In 2007, the KCRC merged with MTR to oversee all the railway lines in Hong Kong. After the merger, MTR operates 10 heavy rail lines and 12 light rail lines, spanning 218.2km and 152 stations. In 2010, MTR decided to create one AI system to schedule and manage all engineering works for all the rail lines. This new system is an enhanced and updated version of the AI Engine deployed in 2004 for the subway system.
Mass Transit Railway Corporation (MTRC), manages all the railway lines in Hong Kong. In order to maintain service quality and performance, and to support various improvement and rail extension / construction projects, each week over 2,600 individual engineering works need to be performed, involving over 10,000 maintenance personnel.
MTR carries 5.2 million passengers on an average weekday, making it one of the busiest rail system in the world. Despite the large traffic-flow, the MTR has set for itself a very high service quality standard, consistenly delivering a 99.9% on-time rate. To ensure these standards are met, the scheduling of engineering works must be optimized so that all necessary maintenance tasks are done in a timely manner.
MTR is using AI to plan, schedule, and optimize their various engineering works at rail lines in Hong Kong. The AI Engine allows MTR to streamline the scheduling / rescheduling porcesses as well as optimize resource allocation to maximize the number of engineering works performed. AI is delivered as a unique cloud-based "AI as a Service" (AaaS) model, where AI services are provided via client-agnostic and rail-line agnostic REST APIs. The AI services include rule-based conflict-checking, scheduling, rescheduling, and resource optimization. The AaaS model allows AI to be easily deployed to server other rail lines through straightforward REST service calls.
The new enhanced AI Engine has been in daily use to schedule and manage all engineering works for all the rail lines in Hong Kong since July 2013. All engineering works scheduled are ensured to be safe and satisfying all business and operational constraints. The schedule is also optimized so that MTR can perform more engineering works with its existing resources. In addition, valuable domain knowledge and expertise related to these regulations and guidelines are now quantified, coded and preserved within the organization, for use by anyone using the system.
Possession/Engineering Work Scheduling (2004)
Airport Bay/Stand Allocation
Airport Check-in Counter/Desk Allocation
Air Cargo Handling System
Container Terminal Berth Allocation