Federal infrastructure investment has dramatically declined over the past decades, dropping from 30 percent of total spending in 1960 to merely 12 percent in 2018, according to recent data.
Despite the nation's deteriorating infrastructure requiring increased funding to achieve optimal performance, groundbreaking MIT research demonstrates that AI-powered infrastructure sustainability solutions can create more resilient and efficient roadways even within current budget constraints.
The study, conducted by a team of MIT Concrete Sustainability Hub (MIT CSHub) researchers and published in Transportation Research D, reveals that implementing innovative planning strategies enhanced by machine learning models for road maintenance optimization could significantly improve environmental and performance outcomes for pavement networks without additional funding.
The research introduces a cutting-edge budget allocation tool that works in tandem with three revolutionary strategies for managing pavement networks: utilizing diverse paving materials, implementing both short-term and long-term paving solutions, and extending the evaluation period for these interventions.
This artificial intelligence in transportation planning approach delivers substantial advantages. When applied to a 30-year case study of Iowa's U.S. Route network, the MIT CSHub model and management strategies reduced emissions by 20 percent while maintaining current road quality standards. Achieving comparable results with traditional planning methods would require a 32 percent budget increase for the state. The success hinges on addressing a fundamental yet challenging aspect of pavement asset management: uncertainty.
Navigating Unpredictability with Predictive Analytics
The typical roadway must endure for decades while supporting hundreds of thousands or even millions of vehicles. Throughout its lifespan, numerous variables can shift: material costs may fluctuate, budgets may tighten, traffic volumes may increase, and climate change impacts may necessitate unexpected repairs.
Effectively managing these uncertainties requires long-term forecasting and anticipating potential changes—a perfect application for sustainable pavement management with AI technology.
"Incorporating uncertainty impacts is crucial for making effective paving decisions," explains Fengdi Guo, the paper's lead author and former CSHub research assistant.
"However, measuring and correlating these uncertainties with outcomes is computationally intensive and costly. Consequently, many DOTs [departments of transportation] must simplify their analyses for maintenance planning, often resulting in suboptimal expenditures and results."
To provide DOTs with accessible tools for integrating uncertainty into their planning processes, CSHub researchers have developed a streamlined approach that offers greater precision and incorporates several innovative pavement management strategies.
The planning methodology, known as Probabilistic Treatment Path Dependence (PTPD), leverages machine learning and was created by Guo specifically for predictive analytics for infrastructure budget allocation.
"Our PTPD model consists of four sequential steps," he explains. "These include pavement damage prediction; treatment cost forecasting; budget allocation; and pavement network condition evaluation."
The model begins by analyzing every segment within a pavement network and forecasting potential future scenarios for pavement deterioration, costs, and traffic patterns.
"We conduct thousands of simulations for each network segment to determine probable cost and performance outcomes for each initial and subsequent sequence, or 'path,' of treatment actions," says Guo. "The treatment paths demonstrating optimal cost and performance outcomes are selected for individual segments and then implemented across the entire network."
The PTPD model aims to minimize costs not only for agencies but also for users—primarily drivers—who experience excess fuel consumption due to poor road conditions.
"One enhancement in our analysis is the integration of electric vehicle adoption into our cost and environmental impact projections," notes Randolph Kirchain, a principal research scientist at MIT CSHub and MIT Materials Research Laboratory (MRL) and a co-author of the paper. "Since the vehicle fleet will transform over the coming decades with increased electric vehicle adoption, we ensured our methodology accounts for how these changes might affect our predictions of excess energy consumption."
After developing the PTPD model, Guo sought to evaluate how different pavement management strategies might perform in comparison. To accomplish this, he created an advanced deterioration prediction model.
A distinctive feature of this deterioration model is its simultaneous handling of multiple deterioration metrics. Using a multi-output neural network—an artificial intelligence tool—the model can predict various forms of pavement deterioration concurrently, thereby accounting for their interrelationships.
The MIT team selected two key metrics to assess the effectiveness of different treatment paths: pavement quality and greenhouse gas emissions. These metrics were then calculated for all pavement segments within the Iowa network.
Enhancing Performance Through Strategic Variation
While the MIT model can assist DOTs in making superior decisions, the decision-making process is ultimately constrained by the options under consideration.
Guo and his colleagues therefore aimed to expand current decision-making frameworks by exploring a comprehensive set of network management strategies and evaluating them using their PTPD approach. Based on this evaluation, the team determined that networks achieved optimal outcomes when the management strategy incorporated diverse paving materials, a combination of short- and long-term paving repair actions, and extended timeframes for paving decisions.
They then compared this proposed approach with a baseline management strategy reflecting current widespread practices: using exclusively asphalt materials, short-term treatments, and a five-year evaluation period for paving action outcomes.
With these two approaches established, the team used them to plan 30 years of maintenance across Iowa's U.S. Route network. They subsequently measured the resulting road quality and emissions.
Their case study revealed that the MIT approach offered significant benefits. Pavement-related greenhouse gas emissions decreased by approximately 20 percent across the network throughout the entire period. Pavement performance also improved. To achieve the same level of road quality as the MIT approach, the baseline method would require a 32 percent larger budget.
"It's important to note," says Guo, "that since conventional practices employ less effective allocation tools, the difference between them and the CSHub approach should be even more pronounced in real-world applications."
Much of the improvement stemmed from the precision of the CSHub planning model. However, the three treatment strategies also played a crucial role.
"We've found that combining asphalt and concrete paving materials enables DOTs to not only select materials most appropriate for specific projects but also mitigates the risk of material price fluctuations over time," explains Kirchain.
A similar principle applies to mixing paving actions. Employing a combination of short- and long-term solutions provides DOTs with the flexibility to choose the right intervention for the right project.
The final strategy, implementing a long-term evaluation period, allows DOTs to visualize the full scope of their choices. If decision consequences are predicted over only five years, many long-term implications remain unconsidered. Expanding the planning horizon can therefore introduce beneficial, long-term options.
The complexity of paving decisions—given their lasting impacts on the environment, driver safety, and budgets—is understandable. Rather than oversimplifying this challenging process, the CSHub method embraces its complexity. The result is an approach that equips DOTs with tools to accomplish more with limited resources.
This research was supported through the MIT Concrete Sustainability Hub by the Portland Cement Association and the Ready Mixed Concrete Research and Education Foundation.
