Less-than-truckload (LTL) carriers collect freight from various shippers and consolidate that freight to fill trailers for travel to common destinations. LTL carriers run high-volume operations, often spending millions of dollars in transportation and handling costs in a single week.

An LTL motor carrier transports shipments that
typically occupy only 5 to 10 percent of trailer capacity. As a result, LTL carriers collect and consolidate freight from various shippers to increase trailer utilization, referred to as the load factor. Consolidation does come with a cost; by transferring freight between trailers, the carrier incurs a handling cost and increases the total time and distance a shipment requires to reach its destination. Supporting these operations is a system of terminals, tractors, trailers, dockworkers, and drivers***collectively called the linehaul network. As competition increases and shippers raise their expectations for service, LTL carriers must optimize their linehaul networks to remain viable.

Researchers in the Georgia Tech Supply Chain & Logistics Institute (SCL) have been working together with two major U.S. less-than-truckload carriers, Saia and YRC Worldwide, for a number of years on developing a suite of decision support tools to optimize the linehaul network. The suite of decision-support tools contains, among other modules:

*An optimization model that determines the appropriate role of a terminal in the network, i.e., should a terminal be used as an entry and exit point for freight into the system "end-of-line terminal" or should it be used as a major consolidation point where freight from various end-of**-lines is cross-docked "breakbulk (hub) terminal"

*An optimization model that determines optimal freight paths for all origin-destination combinations, i.e., a set of freight paths respecting service commitments leading to the highest utilization of trailer capacity

*An optimization model that determines the minimum number of tractors required to ensure that all freight moves through the linehaul system on time

*An optimization model that schedules the drivers to move freight through the network, respecting hours-of-**service regulations and company policies

*An optimization model that determines driver domiciles, i.e., terminals where drivers should be located to most cost-effectively move freight through the network

One of the most critical decisions in operating a linehaul network is how to route freight from origin to destination as it directly impacts consolidation opportunities. Shipments are quoted a service standard from origin to destination in business days. Historically these standards were long enough (often five business days) that service only loosely constrained how a carrier chose to route a shipment from origin to destination. In today's environment, service standards of one, two, and three days are common, and service must play a significant role in a shipment's path. At the same time, shorter service standards reduce consolidation opportunities due to the handling time and circuity that consolidation requires. As a result, carriers need optimization techniques for designing load plans that accurately model how short service standards constrain freight routing and the consolidation opportunities that exist. An integer programming heuristic has been developed that constructs high-quality, service-feasible load plans.

To be able to compete on price, carriers have to find ways to increase the utilization of their current infrastructure. One possible way to do so is to relax some of the self-imposed rules that constrain how freight flows through the system. For example, a traditional load plan assumes the same freight routing decisions are made every day. However, a load plan that accounts for predictable daily freight variations by allowing for different freight routing decisions on different days may substantially increase utilization. In general, as information technology infrastructure improves, a carrier can consider different rules of operation. Yet without load plan design technology that is adaptable to changing operational constraints, it is difficult for a carrier to quantify how changing or relaxing these operational constraints will affect business. An analysis reveals that savings of up to 4 percent can be achieved by allowing day-differentiated load plans.

One of the most complex decision problems faced by an LTL carrier is scheduling its drivers. This is due to the various rules governing the feasibility of driver duties. Hours-of-service regulations imposed by the Department of Transportation to ensure the safety of drivers and others on the roads, for example, specify that drivers cannot drive more than eleven hours and cannot work more than fourteen hours before a mandatory rest of at least ten hours is required. LTL driver scheduling is further complicated by the fact that trucking moves are not pre-scheduled. The decision technology developed for LTL carriers combines greedy search with enumeration of time-feasible driver duties and is capable of generating in a matter of minutes cost-effective driver schedules covering 15,000 to 20,000 loads satisfying a variety of real-life driver constraints. A comparison with real-world dispatch data indicates that the technology produces driver schedules of very high quality.

The driver scheduling technology is embedded in an iterative scheme to determine the best possible home locations, or domiciles, of truck drivers for an LTL carrier. Domiciling decisions are complex, in part due to regulations and union rules restricting driver schedules, but have a significant impact on the operating costs of LTL carriers. In each iteration of our scheme drivers are allocated to terminals, and drivers' bids are determined so as to satisfy union requirements. An analysis of the resulting driver schedules is used to guide the next iteration. Computational experiments demonstrate the value of the iterative scheme and also quantify the impact of union rules on the number of drivers required (and thus on operating costs).

The LTL industry is essential for our economy, and the research conducted at SCL aims to increase the industry's efficiency and effectiveness. The research of Professors Alan Erera and Martin Savelsbergh has involved a number of PhD students, several of whom have written their PhD theses focusing on optimization problems in linehaul networks.

This article was written by Alan Erera and Martin Savelsbergh and first appeared in the Fall 2009 issue of Industrial and Systems Engineering, the alumni magazine of the Stewart School of ISyE.