayer video game, that allows
easy control of players in the game from external
software, using simple TCP/IP commands.
26
Gamebots Shot
27
UTSAF Architecture
28
SAF Broker Agent
29
Gamebots Agent
30
A BioChem Depot in ModSAF
31
A BioChem Depot in UTSAF
32
A Typical UTSAF Screenshot
33
Communication Lag

• There is a tremendous amount of packets that come
  out of ModSAF
• If the broker cannot keep up with the traffic, a
  gap in communication will develop between what
  ModSAF is sending, and what the broker is telling
  the UT agents

34
User Disorientation

• When ModSAF environments are large, people can
  easily get lost and disoriented
• Common problem Where is north?
• There is a need for a world-centered frame of
  reference of the entire virtual environment
• An overall map, for example

35
Future Work

• We are currently investigating the effects of
  network packet filtering
• Preliminary results suggest that filtering out
  even 80 of incoming packets has little
  noticeable effect on the simulation
• We are working to expand the size of ModSAF
  terrain represented
• Our major focus for UTSAF currently is to
  represent uninhabited aerial vehicles (UAVs) in
  realistic 3D