Wing-Kwong Wong

Reciprocal tutoring using cognitive tools

Reciprocal tutoring, where peers take turns to tutor each other, is an interesting style of social learning. In the Reciprocal Tutoring System (RTS), three computational cognitive tools were designed to facilitate reciprocal tutoring of Lisp programming on the network. The first is a Petal-style of code-chunk interface, with which a tutee can enter Lisp code without making syntactic errors. The second tool is Diagnosis- Hint Tree, with which a tutor can diagnose and comment on the errors in the tutees program. The third one is a list of dialogue templates, with which the tutee and the tutor can communicate during the tutoring process. A three-phase experiment was conducted, with each phase using different cognitive tools. In addition, with the help of the cognitive tools, RTS provides a virtual learning companion that can play tutor or tutee. Evaluation results reveal both the strengths and weaknesses of peer-based learning and intelligent tutoring, with supports of different cognitive tools. Peer-based learning supported by cognitive tools is a practical and attractive alternative to intelligent tutoring systems. Exactly which type of tutor is preferred depends on the tutees cognitive, communication, and emotional needs in the tutorial context.

Reinforcement Learning of Robotic Motion with Genetic Programming, Simulated Annealing and Self-Organizing Map

Reinforcement learning, a sub-area of machine learning, is a method of actively exploring feasible tactics and exploiting already known reward experiences in order to acquire a near-optimal policy. The Q-table of all state-action pairs forms the basis of policy of taking optimal action at each state. But an enormous amount of learning time is required for building the Q-table of considerable size. Moreover, Q-learning can only be applied to problems with discrete state and action spaces. This study proposes a method of genetic programming with simulated annealing to acquire a fairly good program for an agent as a basis for further improvement that adapts to the constraints of an environment. We also propose an implementation of Q-learning to solve problems with continuous state and action spaces using Self-Organizing Map (SOM). An experiment was done by simulating a robotic task with the Player/Stage/Gazebo (PSG) simulator. Experimental results showed the proposed approaches were both effective and efficient.

A dynamic geometry environment for learning theorem proving

Many students of geometry theorem proving do not perform satisfactorily even when asked to prove simple theorems. Their problems might arise from their poor understanding of the involved geometric concepts, or from their ignorance of intermediate geometric propositions that are needed to derive the final conclusion. To address these issues, we propose a dynamic geometry environment for theorem proving that integrates the elements of exploration and theorem proving.

A Web-based Diagnosis System with Cognitive Knowledge for Problem Posing

Problem posing is considered an important training for students by mathematics educators. This paper proposes a web-based diagnosis system with cognitive knowledge for problem posing. The system is established with. Net web service architecture. The subject domain is restricted to Chinese geometry word problems. Students can pose a problem in natural language. Then the system diagnoses the linguistic knowledge, semantic knowledge, and schematic knowledge involved in the problem and provides the information for students to correct the posed problem, if needed. In this system, the cognitive knowledge needed for comprehending and diagnosing geometry word problems is constructed with an ontology-based tool called InfoMap. We conducted an empirical experiment to examine the systems ability to diagnose geometry word problems. Results show that the system can diagnose most errors correctly.

Using LC-3 Soft Core on an FPGA Development Board for Microprocessor Labs

In this study, a Little Computer 3 (LC-3) core is ported to Altera’s Cyclone III FPGA on the DE0 development board. DE0 input/output devices such as seven-segment display, LED, switch, GPIO devices, and Lego Mindstorms NXT devices are added as memory-mapped devices. The LC-3 core was designed with VHDL on Quartus II 9.0 Web Edition. The input/output devices on DE0 can be manipulated by LC-3 instructions. With these additional devices in LC-3, more diversified and interesting labs can be designed. Students can first test their programs on the simulator and then on DE0. This should add more fun to learning assembly language programming and computer organization and flatten the learning curve.

Investigating readers' mental maps of references in an online system

Referential identification and resolution are considered the keys to help readers grasp the main idea of a text and solve lexical ambiguities. The goal of this study is to design a computer system for helping college students who learn English as a Foreign Language (EFL) develop mental maps of referential identification and resolution in reading. Four modules, Natural Language Processing (NLP), User Interface, Recording, and Feedback Tool, are implemented in the system. Results of this study showed that the more-proficient EFL readers were able to identify and resolve most of the references to form a coherent mental map from different parts of a text. The less-proficient readers commonly resolved references by relying on grammatical rules instead of semantic contextual clues. They often referred references to incorrect objects. To overcome the difficulties in figuring out the relationship between two words, the less-proficient readers usually asked for more feedbacks. As students progressed in reading, they requested fewer feedbacks in the online system. Some recommendations for future studies are discussed.

A Study of High School Students Doing Physics Experiments with Arduino and Other Data-Logging Devices

Conventional high school physics labs rely on manual data logging or logging devices with a slow data acquisition rate and low precision. This study aims to design data-logging devices and a modelling tool for high school physics labs with low-cost modern electronics, including smartphones, Lego Mindstorms NXT and Arduino, equipped with an ultrasonic sensor. For NXT and smartphones, experimental data were first logged in the devices and then manually copied to a personal computer for data analysis. For Arduino, experimental data were transmitted to a PC via Bluetooth in real time. With the data in a PC, each student used a modelling tool on a Web browser to try to find an equation that fitted the data with a small error. The equation was a function that related one variable to another. For example, in a free-fall experiment, the equation expressed distance as a function of time. With each submitted equation, the tool plots the model against a background of lab data with a measure of error. Based on the visual plot and the error information, the student can then try to reduce the error by revising the equation. The results indicated that both students and the instructor enjoyed using the modern data loggers and the acquired data to find equations that fitted the data well.

A Matrix Robot for Block Stacking

This research studies how to control a robot to stack blocks according to the block structure shown in a given image. In this project, a LabVIEW program is designed and run on a Lego Mindstorms NXT, which in turn controls a robot built with components of Matrix Robotics. There are two operation modes for the block stacking robot. In the first mode, a user uses an Android smartphone to control the robot with Bluetooth to move around, pick up blocks, and stack them. In the second mode, the robot identifies the block structure in a given image with OpenCV and picks up the needed blocks to stack them to build the block structure. The robot was tested with two images of simple block structures in the second mode. The success rates ranged from 7 to 80 %. Future work should further identify the weaknesses of the robot and improve its performance.

Using an Android Smartphone for Robotic Image Recognition

In this study, an Android smartphone was used to control a robot of Lego NXT running leJOS by automatically tracking and pushing a ball forward. OpenCV was run on the smartphone for ball recognition. In addition, HCT (Hough Circle Transform) was used to recognize circular objects of particular colors in the environment. Furthermore, the smartphone controlled the robot with Bluetooth wireless communication.

A virtual computational paper folding environment based on computer algebraic system

Many people enjoy origami, an art of paper folding, since childhood. Origami is a more powerful geometry construction tool than straight and compass. But there are some inconvenience when you practice traditional origami on geometry. In this study, a computational origami environment has been developed. Huzita axioms are implemented with a computer algebra system (CAS). CAS not only deals with fundamental computation of axioms but also can prove some geometric consequences of folding steps. Furthermore, the process of paper folding is visualized. Users can observe the 3D animation of folding steps from different viewpoints.