Omar Swei

Pavement life cycle assessments (LCAs) enable decision-makers to evaluate theenvironmental impact of alternative maintenance, rehabilitation, and reconstruction strategies.This thesis explores the viability of deep reinforcement learning (DRL), a framework that enablesagents to learn optimal actions within a given situation, to identify environmentally benignpavement management strategies. More specifically, this dissertation utilizes proximal-policyoptimization (PPO), a subtype of DRL algorithms, to identify a management strategy thatminimizes the expected global warming impact of a pavement facility over its lifecycle. Throughan urban Interstate case study, this thesis shows that the proposed PPO algorithm identifiesmanagement strategies that are anticipated to reduce the expected global warming impact of apavement facility over its planning horizon by 16 percent relative to traditional practice.Furthermore, the PPO algorithm is able to identify this management strategy in only 25 learningiterations, which is in stark comparison to Q-learning, a common reinforcement learningalgorithm, that requires 70,000 learning iterations. The results of this thesis highlight the viabilityof DRL to integrate within complex LCA models to determine environmentally sustainablepavement management strategies.

View record

Transport agencies must balance a range of performance objectives in managing their roadway assets. These performance objectives increasingly include mitigating greenhouse gas emissions and the environmental footprint associated with their roadway infrastructure. This dissertation outlines the creation of a network-level management tool that can support transport agencies in achieving two objectives for roadway systems: (1) maximizing the number of facilities in a “good” state-of-repair; and (2) minimizing their global warming impact. The former objective reflects the traditional priority for transport agencies, while the consideration of global warming impact is a new, emerging interest for these governmental entities. Unlike several past efforts in this domain, the network-level model accounts for a range of uncertainties (e.g., changes in traffic demand) and embeds flexible management strategies that can help transport agencies adapt to an unknown future. The network management tool is subsequently applied to a realistic case study based on data made available by the U.S. Department of Transportation via its Highway Performance Monitoring System program. The results of this dissertation highlight that, over a 20-year analysis period, maximizing the number of facilities in a “good” state-of-repair across the network increases its expected global warming impact by 1% to 8%. In other words, an unintended consequence of transport agencies emphasizing their efforts towards purely improving the physical condition of their roadway infrastructure is that it can substantially increase their environmental footprint. In addition, the case study demonstrates that, invariant to the available budget, increasing the allocation of funds towards rehabilitation rather than reconstruction treatments can reduce the expected global warming impact of the pavement network by as much as 7%. Simply put, by increasing funding allocation programs from capital outlays towards maintenance and rehabilitation activities, transport agencies have the potential to considerably reduce the global warming impact of their roadway systems. The contributions of this dissertation are two-fold. First, the proposed model provides decision-makers with a new framework to optimize across important performance targets while embedding adaptive roadway management strategies. Second, the case study findings provide transport agencies with valuable insights around the effects of key policy choices on their environmental footprint.

View record

The rising price to build transport infrastructure poses a challenge to policymakers who are grappling with increasingly limited available resources. The United States now gets less per dollar of infrastructure spending than it used to. Although transport infrastructure is a key input into long-term growth of the broader economy, the knowledge around the infrastructure prices and their potential drivers is limited. Therefore, this thesis aims to fill this gap by documenting and analyzing highway infrastructure spending between 2005 and 2017. More specifically, this thesis aims to improve current measures of productivity and price growth, and identify explanations for the increasing price of highway infrastructure. To improve the productivity measures in the sector, quality-adjusted producer price index of highway infrastructure is generated. The indicator of quality is the deterioration rate of a roadway. The developed deterioration model captures the effect of pavement performance improvements (i.e., quality) across time. By using an iterative-reweighted least squares approach, the model is applied to publicly available roughness data of asphalt concrete pavements collected as part of the FHWA’s long-term pavement performance program. Through analysis of highway projects across the contiguous U.S., the proposed quality-adjusted index reveals that annual productivity growth in the sector was underestimated by 2.0 percent. An econometric analysis of the highway sector aims to evaluate for decision-makers possible explanations for the price growth. The developed model identifies that highway prices are largely influenced by changes in labor, material input prices, and demand for more expensive roads. In addition, the results strongly suggest the presence of the Baumol effect, a phenomenon in which labor compensation growth outpaces productivity growth leading to an increase in prices. The primary contribution of this research is two-fold. First, the findings around price-drivers of the highway construction provide policymakers with new information that may help them evaluate opportunities to mitigate price growth. Second, the proposed methodology to improve productivity and price growth measures motivate economists and statistical agencies to reconsider their current approaches. The new insights around productivity growth in the sector also motivate further investigation of the topic by the research community.

View record

The management of sewage sludge is a major global issue due to the presence of contaminants in the sludge, such as heavy metals and polybrominated diphenyl ethers (PBDEs), which are harmful to human health and the environment. Due to these concerns, this study aims to create a decision-support tool for municipalities when evaluating alternative sludge treatment techniques. The environmental and economic impacts of four common treatment techniques (anaerobic digestion, incineration, composting and pyrolysis) and three end-of-life uses (landfill, agricultural application and energy recovery) are evaluated by the use of life cycle assessment (LCA) and life cycle costs analysis (LCCA). In order to deliver credible results, the uncertainties inherent in LCA and LCCA are assessed via probabilistic approaches. The global warming potential (GWP) for each scenario is studied by using the LCA method. The results demonstrate that pyrolysis has the lowest (deterministic) GWP after capturing environmental credits due to energy recovery and fertilizer substitution. Incineration is the worst option in terms of GWP, primarily due to the greenhouse gas (GHG) emissions from the process. The findings from the probabilistic analysis indicate that pyrolysis process and agricultural application of anaerobically digested sludge can achieve net negative GHG emissions under some circumstances. The economic assessment shows that composting has the lowest life cycle costs among these studied technologies due to its low capital investment costs. Incineration is the least preferred alternative due to its high waste management and transportation costs. The results also indicate that capital costs are the most dominant contributor to life cycle costs across all technologies. Pyrolysis process can generate more profits compared to the other alternatives given that valuable resources, such as energy, fertilizer and fuel, can be recovered from the process. Overall, by considering both environmental impacts and economic costs, this study suggests that pyrolysis is the most environmentally optimal and economically affordable sewage sludge treatment method due to its low life cycle costs and desirable performance in terms of GWP. The incineration process is the worst option since it is the most expensive option and has the highest GHG emissions among these considered treatment processes.

View record

Transportation agencies have limited fiscal resources to manage their pavement infrastructure. Planning for the future includes uncertainty, such as the uncertainty of future traffic levels, cost of rehabilitation actions, price indices, among others. Deterioration modeling also includes uncertainty, such as random and measurement uncertainty. Failing to consider these uncertainties may lead to sub-optimal management policies that are unable to adapt to the future. Thus, the objective of this thesis is to develop a reinforcement learning algorithm to manage pavement systems at the project-level that minimizes the life-cycle cost.The deterioration model developed uses an iterative-methods approach to estimate infrastructure performance models based on sampling theory. The model addresses the issue around measurement uncertainty underlying infrastructure condition assessments for continuous distress indicators and its effect on the parametric models underlying decision-support tools. Through a case study of pavement roughness data collected as part of Federal Highway Administration’s long-term pavement performance program, the new approach reduces the unexplained variance that would typically enter decision-support tools by 14%. It also addresses concerns around heteroscedasticity surrounding conventional methods, allowing modelers to recover efficiency in their statistical estimates. Finally, the Q-learning algorithm with an ε-greedy policy efficiently learns an optimal management policy for infrastructure assets while simultaneously incorporating several sources of uncertainty. An important advantage of this approach is that it is model-free and non-parametric, imposing no restrictions on the structure of the uncertain inputs. This study subsequently implements the Q-learning approach across three separate case studies. The proposed algorithm leads to the selection of a management policy that, on average, reduces expected life-cycle costs between 3% and 15% compared to traditional infrastructure management approaches. This research contributes to the pavement management literature by creating improved performance models and providing a holistic view of uncertainties in the management process. There are several opportunities to expand upon this research which are discussed.

View record