Using AI for Olympic Equestrian Event Preparation

Andy Hon Wai ChunAAAI 2008

Introduction

Organizing Olympic-scale events requires massive manpower, yet managing a volunteer workforce presents unique challenges compared to professional staff. For the 2008 "Good Luck Beijing" equestrian events in Hong Kong, organizers faced fragmented schedules, complex skill-matching requirements, and dynamic shift changes driven by weather and competition needs. With off-the-shelf software failing to meet these specific constraints, researchers at the City University of Hong Kong developed a high-speed, AI-driven Workforce Management System (WMS) from concept to deployment in just six weeks.

System Architecture and Methodology

The team employed a "non-intrusive" AI development approach, prioritizing rapid prototyping and modularity. The system was built using a Model-View-Controller (MVC) pattern on a custom, lightweight Python framework known as the "FOA Application Framework."

The system architecture utilizes an MVC pattern, integrating a custom Python framework with a dedicated AI engine for rostering.

To ensure interoperability with other Olympic systems, the team adopted the HR-XML standard. By using tools like generateDS, they automatically transformed XML schemas into Python objects, significantly accelerating the data-modeling phase.

AI-Driven Rostering

The core of the WMS is a rostering engine that treats scheduling as a Constraint-Satisfaction Problem (CSP). Constraints were categorized into three distinct levels to balance automation with human oversight:

  • Hard Constraints: Non-negotiable requirements such as maximum working hours and mandatory skills.
  • Overrideable Constraints: Rules that a human planner can bypass, such as cross-division assignments.
  • Soft Constraints: Preferences used for optimization, like minimizing travel time.
The AI engine evaluates various staff-related constraints, ranging from mandatory skills to personal preferences, to generate optimal rosters.

These constraints were coded in RDF/XML and compiled into Python, allowing for logic updates without altering the core application code.

Key Insights and Operational Success

The project demonstrated the power of rapid prototyping; a functional "click-through" prototype was delivered in just two days, building stakeholder confidence and allowing for early feedback on AI constraints. The use of custom template engines and Python decorators resulted in a reported 90% reduction in code size compared to traditional methods.

The system's agility was put to the test when a typhoon struck Hong Kong just before the event. The AI engine allowed planners to instantly re-roster the entire 1,000-person workforce to accommodate delays—a feat that would have been impossible using manual spreadsheets.

Results and Impact

The WMS was fully deployed within six weeks and successfully managed approximately 1,000 volunteers daily during the August 2007 preparatory games. The AI rostering algorithm generated complex assignments in a fraction of a second, replacing a fragmented manual process. Ultimately, the system enabled Hong Kong organizers to meet the International Olympic Committee’s (IOC) readiness requirements with high marks.

Limitations and Future Directions

While highly effective, the system currently maintains real-time attendance tracking through an external process. Future iterations could integrate these functions for better feedback loops. Additionally, while human planners currently handle "overrideable" constraints, future work could explore automated negotiation via SMS or email to further streamline the scheduling process.

Summary updated

BibTeX
@inproceedings{chun2008using,
  author = {Andy Hon Wai Chun},
  title = {{Using AI for Olympic Equestrian Event Preparation}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2008},
}
Figure