CASTLE Lab combines sponsorship from government sources, which helps to support the basic research, with substantial industry support from our corporate sponsors. The early stages of our work, primarily in the 1980's, was supported by a variety of grants from the National Science Foundation. Since the early 1990's, our work has been supported by the Air Force Office of Scientific Research, with additional support from the National Science Foundation. This funding has created our most dramatic theoretical and algorithmic breakthroughs. In particular, the concept of optimizing simulators, the CAVE algorithm for approximating concave value functions, the modeling paradigm for dynamic resource transformation problems, the Java library for modeling this problem class, and the successes we are having in the solution of integer, multicommodity network flow problems (for both deterministic and stochastic problems) all can be traced to the financial support of AFOSR.
Our Corporate partners
CASTLE Laboratory could not survive without the active participation of its corporate partners. Our modeling methodology is based on the premise that the best information about how a company can and should be run rests with the people who work there every day.
The Lab focuses on three dimensions of research: theory, laboratory experimentation, and field experimentation. We have not adopted the conventional paradigm of "theory and application." In our work, the testing of new ideas in actual physical settings is central to our work. Not surprisingly, the challenge of watching our systems run in the real world introduces issues that change fundamental modeling assumptions. Nowhere has this been more significant than the realization of the importance of information: changing information, missing information and bad information.
Each of the links below contains a short summary of the work being undertaken with each company. Note that each link contains a bitmapped image. If you are not on a fast connection, this may take a while.
United Parcel Service - The largest freight transportation company in the U.S., UPS moves over two million packages each day. By modeling the organization of information and decisions, we are developing modeling technologies that scale to the size of the UPS network.
Netjets - The largest operator in the "fractional jet" business, operating a fleet of over 500 aircraft and over 2,000 pilots.
Embraer - Based in Brazil, Embraer is a major manufacturer of smaller aircraft. Embraer has contracted with CASTLE Lab to develop an advanced system for managing over 10,000 spare parts for a new class of jets.
Schneider National - The one of the three largest truckload motor carriers in the U.S. with a fleet of over 15,000 trucks. Schneider pioneered the first use of optimization models in truckload trucking and the first use of satellite communication. CASTLE Lab has developed a planning system that required solving a dynamic program exhibit a state variable with millions of dimensions.
Yellow Corporation (now YRC Wordwide) - The largest less-than-truckload motor carrier, Yellow was the founding Corporate Partner for CASTLE Lab. Yellow provided several of the truly large-scale applications which motivated our research.
The Air Mobility Command - The AMC handles the movement of people and goods for the military around the world. The funding for our work with AMC comes from the Air Force Office of Scientific Research, but the AMC has been a source of challenging problems.
The Norfolk Southern Railroad - Norfolk Southern has provided sustained support in the development of models for flatcar management, locomotive optimization and freight car distribution.
The Burlington Northern Sante Fe Railroad - The BNSF railroad participated in the development of our first-generation locomotive optimization model.
Air Products and Chemicals - Air Products supported the lab for many years in the development of a production routing and scheduling system to manage drivers, tractors and trailers moving liquified gases to serve customers (our first "five-layer" resource allocation problem).
Burlington Motor Carriers - We worked with BMC from when they consisted of five employees. BMC participated in the development of our first, aggregate fleet management model ("LOADMAP") and later in the development of a real-time driver matching system. BMC also used a dynamic network model ("CASPER") developed at CASTLE Lab to perform account evaluations. When it was first used, CASPER helped reduce BMC's operating ratio from over 110 (costs were 10 percent higher than revenues) to almost 95.
From the 1980's:
Yellow Freight System - Yellow was a rapidly growing natoinal LTL carrier. Yellow picked up the development of a network planning model, originally called "APOLLO" (see below). Yellow implemented the system as SYSNET (SYSNET is a registered trademark of Yellow Corporation) which was an interactive optimization system for the service network design problem faced by LTL carriers. This system was one of the original products of Princeton Transportation Consulting Group (now a part of Manhattan Associates) and is still in use by LTL carriers over 20 years later (it was originally marketed as "SUPERSPIN").
Sea-Land - Once one of the largest ocean cargo companies, Sea-Land managed over 300,000 containers to move freight around the world. The Sea-Land operations have gone through several mergers, and are now a small part of Maersk.
North American Van Lines - NAVL, best known for its household moving services, had three divisions in the 1980's: household moving, high value products (less-than-truckload trucking for computers) and truckload trucking (its Commercial Transport division). We worked with all three divisions. We were a finalist in the Franz Edelman competition with our implementation of a fleet management system for the commercial transport division.
IU International - IU owned two national LTL carriers - Ryder Truckline, based in Jacksonville Florida, and Pacific Intermountain Express (P.I.E.). IU funded a research project at Princeton University which produced "APOLLO" - the first interactive optimization problem for network design in transportation. APOLLO became the foundation of a system that was implemented at Yellow Freight (over the 1986-1989 period) and became a product sold by the first consulting firm started by CASTLE Lab - the Princeton Transportation Consulting Group.
(c) Princeton University - No images may be used without the express written permission of Warren B. Powell (firstname.lastname@example.org).