|
In the past 5 years,
the principal investigator has continued to conduct research in three
directions: Network Management and Control, Production System Design,
Scheduling and Control and Parallel Computing for Large-Scale,
Distributed Decision-Making. Results are summarized as follows.
I. Production
System Design, Scheduling and Control
I.1 Research on
Optimization Theory with Applications to Production Scheduling and
Control
- Is investigating
auction-based scheduling algorithm for distributed applications.
- Combined ordinal
optimization-based simulation and a design of experiment method into a
methodology for efficiently selecting a good enough solution from
combinatorial options and is conducting further study.
- Combined Lagrange
relaxation and minimum cost network flow algorithms into new and
optimization-based scheduling algorithms for flow shops with setup
effects.
- Developed production
scheduling systems based on the algorithms developed for real
applications to semiconductor manufacturing factories.
I.2 Design
and Implementation of Decision Support Systems for Semiconductor
Manufacturing:
- Developed with TSMC a learning software
agent for supervisors of semiconductor tool dispatching.
- Designed a queueing analysis and
fixed-point iteration-based algorithm for daily production resource
allocation.
II. Parallel
Computing for Large-Scale, Distributed
Decision-Making
Realization of a Lagrange relaxation neural network-based scheduling
algorithm into a parallel VLSI implementation (745mW, 3.3V, 100MHz,
and 4mmx4mm in chip size)
III.
Network Management and Control
III.1 Research
on Quality of Service (QoS)-based Network Service Pricing and Its
Enabling Technologies
- Is investigating QoS provisioning over
private virtual networks and the associated pricing policy and
enabling techniques.
- Designed and implemented a quota-based
priority control system (http://vip.ee.ntu.edu.tw)
that achieved fairness improvement and abuse reduction over a
5000-user dormitory network of National Taiwan University.
- Designed single-domain and
auction-based multi-domain pricing policies for the DiffServ
architecture with lab implementations.
III.2 Design
and Development of ATM Network Management and Control: Routing and
Reconfiguration
- Designed a heuristic routing algorithm
for reducing delay while satisfying loss probability constraint over a
hierarchical ATM network management architecture.
- Designed a self-healing algorithm for
fast virtual path reconfiguration satisfying bandwidth requirements
under a network fault.
- Implemented the designs on a FORETM
switch-based ATM network. |