Intelligent Systems Group


Introduction

Simon Fraser University, located 15 km east of Vancouver, enjoys a spectacular view of the Lower Mainland from Burnaby mountain. The School of Engineering Science, a unit of the Faculty of Applied Science, has 22 full-time faculty and 8 instructors. The school offers the degrees of Bachelor of Applied Science, Master of Applied Science, Master of Engineering, and PhD.

This brochure describes current research in Intelligent Systems by faculty and students in the School of Engineering Science. Details of the faculty research and preprints of publications can be obtained by calling or writing the faculty listed below. Additional information concerning entrance requirements and financial assistance can be obtained from the departmental office. Information about the academic programs at the graduate level can be obtained by contacting the Graduate Program Chairman, the Chairman of the Intelligent Systems Group, or the faculty directly.

Contacts

Director, School of Engineering Science
Dr. John Jones, 604-291-4679, jones@sfu.cs.ca
Chairman, Graduate Program
Dr. Mehrdad Saif, 604-291-3119, saif@cs.sfu.ca
Chairman, Intelligent Systems Group
Dr. John Dill, 604-291-3574, dill@cs.sfu.ca
Dr. William A. Gruver
604-291-4339, gruver@cs.sfu.ca
Dr. Kamal K. Gupta
604-291-3118, kamal@cs.sfu.ca
Dr. Paul Ho
604-291-3822, ho@cs.sfu.ca
Dr. Albert Leung
604-291-4296, leung@cs.sfu.ca
Dr. Shahram Payandeh
604-291-4290, shahram@cs.sfu.ca
Dr. Andrew Rawicz
604-291-3819, andrew@schmitt.ensc.sfu.ca

Current Research Projects in Intelligent Systems

Artificial Eye Research The ultimate goal of this work is fabrication of a functioning eye prosthesis, an electronic equivalent of the vertebrate eye. Results of near-term efforts may help with robotic vision problems with image processing and may help explain certain human vision defects. This is a cooperative effort with scientists from physics, psychology, ophthalmology, and computing science.

Automatic Control Systems

This research is developing systems for automated diagnosis of faults. Diagnostic reasoning requires search over a large space of possible diagnoses. Bayesian nets can be used to order this search. A prototype system for diagnosis of automobile engines has been implemented in collaboration with McCarney Technologies. Research also is focused on large scale and decentralized control, optimal control, and fault tolerant control system design. A variety of questions dealing with the state/output feedback design of optimal controllers with pole placement and weight selection capabilities is being investigated. A second area is that of fault-tolerant control system design for applications where continuous and safe operation of the system under control is important for safety or economic reasons. Other work includes the study of robust and insensitive controllers, and the development of a hydraulic pressure controller for compliance and contact force control in underwater robotics applications in collaboration with ISE, Ltd.

Autonomous Underwater Vehicles

An autonomous underwater vehicle (AUV) is being developed as a sensor platform for experiments in signal processing and ocean exploration. Software and hardware for autonomous control of a tethered remotely operated vehicle has been completed and transferred to the AUV platform . The system is being tested in our large in-house test tank and then at a covered barge for initial sea trials. Students also have access to a 40 foot research vessel for coastal work. Research in this area is done in collaboration with the local subsea industry which leads the world in many subsea technologies.

Design Visualization

Existing CAD systems are effective only at the lowest levels of design. The next generation of systems must, by contrast, provide help with design synthesis and analysis. To achieve this, several lines of research are being pursued, including the logical and psychological nature of the design task, and the behaviour of expert and novice designers. Prototype knowledge bases to support mechanical and architectural design are being developed, using an expert system shell with constraint propagation capabilities. Critical attention is being given to the issues of user interface management and visualization techniques.

Industrial Test and Inspection Systems

New transducers and instrumentation are being developed for special applications such as a diagnostic system for rough ride testing of heavy trucks. The diagnostic inference is made based on the analysis of vibration spectra of a truck interpreted by a trained neural network (lower level inference) and a causal-heuristic expert system (higher inference level). Other examples include a microprocessor controlled capacitive transducer for automated fish inventory. Research is being conducted on the implementation of systems and algorithmic techniques for industrial inspection, defect recognition, and the integration of data from multiple sensors. Recent applications have utilized techniques based on morphology, neural networks, and fuzzy logic. Sensor development is being conducted in collaboration with researchers from the SFU Micromachining Facility.

Intelligent Design

Research in this area explores the use of model-based reasoning for design and diagnose faults. In collaboration with McCarney Technologies, we are developing a diagnostic system for automobile engines. This research integrates engineering knowledge of the artifact with constraint-based logic programming. Search through the space of possible diagnoses is guided by heuristic rules. In particular, we are evaluating Bayesian probabilistic reasoning as a way of ordering the search. In both design and diagnosis we take a mixed-initiative approach; the responsibility for the task is shared between the system and a human engineer. For this approach to be successful, the system must be able to explain its decisions to the user. This involves study of automatic explanation and the development of a powerful graphic user interface.

Intelligent Graphic Interface

The thrust of this joint university-industry project, is to bring intelligence and advanced visualization methods to help inprove graphic user interfaces for very large real time systems, such as power transmission and telecommunications systems. Part of this work has been in visualization methods for representing large, hierarchically organizednetworks. The major industrial partner is MPR Teltech, though several other organizations including HP Canada, BC Tel, TransAlta Utilities, Alberta Research Council, DEC Canada, Shell Canada, and HA Simons are involved.

Manufacturing Automation

Research in this area treats automation of discrete parts assembly with emphasis on flexibility through sensor-based handling of parts and the development of mechanisms that facilitate mechanical and electronic assembly. Methods are being developed for industrial inspection, defect recogntion, and recognition of parts being handled. This research has received collaboration and support from Weiser Inc. for the manufacture of locksets. Another direction of research is the development of software environments for programming manufacturing systems. Building on textual languages developed for industrial robots and numerically controlled machine tools, manufacturing programming environments attempt to provide a common interface to industrial devices and sensors that would otherwise require specialized programming by different people with different levels of skills. To fabricate micromechanisms from biocompatible materials for medical uses, a Nd:YAG laser with a fine focussing optical system has been coupled with two precision linear motors. The motors are computer controlled providing x-y motion and the laser duty cycle is controlled using an acousto-optical modulator. Reliability research is also being conducted to search for correlations between easily and nondestructively measurable parameters of electronic devices and their lifetime. Correlation models are built on the knowledge of microscopic, physical properties of materials and analysis of failures, and this models are used to determine reliability measures of the product through fast measurement techniques.

Neural Networks and Fuzzy Logic

Work in this area is investigating the properties of non-linear adaptive systems, and their potential in solving pattern recognition and control problems. Advanced computer graphics visualization techniques are being applied to study how such networks learn.

Robot Path Planning and Control

Research focuses on various computational and geometric issues in task level robot planning, in particular robot motion planning efficient algorithms for planning collision-free motions for manipulator arms with many degrees of freedom, planning robot motions among moving obstacles, and dealing with uncertainty in robot motions. A second area of interest is that of building 3-dimensional descriptions of a robot's environment for motion planning, primarily using range and intensity data. With the collaboration of ISE, Ltd., the algorithms designed at SFU are being implemented into a practical motion planner for a space robot having 7 degrees of freedom.

Robotics for Personal Service

Research in this area deals with mechanical design, intelligent control, and integration of devices for manipulation and mobility. These devices include robotic hands and end-effectors, walking machines, and mobile vehicles. A three-fingered, tendon operated, 9 dof dextrous robotic hand is being developed for use in experimental validation of task oriented grasping. A limited dexterity, 3-fingered hand is being developed for use as a prosthesis and as a flexible end-effector for service robots. Other devices being developed include a four-fingered hand prosthesis, an electro-myographically controlled below-elbow prosthesis, and a powered walking aid to assist paraplegics. This research is being jointly conducted with Tsinghua University and Beijing University of Aeronautics and Astronautics in China. With collaboration from Siemens Corporation, a wheeled mobile vehicle is being developed for use as a personal robot for operation in the home and office. Vehicle control incorporates feedback from ultrasonic and infrared transducers.

Facilities

Design and simulation software includes symbolic numerical manipulation (Matlab, IMSL), kinematic simulators (Mechanical Advantage), finite element modelling and analysis (ANSYS), and electronic and mechanical CAD/CAM (OrCAD, AutoCAD, Personal Designer, Personal Machinist). A large selection of Sun SparcStations, 486-class PCs, and an SGI Indigo workstation provide the compute power to run these packages. Computer graphics equipment includes three Silicon Graphics Indigo machines, two Indigo Extremes, and associated video equipment for simulation and animation. Research facilities available for graphics include a pair of Sun Sparc-Stations and four Silicon Graphics Indigo graphics workstations and video equipment in the Computer Graphics Research Laboratory. Software resources include ECHIDNA, developed by Bill Havens, LIFE FORMS - a system for human animation research developed by Tom Calvert, Mechanical Advantage from Cognition Inc., and AutoCad from AutoDesk.

Robotics research is supported by several robot manipulators, Unimation PUMA 560 industrial robot, a Silicon Graphics Indigo workstation for graphical simulations, a laser range scanner (modified 100-X) that generates a full 2-D range image, a 24 bit coulour frame grabber connected to a Sun SPARCstation, two solid state CCD cameras and a high resolution colour display monitor. Projects in underwater vehicles can utilize the Underwater Research Laboratory's 4m x 4m x2 m test tank, an underwater acoustic testbed, a CTD profiler, and communications test equipment. Manufacturing process and reliability research is supported by three environmental chambers, two vibration tables, scanning electron microscope, several light microscopes and a variety of materials' test and inspection devices.

A departmental machine shop is available for small projects, and a larger, fully-staffed university machine shop is equipped with modern NC tools for complex mechanical implementations.

Industrial Collaboration

Publications (1990-present)

Books Journals Book Chapters Conference Publications Theses and Dissertations Other Publications