Full Name:  Brian      Nationality: Malaysian      City of Origin: Kuala Lumpur, Malaysia

Skills:

• Programming: C/C++, C#, Java, Lua, PHP, Python, ASP, ASP.NET, SQL
• Software design and architecture

Field/Industries:

• AI
• 3D Visualization and Game Programming
• Web application and portal development

Poster

Year 2008 - 2007

• Online portal development for TQMotor
• Baja: Edge of Control game development
• R&D project on optimization for Xbox 360 and Playstation 3
• Conference poster for Pacific Graphics 2008 at Tokyo, Japan

Year 2006 & earlier

• Consultant at G Element Private Ltd., Singapore
• Technical Lead for the SAP Asia Pacific Small & Medium Business (SMB) Marketing Portal project
• Technical Lead for the Quality Management Information System (QMIS) in-house web portal for Tan Tock Seng Hospital, Singapore.
• Freelance programmer for the Hong Kong MTR Corp Ltd’s LYNX Concrete Cube Test System that won the Merit Award in the Asia Pacific ICT Awards 2004 (APICTA 2004)

Thesis

Title:

Identifying and Overcoming Performance Bottlenecks for Modern Console Hardware

Description:

The new generation of video game consoles (Microsoft Xbox 360 and Sony Playstation 3) provides an unprecedented increase in processing power over its predecessors. However, this increase in computational power comes at the price of complex multi-core CPU that requires games to be optimized for parallel computation and multi-threading. Game subsystems that cannot take advantage of parallel computation effectively would then become a performance bottleneck.

The aim of the project is to investigate the general hardware architecture of modern video
game consoles (Microsoft Xbox 360 and Sony Playstation 3) in order to identify the possible performance bottlenecks that games running on these platforms are most likely to encounter. By identifying the game subsystem that would under-perform in the modern consoles, the performance issues particular to this subsystem can be resolved resulting in a major improvement for the overall performance of the game.

Due to the inherent nature of game Artificial Intelligence (AI), game AI code contains a lot of decision logic, which results in numerous execution branches, interactions and dependencies on other game subsystems. These attributes made it difficult for game AI to harness the additional power provided by the modern consoles. It has become a challenge for game developers to create game AI which can meet the rigorous demands of modern games without using computational power beyond the limit given to them.

Therefore, the focus of this research is to overcome the performance limitations of game AI in modern consoles. In this thesis, a novel load management technique is proposed to manage the updates of the game AI characters during runtime. Based on the elastic task
model, this technique allows games to control the total CPU utilization of the AI characters and limit the processing time consumed per frame to a specified limit. The elastic nature of the technique is also exploited to provide Level of Detail (LOD) effect for the AI characters. Empirical results from test simulations performed on Microsoft Xbox 360 hardware confirmed the benefits of the proposed load management technique. The processing time required for simulation of the game AI characters can be constrained to a specified threshold without sacrificing the quality of simulation perceived by the players. This results in a higher and more stable overall frame rate giving players a better game experience.

Conference:

Research results from the project have been accepted for conference poster proceedings in the Pacific Graphics 2008 conference at Tokyo University, Tokyo, Japan.

Title:

Effective Load Management Technique for AI Characters in Games

Abstract:

Creating large populations of AI characters for game environment is a major challenge because either insufficient CPU processing time is available or it is difficult to balance the computational needs of the game AI against the requirements of other game components. We present a novel technique that allows games to manage the functional updates of AI characters efficiently during runtime. The technique is an enhancement of the elastic task model and scheduling [BLCA02] which allows the total CPU utilization of the AI characters to be adjusted to the current workload of the game. We exploited the elastic nature of the technique to provide Level of Detail (LOD) effect for the AI characters. A prototype implementation on Microsoft Xbox 360 hardware is described in this paper.

Citation:

Brian Tan and Gabriyel Wong, Effective Load Management for AI Characters in Game. In Poster Proceedings of Pacific Graphics 2008 (Tokyo, Japan, 2008).