Readings

Read by Damian Johnson on March 4th, 2009

Conventionally game environments are crafted by developers by blending rendered 3D models and scripted events to provide the desired scenes. Though there's been a few commercial applications of auto-generated environments (for instance dungeons in Diablo II) this has mostly stayed in the 2D realm to cut down on complexity. This paper discusses the use of situated rational agents (AIs with a spacial sense) to craft 2D and 3D environments using a predefined design grammar (viable geometric transformations like rotation and mirroring) and descriptions of function (tags for their attributes, purposes, and associated scripts) to automatically structure scenes according to given constraints and objectives.

The port simulation is at heart is an analysis and optimization problem. Concepts proposed by the paper have little to do with analysis (since we're reflecting an established sea port, not making one of our own). However, for optimization this has the added benefit in that if the reasoning was provided to a player agent (ie, someone that needs to obey physical laws like gravity and collisions) then we can draw added information from the process of optimizing the scene. For instance, if we are trying to rearrange cargo to optimize sensor coverage then a simple optimization algorithm might make several decisions that, though technically correct, are completely unhelpful as a real world solution. For instance its tactics might include...
... placing all cargo on a distant mountain.
... placing all cargo under the ground or suspended high above the scene.
... stacking all cargo into a single, high tower.

Though we could take an iterative approach where we script in rules disallowing these sorts of solutions, forcing the AI to use a player's body to transport the cargo would eliminate most of these problems. For instance players can't climb mountains, go underground, or fly up to place the cargo in impossible locations. In addition, since the player agent is constrained to specific movement speeds and activities, we could log what it does to optimize the arrangement so it can be used later to exemplify how the rearrangement can be done in the real world and to provide an estimate for its viability. With a weighting of real-time and price to game time activities estimating the time and cost of such changes would be trivial. In addition physical constraints force the agent to take realistic approaches to their ends. For instance swapping two pieces of cargo would require that the agent have a temporary location for the exchange which might radically change a solution's viability.

Read by Allen Christiansen on February 11th, 2009

This paper they presented a simulation prototype of the customers' shopping behavior in a mall using a knowledge-based multiagent geosimulation approach. The shopping behavior in a shopping mall is performed in a geographic environment (a shopping mall) and is influenced by several shopper's characteristics (internal factors) and factors which are related to the shopping mall (external or situational factors). After identifying these factors from a large literature review they grouped them in "dimensions". Then they used these dimensions to design the knowledge-based agents' models for the shopping behavior simulation. These models are created from empirical data and implemented in the MAGS geosimulation platform. The empirical data have been collected from questionnaires in the Square One shopping mall in Toronto (Canada). After presenting the main characteristics of our prototype, we discuss how mall’s managers of the Square One can use the Mall_MAGS prototype to make decisions about the mall spatial configuration by comparing different simulation scenarios. The simulation results are presented to mall’s managers through a user-friendly tool that they developped to carry out data analysis.

Read by Damian Johnson on December 1st, 2008

The state of the art in RTS (real time strategy) game AIs are at heart simple reflex agents. Even most avenues of research are toward pre-learned or map based tactics due to limited resources and game complexity. This paper provides a different approach using a combination of case-based and reinforcement learning to allow AIs to grow from experience and translate those experiences to different environments.

The proposed design delegated the scope of decision making to have a multi-tiered approach where each layer's output are the goals for the next with temporal progression (ie, lower layers run more frequently). This allows for occasional strategic thought at the top and fast tactical decision making at the lowest levels. While this could easily be applicable to infiltrators for our port scenario (ie, combatants trained in urban environments trying to take control of the port) it offers limited usefulness to cargo inspection. The issue is scope - sneaking a given piece of cargo past static sensors is a matter of finding gaps, not high level strategic planning. However, the port simulation is simply a component in a larger scheme which does include strategic planning which could likely benefit from transfer learning.

Read by Allen Christiansen on November 3rd, 2008

In this paper a simulation tool is proposed to assess the level of underwater security in civilian harbour installations. The tool is based on a geographical information system, and it allows to improve port security verifying the impact of both new sensors and modification in the sensors placement in the overall harbour protection system. The tool works under the Microsoft Windows Operating System and it is based on the ESRI MapObjects 2.3 library. It permits to load a geographical area with any kind of environmental information, add several type of underwater sensors and several type of intruders and verify what security level is guaranteed by the sensors configuration chosen against the particular intruder selected. Moreover the simulation tool is strictly connected with the Matlab environment to improve the tool flexibility when a new sensor or a new kind of intruder is added. Finally, a genetic algorithm, built in the Matlab environment, allows to find the optimal sensors configuration to obtain a desired security level in the area at the lowest cost..

Read by Damian Johnson on October 6rd, 2008

Paper discusses the advantages of using game engines to train first responders in the case of radiological emergencies. Specifically, this is concerned with the use of separate simulations to provide realistic input to the game engine for things like radiation propagation. Much of the paper was at a fairly high level and unfortunately I wasn't able to find further information concerning their work.

Read by Allen Christiansen on April 7rd, 2008

Around the world, there is great concern about the movement of threat materials using seaport shipping containers. The benefit of early detection of weapons of mass destruction is obvious. However, the inspection process needs to be conducted in such a way as to not unreasonably impede normal commerce. Prior to actual deployment of new detection systems, policies, or procedures, it is useful to construct an operational and cost model of the port facility and to run simulations to gage the impact. Using a simulation model beforehand aids decision makers in evaluating trade-offs. PortSim was developed at ORNL by the author to allow a user to investigate a number of parameters in order to see the impact on port operations and cost. It consolidates a conceptual operations model, cost information, policy and procedures database, a real-time data acquisition capability, and information flow tracking and provides a visualization of port operations in a geospatial environment. This paper describes the use of PortSim to simulate and visualize a typical port.

Read by Damian Johnson on March 3rd, 2008

This paper concerns the technical details of L3DGEWorld 2.3, a data visualization tool utilizing OpenArena (a derivative of the Quake III Arena game engine). The tool consists of four components allowing a measure of flexibility in how its applied. These include an input daemon, the L3DGEWorld client & server, and an output daemon for translating in-game actions to the appropriate commands. The paper presents a sample scenario where network administrators detected and blocked a port scan on a greynet via the mechanics provided in the FPS engine. In this case pyramids represented IP ranges, spinning indicated network activity, and being shot caused a block. The tool operates by metaphors which are occasionally somewhat labored. Still, this is an interesting and novel approach to an alternative means of administration.

Read by Allen Christiansen on February 4th, 2008

In this paper they are developing interactive simulations of the National Institute of Standards and Technology (NIST) Reference Test Facility for Autonomous Mobile Robots (Urban Search and Rescue). The NIST USAR Test Facility is a standardized disaster environment consisting of three scenarios of progressive difficulty: Yellow, Orange, and Red arenas. The USAR task focuses on robot behaviors, and physical interaction with standardized but disorderly rubble filled environments. The simulation will be used to test and evaluate designs for teleoperation interfaces and robot sensing and cooperation that will subsequently be incorporated into experimental robots. This paper describes our novel simulation approach using an inexpensive game engine to rapidly construct a visually and dynamically accurate simulation for both individual robots and robot teams.

Read by Damian Johnson on December 3rd, 2007

This looked at the trade-offs between various stakeholder concerns (such as security and throughput) using the response surface method (RSM) to provide a set of Pareto optimal solutions. Analysis was mostly concerned with a simulated supply chain under a variety of loads. This was of great interest since it illustrated concerns our simulation should take into account (interests of port authorities and the need to investigate varying loads) as well as illustrate an off-line means of balancing trade-offs at a high level (ie, input to their mathematical analysis being generic attributes rather than the use of actual physical layouts as our project does).

Read by Allen Christiansen on November 5th, 2007

In this paper they present a new approach for simulating real-time sound propagation in complex, virtual scenes with dynamic sources and objects. Their approach combines the efficiency of interactive ray tracing with the accuracy of tracing a volumetric representation. They use a four sided convex frustum and perform clipping and intersection tests using ray packet tracing. A simple and efficient formulation is used to compute secondary frusta and perform hierarchical traversal. In the paper, they suggest that their algorithm can simulate and render sounds at interactive rates on a high-end PC. More information can be found here.