Ray H. Kemp

Using planning techniques to provide feedback in interactive learning environments

A scheme for domain and task representation in an interactive learning environment is described. The approach, based on recent enhancements to a classic AI planning formulation, enables suitable feedback to be incorporated economically into tutors for a range of procedural domains. The method is applied to the problem of simulating the operation of a video cassette recorder and representing tasks within this domain.<<ETX>>

A genetic-algorithm-based neural network approach for short-term traffic flow forecasting

In this paper, a Genetic-Algorithm-based Artificial Neural Network (GAANN) model for short-term traffic flow forecasting is proposed. GAANN can integrate capabilities of approximation of Artificial Neural Networks (ANN) and of global optimization of Genetic Algorithms (GA) so that the hybrid model can enhance capability of generalization and prediction accuracy, theoretically. With this model, both the number of hidden nodes and connection weights matrix in ANN are optimized using genetic operation. The real data sets are applied to the introduced method and the results are discussed and compared with the traditional Back Propagation (BP) neural network, showing the feasibility and validity of the proposed approach.

Towards a realistic microscopic traffic simulation at an unsignalised intersection

In this paper we propose a microscopic traffic flow model to simulate the flow at an unsignalised intersection. The model is built on fine grid cellular automata (CA), and is able to simulate actual traffic flow. Several important novel features are employed in our model. Firstly, the average car-following headway (=distance /velocity) 1.5 seconds has been observed in local urban networks and this 1.5-second rule is built to our model. Secondly, vehicle movement on urban streets is simulated, based on the assumption of velocity following a Gaussian (normal) distribution and is calibrated by field data. Thirdly, driver behaviour is modelled, using a truncated Gaussian distribution. Finally, the limited priority mechanism is involved in this paper. The model has been validated against real data for its several components.

From electronic textbook to multidimensional learning environment: overcoming the loneliness of the distance learner

Traditional textbook-centred distance education has been criticised for providing a one dimensional and isolated learning experience for the remote student. Computer-based distance learning is promoted as a more interactive and collaborative alternative. However, web-based systems have yet to realise their potential for overcoming the isolation of the student, in part because of their heavy reliance on the text book approach. This paper reviews existing computer-based learning systems from the perspective of the student user, and proposes some guidelines for improving their effectiveness as collaborative and interactive distance learning environments.

Domain and task representation for tutorial process models

Abstract The teaching of procedural skills is a fruitful area of intelligent tutoring system research. Students often learn such skills most effectively by computer when a guided discovery mode is employed. So far, no general aids to developing associated software have emerged. A domain and task representation scheme is proposed that facilitates a cognitive modelling approach to the development of discovery learning systems. It is based on a classical formalism used in AI planning and utilizes recent modifications to the technique to allow hierarchical feedback to be included, efficiency to be improved, the important distinction between domain and task to be made, and task representation to be simplified.

Learning by Creating: Letting the Student Do the Work.

There is evidence to suggest that mature students, perhaps studying at home where human tutors are not readily available, learn particularly effectively in a problem-solving environment. Also, they appear to be especially responsive to analogies and metaphors. We seek to exploit these effects by requiring students to transform problems into alternative forms before solving them. Three examples are given of domains where the student has to convert the formulation into an equivalent diagram. Arguably, the process of performing this transformation increases their understanding of the subject area.

On Developing Case-Based Tutorial Systems with Conceptual Graphs

In this paper, a case-based intelligent tutoring system (CBITS) using conceptual graphs to represent the cases is described. Among others, two core issues to be addressed by CBITS developers are indexing and how learning activities can be constructed from the cases. Addressing the former, the authors discovered that the minimum common generalization of a set of graphs is a powerful means of embedding different tutorial primitives into the cases. This approach provides an extremely rich indexing vocabulary for the cases and relieves the developers from speculative assignment of indexes. For constructing learning activities an operational semantics for the case graphs is defined and with this semantics the instructor can reason about the graph structure and provide intelligent guidance to the students in exploring the case contents. Newtonian mechanics is the testing domain of the proposal but its underlying methodology should be equally applicable to other subject domains.

Computerised problem-based scenarios in practice-a decade of DIAGNOSIS

The paper discusses aspects of problem-based scenarios (PBS) for teaching based on experiences with a computer package called DIAGNOSIS. This package has been used for teaching the diagnosis of plant pests and disease problems for a decade. The software is described, and its evolution in response to changing demands and software platforms is also covered.

An expert system for diagnosis of pests, diseases, and disorders in apple crops

Abstract A prototype expert system is being developed to diagnose pests, diseases, and disorders in apple trees. The system has the dual aims of supplying diagnostic advice to apple growers and teaching students to diagnose problems effectively. Notable features of the system include: clear separation between heuristic knowledge and domain/real world facts, a flexible certainty factor system, and a user-friendly environment.

MODELING DRIVER BEHAVIOR ON URBAN STREETS

Traffic flow on straight roads is the most common traffic phenomenon in urban road traffic networks. In this paper, a realistic cellular automaton (CA) model is proposed to investigate driver behavior on urban straight roads based on our field observations. Two types of driver behavior, free and car-following, are simulated. Free driving behavior is modeled by a novel five-stage speeding model (two acceleration stages, one steady stage and two deceleration stages). Car-following processes are simulated by using 1.5-s as the average headway (1.5-s rule), which is observed in local urban networks. Vehicular mechanical restrictions (acceleration and deceleration capabilities) are appropriately reflected by a five-stage speeding model, which has the dual-regimes of acceleration and deceleration. A fine grid (the length of each cell corresponds to 1 m) is used. Our simulation results demonstrate that the introduction of the dual-regimes of acceleration and deceleration, 1.5-s rule and fine grid matches actual driver behavior well on urban straight roads.