This paper describes how AI is used to plan, schedule, and optimise nightly engineering works for both the commuter and rapid transit lines in Hoag Kong. The MTR Corporation Limited operates and manages all the rail lines in Hong Kong. Its Engineering Works and Traffic Information Management System (ETMS) is a mission critical systemthat manages all information related to engineering works and their related track possessions and engineering train movements. The Al Engine described in this paper is a component of this ETMS. In Hong Kong, the maintenance, inspection, repair, or installation works along the rail lines are done during the very short non-traffic hours (NTH) of roughly 4 to 5 hours each night. These engineering works can be along the running tracks, track-side, tunnel, freightyards, sub-depots, depot maintenance tracks, etc. The proper scheduling of necessary engineering works is crucial to maintaining a reliable and safe train service during normal hours. The Al Engine optimizes resource allocation to maximize the number of engineering works that can be performed, while ensuring all safety, environment, and operational rules and constraints are met. The work described is part of a project to redesign and replace the existing ETMS, deployed in 2004 with an updated technology platform and modern IT architecture, to provide a more robust and scalable system that potentially can be deployed to other cities around the world.

This paper outlines the design of a scheduling algorithm that allocates outstabling locations to railway trains. From time to time railway trains may need to be outstabled to temporary locations, such as stations, sidings, depots, etc., until they are needed for regular operations. This is common for urban rail transit, and especially so for those that do not operate 24 hours. During non-traffic hours (NTH), trains are outstabled to various locations along the rail network so that when operations start again next day, the trains will be nearby their originating station or conveniently located so that they can be put into service whenever needed. However, this is complicated by the fact that engineering works, such as rail testing, installation, regular maintenance, etc. are done during the NTH. Therefore, passenger trains must be outstabled in such a way that they do not interfere with night-time engineering works or the movements of associated engineering trains. Since the engineering works scheduling is done separate to outstabling, this is a mixed-system problem. This paper shows how we modeled this as a constraint-satisfaction problem (CSP) and implemented into an “Outstabling System” (OSS) for the Hong Kong Mass Transit Railway (MTR) using a two-stage search algorithm

The ability to maximize service quality while minimizing cost is very important to service-oriented businesses, such as lorry leasing. Very often, the ability to consistently offer higher quality service is the main differentiating factor between a business and its competitors. For lorry leasing businesses, service quality and cost are directly related to how resources-vehicles, cranes, and drivers-are allocated to jobs. The ability to assign the right combination of resources is crucial to daily operations. This paper presents how we modeled this assignment problem as a constraint-satisfaction problem (CSP) and implemented using constraint programming (CP) with an algorithm that we call the crane-lorry optimizing engine (CLOE). CLOE was implemented for the largest crane-lorry company in Hong Kong. Furthermore, plans are generated within seconds compared to close to an hour if done manually. All necessary constraints and criteria are considered systematically. We have experimented with many different types of search heuristics and have analyzed their effects on plan quality. We have found that by considering both the experience of the crane-lorry-driver combination and maximizing on the order assignment, we were able to generate plans that were significantly better than those produced by human planners, and within a substantially shorter time. Although the use of constraint-based assignment techniques is still limited in the vehicle leasing industry, we hope our combination of automated assignment with Internet portal technology that streamlines business-to-business, business-to-consumer, and business-to-employee communications can excite interest in this area

Systematic tree search augmented with constraint propagation has been regarded as the de facto standard approach to solve constraint satisfaction problems (CSPs). The property of completeness of tree search is superior to incomplete stochastic local search, although local search approach is more efficient in general. Many heuristics techniques have been developed to improve the efficiency of the tree search approach. We propose a framework for combining and coordinating a complete tree search solver and a different solver in order to produce a complete and efficient CSP solver. Three different instances of the framework have been suggested including combining complete tree search with stochastic search, mathematical programming approach respectively. The experimental results show that this highly integrated hybrid scheme greatly improve the efficiency of constraint solving process in terms of both computation time and number of backtracking.

The objective of resource allocation systems is to assign valuable corporate resources to meet business demands. However, in applications involving stochastic events, demands cannot be determined explicitly beforehand. For those applications, computer simulation is often used to predict resource demands. The paper describes research in developing a knowledge based simulation system to predict resource requirements at an international airport. Our intelligent resource simulation system (IRSS) determines how many check-in counters should be allocated to each departure flight while providing passengers with sufficient quality of service. This predicted resource requirement is then used by a constraint-based resource allocation system to allocate the actual check-in counters. Because IRSS considers many more factors than a human can, the resulting allocation schedule is more efficient. These factors include: 1) different service rates for different destinations, airlines, or handling agents; 2) different passenger arrival rates for different times of the day or days of the week; and 3) different requirements for different service levels, etc. Our experiments show that there are substantial resource savings by combining a resource allocation system with an intelligent resource simulation system.

In order to manage the daily operation of a bus company, different types of scheduling must be performed on a regular basis: bus timetabling, duty assignment, bus assignment, rostering, driver assignment, etc. Due to problem complexity and limitations of available computer technologies, these scheduling activities are often performed by separate software systems. Even within a single type of scheduling, the schedule for each route is generated independently of each other. This mode of isolated scheduling leads to inefficient schedules since interrelated constraints are not considered and resources cannot be shared among routes. This paper documents research that explores how a distributed constraint-based scheduling architecture might shed some light on this integration problem. Our current software prototype integrates the scheduling tasks of bus timetable generation with duty assignment and permits bus and bus driver resources to be shared among routes. This approach guarantees that the resulting timetable satisfies both service requirements as well as bus driver labour constraints.

Object-oriented constraint programming (OOCP) combines two orthogonal but complementary programming paradigms into one-object-oriented programming (OOP) and logic constraint programming (LCP). Most practitioners of OOCP would agree that the design of a constraint-based model is the most time consuming and difficult part of OOCP. There is currently no systematic approach or methodology to follow when designing a model suitable for OOCP. The paper outlines a simple methodology that extends traditional object-oriented analysis and design (OOAD) to cover constrained variables and constraints used in OOCP. The proposed OOCP Methodology is an initial investigation into providing a more rigid set of guidelines to follow when designing constraint-based models. The paper presents the OOCP methodology using a simplified nurse rostering system as a case study.

The paper describes an intelligent computer aided architectural design system (ICAAD) called ICADS. ICADS encapsulates different types of design knowledge into independent "critic" modules. Each "critic" module possesses expertise in evaluating an architect's work in different areas of architectural design and can offer expert advice when needed. This research focuses on the representation of spatial information encoded in architectural floor plans and the representation of expert design knowledge. Described in the paper is our research in designing and developing two particular "critic" modules. The first module, FPDX, checks a residential apartment floor plan, verifies that the plan meets a set of government regulations, and offers suggestions for floor plan changes if regulations are not met. The second module, IDX, analyzes room and furniture layout according to a set of interior design guidelines and offers ideas on how furniture should be moved if the placement does not follow good design principles.

This paper describes a massively parallel system for word recognition. Based on the connectionist network model adopted from cognitive science and artificial intelligence, the system consists of a large number of simple neuron-like processing units, or nodes, which represent words, phonetic segments, or phonetic features. The computation consists of constant updating of activation levels of all nodes, resulting from the excitatory links and inhibitory links between the nodes. Input to the system consists of frame-by-frame scores of similarity to a set of pre-defined spectral filters, which represents the set of phonetic segments necessary for distinguishing between words in the vocabulary. These similarity scores are combined into phonetic feature indexes for each frame of speech as input to the feature nodes in the network. A linguistic knowledge base is built into the network, allowing both data-driven processing and top-down prediction to cooperate or compete in working toward the correct lexical hypothesis.