«BASIC HUMAN DECISION MAKING: An Analysis of Route Choice Decisions by Long-Haul Truckers John Holland Knorring Advisor: Professor Alain L. Kornhauser ...»
2.7.4 Ambiguity Aversion Lastly, in an interesting study published by Daniel Ellsberg, the concept of ambiguity aversion is discussed.39 This concept refers to the idea that people are more comfortable with risk and taking risks when they know what the probabilities of outcomes are. This is entirely logical. When faced with the option to bet on a coin flip versus picking a black ball out of an urn filled with a number of balls whose colors he does not know, the average person would chose to bet on the coin flip. The implication to truck driver route choice is that when drivers are in unfamiliar areas, they are less likely to switch to alternate routes than when they are in familiar areas even if the level of information that they have on the alternate routes is the same. This can be simplified into saying that information, in this case the driver’s familiarity with their surroundings, is important in the decision making process.
2.7.5 Prospect Theory as it Relates to this Study So far, Kahnemen’s and Tversky’s Prospect Theory has been shown to be influential in the decision making process whether it be in terms of risk aversion or in irrational behavior. Prospect theory is important to this study because it provides some explanation of the decision-making processes that individuals use. This thesis is an attempt to examine the basic human decision making processes. Kahneman and Tversky suggest that humans do not always follow rational decision making processes in that they change their preferences based on how the decision will affect their wealth and how the 39 Ellsberg, Daniel, “Risk, Ambiguity, and the Savage Axioms.” Quarterly Journal of Economic, Volume LXXV,1961, pp. 643-669
question is proposed. This is an interesting idea that this thesis will attempt to shed light on.
2.8 Other Scholarly Truck Specific Studies An enormous number of studies have been done on passenger vehicle route choice. There are however, relatively few studies that have focused solely on truck driver route choice. In addition, many studies have treated all vehicles in the system as the same type of driver. Passenger automobiles are grouped together with long haul truckers.
Other studies have entirely neglected the trucks in the system when doing their analysis.
As stated earlier, trucks are not identical to passenger automobiles in terms of the driver’s risk behavior and the type of routes that they pick among other characteristics.
Additionally, passenger vehicles have significantly higher flexibility than long haul truckers in terms of destinations. Trucks generally have to be at a certain place at a certain time, and these characteristics are beyond the control of the driver. For passenger vehicles, excluding the typical home to work trip, they have more flexibility on their destinations. For example, an individual is at home and he wants to get some gas for his car. He may have a preferred gas station to go to, but if it is impossible to get there, it is possible that the driver would divert to another gas station. Likewise, passenger vehicles generally have fewer restricted roads than long haul truckers due to height, weight, or hazardous material restrictions on trucks.
The primary reason for so few studies on trucks, according to Robert Stammer et al., is that there has been a lack of reliable Origin-Destination information on trucks.40 As seen in the previous chapter, O-D pairs are crucial in the network assignment process as they tell where the vehicle starts from and where it ends. Without this information, it is impossible to do the route determination stage of the network assignment process.
This data set, while lacking in actual O-D pairs as well, still proves very helpful in the analysis that will be done. A heuristic will be developed that will allow for generation of pseudo O-D pairs. This thesis will not look to build a model to incorporate the entire network; rather, case studies of small, representative, sections of the network will be examined. As a result, only trucks that pass through certain points on the network will be examined and those points will be used as the representative Origin-Destination pairs.
40 Stammer, Robert et al. Conducting Truck Routing Studies from a New Perspective. In Transportation Research Record 1038, TRB, National Research Council, Washington, D.C., 1985, pg. 59.
3 Explanation of the Data Set The data set this thesis uses, along with the analysis of it, is the most valuable portion of this study. The data set, after significant data reduction, which will be explained later, consists of over 60,000,000 observations of truck time and location data for over 249,465 unique trucks over a thirteen-day period between August 29, 2002 and September 10, 2002. These observations shed light on the movements of long haul truckers across the United States highway network. By analyzing these movements, this thesis will explore the basic decision making processes that the drivers use. Up until recently, the measuring and modeling of human behavior has been very difficult.
Acquiring a sufficient amount of unbiased, complete, and accurate data has been the major shortcoming that has plagued other studies. This thesis is one of the first academic studies that has access to a data set large enough that random sampling and scarcity of observations considerations are only minor factors. While this revealed preference data set does not show many of the conditions that are factors in the utility function, and thus the decision making process, it does allow for many characteristics to be derived. Each
observation in the data set is a vector that consists of the following:
Ø A unique alphanumeric truck identifier (it remains constant throughout the
Ø A longitude position in the format “ddd.mm’ss” The following sections will discuss ways to collect this type of data, problems associated with using this type of data in determining route choice decision making behavior for the general truck population, and other hurdles that this data set posed.
3.1 Collection of Data Though the exact method used to collect this data is not known, there are three likely candidates as possible methods of collection for position data: Inertial Navigation Systems, Loran, and Global Positioning System (GPS). The first system relies on very sensitive motion sensing equipment and accelerometers to determine location information.41 Loran and GPS rely on a concept known as Time of Arrival ranging to calculate location information.42 The following sections will discuss the intricacies of Time of Arrival Ranging, Loran, GPS, Inertial Navigation Systems and lastly some of the problems involved with data collection using these systems.
3.1.1 Time of Arrival Ranging Time of Arrival Ranging, or TOA, is one of the most popular methods of determining one’s position relative to his surroundings. TOA relies on the measurement of the time that it takes for some sort of signal to travel from a known location such as a 41 Farrell, Jay A. and Matthew Barth. The Global Positioning System and Inertial Navigation, United States of America: R. R. Donnelley & Sons, 1999, pg. 13.
42 Kaplan, Elliot D. Understanding GPS: Principles and Applications, United States of America: Artech House, Inc., 1996 pg. 15
radio tower, satellite, or lighthouse to the observer’s location. The observer calculates the time that it takes for the signal to arrive and that time is then multiplied by the speed of the signal that results in the distance to the signal source or transmitter. The speed of the signal is generally either the speed of light or the speed of sound depending on the type of signal generated by the transmitter.43 After the distance to multiple transmitters has been determined, the user uses the principal of trilateration to determine his exact location.44 In determining a latitude and longitude position for a two-dimensional surface like this thesis uses, a number of transmitters are required for exact locations. When the user picks up the signal and determines the range to the transmitter, he knows that he is a specific distance from the known location of the transmitter T1, but he does not know what direction he is. In other words, the user (P) knows that he is somewhere on the perimeter of a circle with radius R1 centered at T1.
In order to determine the user’s location more accurately, he will need to attain signals from multiple sources. In the following case, the user has picked up a signal from two different transmitters, T1 and T2. The user can narrow down his position from somewhere on the perimeter of circle 1 to one of the two points at the intersection of circles 1 and 2, P.1 or P.2. He can do this because he knows how far he is from both transmitters. The only locations on a flat plane that satisfy the distance locations are P.1 and P.2. Additionally, there is a special case that only requires two transmitters to determine one’s exact position. This only happens though when the user is standing exactly on a line that passes through both T1 and T2.
The exact location can now be determined with the help of a third transmitter as long as the three transmitters are not located in a straight line. With the third transmitter, the location P.1 is eliminated and P is selected as the location of the user, as P.1 is not R3
Loran and the Global Positioning System use this technique of trilateration to generate position information. If we generalize the surface of the surface of the earth to be a flat plane this exact technique can be used to generate position information. Loran outputs only latitude and longitude for a position, not altitude, and the area of coverage is small enough that approximating the surface of the earth to be a flat plane does not reduce the effectiveness of the system a significant amount.
Extending this example of trilateration to GPS is not difficult. Fundamentally, the mechanics of 3-D positioning is the same as 2-D positioning. If one thinks of the circles generated by the transmitters to now be spheres in three-space, the same technique can be
used. If the user determines that he is 25,000 miles from satellite A, he knows that he is somewhere on the surface of a sphere with a radius of 25,000 miles centered at satellite A. Once the user obtains a second signal from satellite B, he can determine that he is on an exact circle at the intersection of the two spheres. With three satellites, the user will know that he is at one of the two points of intersection of the three spheres. The user only needs to get a signal from one additional transmitter to determine his exact altitude relative to the transmitters. However with GPS, users are usually only on the surface of the earth. As a result, the surface of the earth can also act as a sphere so that only three visible transmitters are needed, though four are usually used to improve accuracy of location and the internal quartz clocks in the receiver units.
3.1.2 Loran LOng RAnge Navigation, or Loran for short, is a system of navigation developed during World War II to help the U.S. Navy to improve their day as well as night navigation.45 Today Loran is a federally funded terrestrial system that provides users with positioning as well as timing information. The system has coverage anywhere in the continental United States as well as much of the costal waters. Currently, the United States Coast guard operates and maintains the Loran system at an annual cost of around $30 million. The current system of Loran, Loran-C, has been in place since 1957.46 Additionally, there are other nationalized Loran systems throughout the world in countries like China and Russia.
45 Pierce, J. et al. Loran: Long Range Navigation. United States of America: McGraw Hill Book Company, Inc. 1948. pg ix.
The Loran system uses the principles of TOA ranging to determine locations.
Radio transmitters are setup in a group of three or more and are separated by hundreds of miles. Each group has one master transmitter and a series of secondary transmitters. The stations constantly broadcast radio signals with precise timing information. These timing signals are gathered by the receiver and compared with the other timing signals from different master-secondary transmitters. The receiver measures the difference between the signals and calculates the distance to the transmitters. With these distances and the precise locations of the transmitters, the receiver calculates its position in the previously mentioned manner.47 In general, the position information generated by Loran systems is very accurate.
While accuracy of 5 meters is possible in conjunction with other navigation devices such as GPS, the accuracy in most areas covered by Loran is closer to 100m-300m range of accuracy. With this level of accuracy, compounded by the fact that Loran is available across the entire continental United States, Loran could be used to determine position information for a truck. Additionally, because the system is accurate enough, it is possible to determine what highway a truck is on with a high level of accuracy.48 3.1.3 Global Positioning System The Global Positioning System, or GPS, is an all weather, worldwide, continuous coverage, satellite-based radio navigation system that utilizes the principles of TOA to determine the location of GPS receivers anywhere in the world with a clear view of the 47 Ibid.