Projecting line-haul truck technology adoption: How heterogeneity among fleets impacts system-wide adoption

Abstract

A System-of-Systems engineering methodology is used to project truck technology adoption behaviors of heterogeneous fleets operating over the U.S. line-haul freight transportation system. A constrained mixed-integer linear program is formulated to optimize total cost of ownership of regional fleets given vehicle highway performance, fleet operations, cost of energy, and freight demand. A design-of-experiments demonstrates adoption sensitivity to economic parameters and individual fleet management constraints. Validation results demonstrate the importance of modeling fleet heterogeneity to achieving 90% prediction accuracy of historical adoption of three different vehicle architectures across 12 representative fleets over a 11-year period.

Introduction

A freight transportation system (FTS) is a critical element of an industrial and globalized economy, but it also imposes a high cost on the environment. In the US alone—where 60% of freight is moved on highway roads by Class 8 trucks—freight transportation trucking generates an estimated 450 million tons of CO₂ emissions annually (U.S. Department of Transportation BTS, 2016). Presently, an overwhelming majority of heavy-duty Class 8 vehicles use diesel internal combustion engines, with less than 2% using natural gas or other alternative fuels (Reinhart, 2016). Adoption of cleaner vehicle technologies in the heavy-duty trucking segment is regarded as a promising strategy to reduce emissions in the transportation sector. However, the market penetration levels that these emerging technologies will ultimately achieve depends on their economic attractiveness as perceived by the fleets that will acquire and operate them. It is also well known that freight technology adoption is a function of not only the acquisition price of a given technology or its performance benefits (e.g. as characterized by an average increase in fuel economy), but also state and federal policies and incentives, regional fuel costs, fleet operation strategies, network characteristics, etc. (Yeh, 2007). Moreover, these factors also affect how a fleet operates the vehicles they purchase given the varied performance capabilities of said vehicles. More importantly, the decision-making behavior of fleets servicing the FTS vary as a function of their size, operating strategy, annual growth, vehicle replacement cycles, etc. Ostensibly, these fleet entities make their acquisition decisions independent of each other, however they operate and interact on a single, shared transportation network, altogether influenced by the evolution of the FTS environment. In other words, in order to capture emergent mixed technology adoption trends in the freight transportation network, its evolution must be modeled across the heterogeneous mix of fleets that service it.

However, while many researchers and stakeholders are interested in modeling technology adoption across the freight transportation sector, the literature on this subject is at best limited to a single fleet’s adoption behaviors, therefore assuming a homogeneity that is not realistic across the FTS. Davis and Figliozzi (2013), for example, project the number of electric vehicles purchased and operated annually for a single delivery fleet by optimizing the cost of ownership over an 11-year period and comparing it to that of a fleet using all diesel vehicles as a measure of competitiveness. Feng and Figliozzi (2012) present an ownership cost minimization model to project the mixed adoption of diesel and electric delivery vehicles similarly for a single fleet. Furthermore, although both studies take into consideration ownership cost sensitivity to vehicle speed, battery life, and tax incentives, they impose daily energy consumption estimates and route lengths, without considering the dynamic effects of the FTS evolution on vehicle utilization. Janic (2007) presents a model to calculate the reduced total costs of an intermodal collection and distribution fleet when heavier and longer vehicles are used. The costs are based on historical local and global aggregated data, and not on the operational decisions made at the fleet level.

Moreover, a majority of the literature on projection of trucking freight technology adoption focuses on evaluating the drive-cycle performance of emerging vehicle technologies and resulting fuel or energy savings as a measure of their economic attractiveness and potential for widespread adoption, rather than the direct effects that vehicle performance and FTS evolution will have on fleet adoption behaviors (Delorme et al., 2010, Lammert et al., 2014, Zhao et al., 2013). Other studies take into account drive-cycle performance and extrapolated historical data—projected number of vehicles, annual vehicle miles traveled, etc.—and project resulting emissions reduction given different assumed market penetration scenarios. Fulton and Miller (2015) explore high levels of market penetration of different low-carbon vehicle technologies, such as alternative fuel architectures and electric vehicles, and the resulting capability to meet 80% reduction of CO₂ emissions in the U.S. by 2050. Schäfer and Jacoby (2006) present rates of adoption for different personal vehicles and heavy-duty Class 8 diesel truck technologies under CO₂ emission constraints and penalty costs. Wadud et al. (2016) present a qualitative estimate of changes in energy consumption and carbon impact of light-duty and heavy-duty vehicles given the levels of automation adopted in four assumed scenarios. While these articles describe projected adoption scenarios and resulting emissions levels, they do not provide any further understanding on the mechanisms that lead FTS fleets to exhibit the vehicle adoption behaviors in the scenarios presented.

The future technology composition of the FTS will likely be mixed as a result of the many independent fleets that purchase and operate vehicles based on their individual management and operational strategies; however, these fleets are altogether influenced by, and collectively affect, the dynamics and evolution of their common environment. System-of-Systems Engineering (SoSE) (Jamshidi, 2008) is an emerging and specialized field of research in complex systems that addresses the design and analysis of capabilities resulting from the interaction and collaboration of large-scale independent and distributed systems. A foundational principle of SoSE is that the constituent systems maintain their independence (e.g., fleets) but contribute to, or take resources from, other systems (e.g., road network, traffic) for mutual benefit (Maier, 1998). Any change in one system, therefore, ultimately impacts other systems and the resultant System-of-Systems (SoS) capability (i.e., the transportation of freight). Even though the evidence of an SoS perspective on freight technology adoption is lacking in the literature, many researchers are beginning to recognize the value of SoSE methods for investigating transportation related problems. Aliubavicius et al. (2016) utilize SoSE concepts to discover emergent behavior in integrated traffic control problems, while, Roncoli et al. (2013) develop an SoS model for real-time scheduling of dangerous good transportation to meet regulatory requirements under business delivery constraints. Mostafavi et al. (2011) claim the SoS framework to be the foundational element for developing innovations for future intelligent transportation systems because of its ability to accommodate a diverse set of stakeholders and complex constituent systems.In addition to the transportation related examples, the SoSE concepts are being applied across diverse application domains where multiple complex independent systems are integrated, e.g., space applications (Martin, 2008), emergency management (Liu, 2011), and supply chain management (Choi et al., 2018), to name a few.

In this paper we contribute an approach for modeling the heterogeneity of fleet vehicle purchase decision-making behaviors by first modeling the FTS as a system-of-systems and then designing and validating a representative set of fleets to project technology adoption trends. The SoSE principles allow for the development of a holistic FTS model that captures the interaction of the multiple complex systems that compose it. Here, we pose the problem of projecting mixed technology adoption across the FTS based as a cost minimization problem in which we consider how the FTS affects, and is affected by, the purchase and vehicle allocation decisions of multiple heterogeneous fleets. The proposed model is parametrized and validated using a Design of Experiments (DOE) and historical adoption data. The results of this SoS model demonstrate 90$\%$ accuracy in prediction outcome when modeling historical technology adoption across a set of 12 heterogeneous representative fleets over an 11-year period.

This paper is organized as follows. Section 2 introduces an overview of the SoS methodology. Section 3 describes the proposed linear program formulation for modeling fleet adoption behaviors based on an optimization of regional fleets’ operational and purchasing costs. Section 4 presents an analysis of the proposed model using technology adoption data between the years 2005 to 2015. Concluding remarks and future work are discussed in Section 5.