Integrated Logistics

1
Integrated Logistics

• R. Ravi, CMU
• Co-organizers Adam Meyerson, CMU
• Moses Charikar, Princeton
• Ted Gifford, Schneider Logistics

2
Motivations

• Integrate models across many application areas
  Databases, Genetic clustering, Supply Chain
  Management
• Bridge across various disciplines working on same
  models CS, OR, Industry
• Postdoc-propelled! Adam Meyerson from Stanford,
  going to the faculty at UCLA

3
Sample Applications

• Trucking logistics (Gifford)
• Clustering expression data (Munagala)
• Databases (Guha, Mettu)
• Survivable Telecomm Networks (Balakrishnan,
  Mirchandani)
• Supply Chain Models (Goetschalckx, Schaefer)
• Disk Placement (Khuller)
• Network Games (Tardos. Wexler)
• Overlay Multicast Networks (Maggs)

4
Affiliations of Participants

• CS Departments CMU, Princeton, MIT, Berkeley,
  Penn, Cornell, UMCP, Dartmouth, Stanford
• CS Research Labs IBM, Bell-Labs, Microsoft
• Business Schools CMU, Pitt, UT Austin, Georgia
  Tech
• Industry

5
Integrated Logistics

• Integrate formulations and approaches in
  Logistics across applications and disciplines
• Integrating theory and practice
• Integrate models of facility location and
  transportation into one comprehensive model (as
  opposed to handling them in two distinct tactical
  stages)

6
Research Goals and Plan

• Bring researchers from CS, OR and practitioners
  together for dialogue
• Hope to stimulate new work motivated by exchange
• Adapt and integrate algorithmic approaches across
  areas
• Disseminate algorithms in course, web depot,
  teaching modules

7
Viewpoints

• Algorithms viewpoints Surveys by Meyerson and
  Shmoys First Workshop
• Industry perspective Teds Presentation
• Business School/OR perspective of Logistics
  Marc's Presentation

8
Online Facility Location

• Given facilities with opening costs f in a
  metric, locate them to minimize total facility
  opening costs plus distances from all clients to
  their resp. closest facility
• We start with some graph and its solution, but we
  will have to add more vertices in the future,
  without disturbing our current setup
• The demands of incoming clients are based on some
  known function, generally of distance
• Goal what do we do with each incoming point as
  it arrives to stay close to optimal?

9
Online Facility Location
What do we do with incoming vertices?

• With each new client, we do one of two things
• Connect our new client to an existing facility,
  or
• Make a new facility at the new point location

10
Theoretical Result (Meyerson)

• The probability that a Facility is created out of
  a given incoming point is d/f
• Where d the distance to the nearest facility
• And f the cost of opening a facility
• Worst case cost is expected 8 times the optimal
  cost

11
Goals of REU Investigation (Bleimes, Garrod,
Meyerson)

• Motivation Rather than a new approach, examine
  the realistic behavior of existing techniques for
  facility location
• Task Run simulations over both real and random
  data sets, to get average data on the performance
  of known algorithms for this problem
• Expected Results
• Both speed and accuracy are important, but for
  different reasons and applications
• Realistic data will help determine how best to
  use these algorithms

12
Research Accomplishments

• 11 papers on topics ranging from networking,
  routing, orienteering, designing mechanisms to
  scheduling
• New ideas on online cost-distance, truth-telling
  mechanisms for network pricing, discount-reward
  TSP

13
Education Outreach

• Guest Lectures in Algorithms in the Real World
  Class and College Teachers Workshop
• Graduate Student Training (Garrod, Dhamdhere,
  Konemann, Sinha) and REU (Bleimes,Kitchin)
• Graduate Class on Planarity (Spring 2003)

14
Future Research

• New results on applying approximation algorithms
  for two-stage stochastic optimization problems in
  Facility Location and Network Design (Ravi
  Sinha, submitted)
• Web depot on Logistics implementations,
  benchmarks and test sets (Meyerson)