Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Innovative Tour-Based Truck Travel Model Using Truck GPS Data Arun Kuppam, Jason Lemp, and Dan Beagan Cambridge Systematics, Inc. Vladimir Livshits, Lavanya Vallabhaneni, and Sreevatsa Nippani Maricopa Association of Governments Presentation Notes: Presented by Arun Kuppam, Cambridge Systematics, Inc. The tour-based truck travel model used non-inclusive data collection, relying on third-party data vendors for information. This included deploying Global Positioning System (GPS) units and collecting records from truck fleets. Commercial GPS data is widely available and common since companies use it for day-to-day logistics; however, privacy limitations require more data processing. The data set included 3.5 million GPS records from approximately 22,000 trucks, primarily heavy trucks. This represented about 60,000 truck tours. The GPS event is a signal that transmits the data back and forth, such as a signal stop, breakdown, or delivery. By collecting latitude and longitude data from American Transportation Research Institute (ATRI), it is possible to derive distance and speed. The origin lat/lon information was very similar to the final lat/lon. In this approach, parcel data were overlaid to identify land uses at these locations. This provides information as to the purpose of the truck trip and the associated land use. Stops were categorized into 10 stop types, with variables based on purpose type, derived from stop length, accessibility to employment, the number of stops on the tour, and the tour purpose. Next steps include further calibration and validation and determining the adequacy of GPS data to sample size and biases. Abstract The concept of truck travel demand forecasting, internal to a region, has always been built upon modeling discrete truck trip ends, distributing truck trip ends to various origins and destinations using travel time impedances and some land use characteristics, and allocating truck trip tables into distinct time periods, using factors derived from observed counts. An innovative enhancement to this approach is to apply activity-based modeling (ABM) principles to truck tour characteristics and develop a tour-based truck travel demand model. In the recent past, Cambridge Systematics (CS) has successfully acquired, processed, and used truck GPS data to update urban truck models for MPOs. CS, through a contract with the Phoenix MPO, Maricopa Association of Governments (MAG), acquired and processed truck GPS data from the American Transportation Research Institute (ATRI) to develop an innovative framework that links trucks into trip chains in a tour-based model. These data had over 3 million GPS event records for the month of April 2011, which were reported by over 20,000 trucks. These trucks yielded about 20,000 truck tours comprising over 62,000 stops at various land uses. About 95% of truck tours are generated by retail establishments, farms, households, wholesale 30
trade, and manufacturing facilities in the region. This truck tour database formed a strong foundation to estimate robust tour-based models for various industry sectors. GPS Devices in Trucks GPS devices are widely deployed in cell phones, autos, and trucks. These devices can display information about the position of the vehicle, often on a map of the area, and the desired destination, based on signals received from GPS satellites. Sometimes these devices not only receive the GPS satellite signals or other information, such as traffic conditions, they also may wirelessly transmit that information back to a central location. The GPS event information is collected to serve the business purposes of the truck fleet operators. Those businesses are under no obligation to share this information with others. In fact, this information, since it contains sensitive information about the business practices of truck fleet operations, is contractually protected when it is collected as a part of a subscription but provided to third parties. However, information pertaining to the GPS locations (in the form of latitude and longitude) and the time stamp at which the transmission was sent is available. This information can be processed to derive a truck trip or tour database that can be used for model estimation. ATRI GPS Data for Phoenix CS processed the ATRI data for the MAG modeling area for the period from April 1, 2011, to April 30, 2011. The raw data delivery from ATRI contained 3,429,603 GPS event records. There are GPS event records reported for 22,657 trucks that indulge in 58,637 tours. At these GPS events, the vehicle may be stopped or moving. In principle, only certain stopped records can be grouped into tours or trips, but tours or trips cannot be precisely computed without further processing. A tour is defined as a sequence of GPS events for a given truck that is only intended for the initial filtering of the GPS records. Subsequent processing was done to determine truck tours consistent with its use in the development of touring models. Truck Tour-Based Model The objective of the truck tour model is to develop truck trip chains by industry sector. These truck trip chains can be grouped into the major linkages based on the land uses the trucks make stops at and the probability of making another stop based on the number of previous stops. For each truck tour, a series of choice models are employed in order to determine the time period of tour start times, propensity to make additional stops, next stop purpose, location of the stop, and stop duration. The model generates the number of stops by industry sector (e.g., retail or manufacturing) and then strings the individual trips together into tours. The number of stops on a tour, the type of stops, and the location and time of day of stops are all estimated based on the type of truck making the tour, the activities conducted by the truck, the characteristics of the stops, and the traffic conditions in the network. All the tour model components were coded in Geographic Information System Developerâs Kit (GISDK) and implemented in TransCAD. Each component 31
was individually assessed and calibrated. The reasonability of the explanatory variables was determined by their magnitude, t-statistic, and relation to the dependent variable. The individual model outputs were also compared with the truck GPS data to assess the model performance. These comparisons indicated that the model components are predicting very closely the observed data for the most part. There are some differences, which can be further improved upon with more rigorous calibration and validation of the model. 32
