Jerry B. Weinberg

ITERATE: a conceptual clustering algorithm for data mining

The data exploration task can be divided into three interrelated subtasks: 1) feature selection, 2) discovery, and 3) interpretation. This paper describes an unsupervised discovery method with biases geared toward partitioning objects into clusters that improve interpretability. The algorithm ITERATE employs: 1) a data ordering scheme and 2) an iterative redistribution operator to produce maximally cohesive and distinct clusters. Cohesion or intraclass similarity is measured in terms of the match between individual objects and their assigned cluster prototype. Distinctness or interclass dissimilarity is measured by an average of the variance of the distribution match between clusters. The authors demonstrate that interpretability, from a problem-solving viewpoint, is addressed by the intraclass and interclass measures. Empirical results demonstrate the properties of the discovery algorithm and its applications to problem solving.

Robotics in education: Low-cost platforms for teaching integrated systems

Robotics has become a major educational tool from secondary school to undergraduate courses to graduate education. What’s more interesting is that the use of robots is not limited to traditional engineering departments but has also found its way into a variety of arts and science courses. The use of robotics by nonengineering, nontechnical instructors has been termed a “robotic revolution” [1]. The success of robots in education has come about as a result of a combination of factors. First, robotics provides a unique learning experience. Robots are a physical embodiment of computation. The students receive strong, visceral feedback from physically experiencing their work. There is a wide design space for students to explore, make hypotheses about how things work, and conduct experiments to validate their beliefs and assumptions. Early work by Seymour Papert, which led to the development of educational robot platforms, termed this style of learning “constructionism” [2]. Another important aspect of robotics is its systems nature. It provides an excellent design experience of an integrated system that includes mechanics, electronics, and computation. A second factor in the success of robots in education is the cost. The cost of computation has dropped exponentially over the last decade. The result has been the ability to design and market robot controllers at prices that are accessible to schools with even modest budgets [3]. Low-cost mechanical parts and electronic sensors have been marketed in combination. While these do not provide the same precision that would be needed in an industrial robot, they are sufficient for educational purposes. The final important factor is the plug-and-play feel of the new robot platforms. The multidisciplinary nature of robotics has previously relegated its study to larger research institutions that have had the range of prerequisite knowledge to engineer complex systems. Robot controllers, such as the Handy Board and the LEGO Mindstorm RCX, have mitigated this need by making it relatively simple to plug in motors and sensors and use well-known or simple programming environments. The development of specialized LEGO parts has made the construction of mechanical components accessible [4], [5]. In this first of two special issues, we present a group of articles that use available inexpensive platforms in secondary school courses, undergraduate school courses, and graduate study courses. These articles focus on the use of the Handy Board and the LEGO Mindstorm RCX plus LEGO construction pieces. The Handy Board, which was developed at MIT, was a predecessor of the LEGO RCX and was used in the innovative MIT 6.270 design course. The LEGO Mindstorm RCX was developed for the toy market but quickly found its way into educational settings. Many academics and hobbyists have been pushing the limits of the RCX through the development of new operating systems, programming environments, and sensors. The use of the RCX is going beyond behavior-based mechanisms to embrace control systems

Applying AI clustering to engineering tasks

An AI clustering system, Cobweb, has been applied to fault diagnosis, bridge design, and human gait analysis. These projects suggest lessons for those interested in engineering applications of clustering.<<ETX>>

Components, Curriculum, and Community: Robots and Robotics in Undergraduate AI Education

This editorial introduction presents an overview of the robotic resources available to AI educators and provides context for the articles in this special issue. We set the stage by addressing the trade-offs among a number of established and emerging hardware and software platforms, curricular topics, and robot contests used to motivate and teach undergraduate AI.

Multidisciplinary teamwork in a robotics course

Real-world systems are comprised of interdependent components creating integrated systems. These systems are developed by multidisciplinary teams. The goal of this project is the development of a comprehensive undergraduate course in robotics that encompasses various fields that are integral to robotic systems: Computer Science, Electrical and Computer Engineering, and Mechanical Engineering. A main pedagogical goal of the course is to teach group dynamics and the skills necessary for interaction with people in different disciplines in multidisciplinary teams. Descriptions of the course and the hands-on lab assignments are presented along with course assessment.

Participatory design in a human-computer interaction course: teaching ethnography methods to computer scientists

Empirical evidence shows the ability for computer technology to deliver on its promises of enhancing our quality of life relies on how well the application fits our understanding of how things work. Software designers need to apply methods that provide insights into the users mental model of the applications target task and to invite the user to be an active participant in the design process. This paper reports on our efforts to design an HCI curriculum around ethnographic techniques of data gathering and paper prototyping. Initial results are presented that study the courses effects on students attitudes regarding approaches to software design and their long term design behavior.

Fault-tolerant formations of mobile robots

The goal of a robot formation control architecture is to get a number of robots into a specified form. To be effective and practical, the control architecture must be able to transition a group of robots from an initial swarm to a final formation. It must then be able to handle real-world events that could disrupt the formation, thus, requiring formation repair, obstacle avoidance, and changes in the formation. In previous work, we presented a distributed, reactive cellular automata-based formation control architecture capable of controlling any number of robots in formation at once. In this paper, we examine our architecture with respect to necessary characteristics to handle real-world occurrences. To address issues of formation repair and obstacle avoidance, the control architecture is extended by a distributed auctioning method that allows the robot formation to reconfigure autonomously.

Breadth-last technical electives: integrating the CS core via computer games and mobile robotics

In this paper, we introduce the concept of breadth-last technical elective courses, which are designed to assist undergraduate CS students in integrating their entire core curriculum into a coherent whole at the end of their degree programs. Specific breadth-last courses in intelligent mobile robotics and computer game development have been implemented and are presented here to demonstrate the pedagogical concepts being discussed.

Continuing adventures in qualitative modeling—a qualitative heart model

This paper discusses the second generation of an ongoing project for developing a model of cardiac electrophysiology with sophisticated reasoning mechanisms and robust explanation capabilities. The framework for this model combines device- and process-oriented ontologies, and contains mechanisms that introduce delays into process definitions for explicitly modeling sequences of temporal activities. The scheme has been applied to develop a model of the electrical subsystem of the heart. The first generation model incorporated a “lumped” view of the cardiac electrical system, and simulated single cardiac action potentials for each component of the system. As a result the electrical behavior of an entire region was collapsed into a single action potential curve, and the model captured only temporal relationships among action potentials of gross anatomical structures. To provide a more accurate picture of cardiac functionality, the second generation model distributes electrical activity through the muscle and conduction units. This distribution allows for derivation of behaviors that are related to local disturbances (e.g., scar tissue) in individual units or regions. The corresponding simulation model scheme will incorporate a three-dimensional spatial representation of the cardiac electrical system and the musculature.

MDS: An Integrated Architecture for Associational and Model-Based Diagnosis

This paper discusses the design and implementation of an integrated diagnosis system, MDS (Multi-level Diagnosis System), which combines associational and model-based approaches to diagnosis. The design and implementation of the associational module is tailored to achieving efficiency in routine diagnostic problem solving, and to providing a desirable interface for the users. The model-based diagnosis module is developed to achieve completeness and consistency in the fault isolation task, and to avoid the brittleness that often occurs in associational systems. MDS addresses the important issue of combining the use of “deeper” knowledge in the form of a system model with “shallow” (or associational) knowledge, using a diagnostic controller to improve completeness and consistency without sacrificing efficiency. The diagnostic controller also employs a methodology for automated knowledge refinement by identifying incomplete and inconsistent rules and diagnostic tests in the associational module, and then by performing updates to correct the problems. This paper focuses on the design and implementation of the diagnostic controller.