- IBM
Journey Management Library
Journey
management - a concept IBM began developing in 1996 - is concerned
with delivering a positive experience to travelers as they proceed
on their journeys. Using simulation techniques, IBM is helping travel-related
service providers investigate new ways to eliminate bottlenecks,
improve service, handle more customers, reduce costs, and enhance
revenues using advanced technologies.
- IBM
Steel Industry Solutions
The IBM
Steel Industry Solutions Group provides leading-edge solutions for
combinatorial optimization problems of steel manufacturers. These
solutions are based on advanced techniques from operations research
and artificial intelligence.
- IBM
Paper Mill Scheduling Solutions
IBM paper
mill scheduling system provides decision support for scheduling
paper manufacturing and distribution. The system, which is built
on an agent-based framework called A-Team, considers multiple scheduling
objectives and different stages of paper manufacturing and distribution
simultaneously, in a global multi-criteria optimization framework
and creates multiple high quality scheduling alternatives. The decision
support system consists of a suite of scheduling solutions. These
solutions have been successfully fielded at several paper mills
within the United States of America and are currently being sold
world-wide.
- IBM
Research in Stochastic Programming
Stochastic programming supports decision making under uncertainty.
It is a methodology for bringing uncertain future scenarios into the
traditional decision making framework of linear programming. Just
as linear programming models the optimal allocation of constrained
resources to meet known demands, stochastic programming models the
allocation of today's resources to meet tomorrow's unknown demands
in such a way that the user can explore the trade offs with respect
to expected risks and rewards and make informed decisions.
Users seeking the greater modeling power and flexibility of stochastic
programming come up against high hurdles: there is more input data
to be managed, the optimization problems are very large, and the solution
arrays are larger and more difficult to analyze.
IBM Stochastic Solutions offers new technology to help the user meet
the demands of modeling stochastic programs. It has an easy-to-use
solver that operates from the command line. It is the first commercial-grade
optimization solution to implement the Stochastic Mathematical Programming
System (SMPS) input format for multistage stochastic programs. The
flexible decomposition solver is robust and fast, and with callable
modules, the user has node-by-node access to data and solutions for
efficient solution analysis.
The components of IBM Stochastic Solutions are a solver, a suite of
callable modules for advanced development, and the User's Guide and
Reference containing documentation, problem examples, and sample drivers.
- IBM Microelectronic
Manufacturing Solutions
The IBM Microelectronics
Division uses an LP-based planning tool for planning production
starts (PROFIT). IBM Research is working closely
with IBM Microelectronics to improve the capabilities of this tool.
The first phase, which has been completed, was to improve the running
times by an order of magnitude. This was done by the use of a new
LP algorithm involving the solution of a sequence of LPs to obtain
an approximate optimum solution - finishing by a normal LP solve.
The second phase, which is undergoing acceptance tests, is to allow
for preferred lotsizes (e.g. only productions of 0, 100, 120, 140,
200, 220, 240 are valid). This would normally be a much larger mixed
integer model, but extending the capabilities of OSL
, this has been implemented with no increase in problem size and this
is giving good results in testing. Even with correct lotsizes, it
is still significantly faster than the original model. Co-operation
between IBM Research and IBM Microelectronics is continuing as PROFIT
is enhanced. "
Matching
Assets with Demand in Supply-Chain Management at IBM Microelectronics
Peter Lyon, R. John Milne, Robert Orzell, Robert Rice
Interfaces 31:1 January-February 2001 (p 108-124)
Available from the Interfaces website at http://silmaril.smeal.psu.edu/interfaces/
- Work-force
Scheduling
IBM Global Services (IGS) is the largest operator of service personnel
in NA. It operates about 7,000 customer service representatives (CSRs)
responding to more than 20,000 service calls per day. Managing this
work-force efficiently is a major challenge. The Theory of Computation
group in the Mathematical Sciences Department in IBM Research developed
an assignment engine that assigns calls to CSRs, striving to optimize
several business objectives (e.g., maximize customer satisfaction
by meeting contractual target times, minimize CSR idle and travel
times, minimize cost of parts shipment, etc). The main challenge that
the engine had to face was the speed: the requirement that the engine
recomputes the schedule every 10 minutes in order to react to changing
conditions.
The key ingredients that differentiate this engine from existing products
are: (1) the ability to optimize a large scale scheduling problem
quickly. (2) the dynamic nature that allows response to changing conditions.
(3) the consideration of a spectrum of factors simultaneously in assigning
calls to CSRs (like skills, CSR availability, utilization, call bunching,
customer priority, target time of calls, parts inventory and parts
shipment cost). (4) a novel way to consider parts availability.
The engine has been piloted for more than a year. The results of the
pilot show a reduction in the total time spent on a call (between
10% to 35%), and a decrease of CSR idle time by a factor of 5(!) to
1.75.
The assignement engine can be integrated quite easily to any environment/legacy
data bases. The interface to the engine is done using subroutine calls.
Each such subroutine call populates a specified record. The implementation
of these subroutines varies according to the environment. Thus, to
integrate the engine only these "middleware" subroutine have to be
rewritten.
Below is a list of the input and output records that are populated
by these subroutines:
INPUT
- Service
call record - information on the service calls: call id, location,
customer, availability hours, projected duration, required parts,
target time and type, etc.
- Alerted
call record - call id, type of alert
- Employee
record - information on employees: employee id, location, local time,
salary, etc.
- Employee
calendar - information on the employee work hours
- Employee
skills
- Inventory
parts record - information on parts inventory for each employee
- Ordered
parts status record
- Current
call assignment record
OUTPUT
- New call
assignment record
- Inventory
parts assignment record
- Part
order record - information on new parts ordered
|
