05_conclusion.qmd

# Conclusions {#sec-conclusions}

This paper contributes to the existing body of knowledge on IMT effectiveness and traffic incident modeling. It highlights the importance of strategic IMT deployment and the necessity for ongoing evaluation and adaptation of these strategies to meet changing demands in highway operational systems. Our study focuses on the Utah Wasatch Front Region, employing simulation modeling to evaluate IMT effectiveness. Previous studies have established IMT's significance in mitigating traffic disruptions, such as incidents and vehicle breakdowns [@bennett2022; @hadfield2021; @schultz2023; @skabardonis1998]. Furthermore, mesoscopic DTA modeling has proven beneficial for in-depth analyses of incident impacts and management strategies [@kaddoura2018; @li2020; sisiopiku2007].

However, there is a noticeable lack of large-scale simulation models applied to evaluating IMT performance. These models offer unique insights and enable scenario testing that may be challenging with traditional methods. To address this gap, we developed a specific simulation model for assessing IMT operations across Utah, seeking to improve our understanding and the effectiveness of these teams.

We used our model to analyze VHD and various IMT performance measures across different incident types and IMT configurations. Our analysis primarily focused on comparing the outcomes based on the number of IMT deployed in each scenario. The results indicated that the existing fleet of 20 IMTs deployed by UDOT effectively respond to incidents and reduce delay. In comparison to scenarios with incidents alone, a 20 IMT configuration reduced highway delays by 18.2%, resulting in an average VHD reduction of 4232 hours per simulated day. Increasing the fleet size to 30 IMTs led to a reduction in delay of 22.9% and an average VHD savings of 5334 hours.

Additionally, in our simulation, a fleet of 30 IMTs showed quicker response times to incidents compared to a fleet of 20 IMTs, with average response times decreasing from 15.0 minutes to 11.0 minutes. Notably, the 15.0-minute average arrival time for the 20 IMT scenarios aligns closely with the actual median arrival time recorded in Hyer's 2023 study on Utah IMT performance [@schultz2023]. Additionally, increasing the IMT fleet to 30 resulted in a cumulative reduction of 28 hours in travel time across 20 scenarios. This translates to an average decrease of 1.4 hours in IMT travel time per simulated day.

## Implications {#sec-implications}

In their study on Utah's IMT performance, @hadfield2021 found that a one-minute delay in IMT response time led to a 0.8-minute increase in RCT, an additional 34.6 minutes in total estimated travel time, and an increase of \$925 in EUC. Building on these findings, we estimate that reducing the IMT response time by an average of 4 minutes --- a potential outcome of adding 10 more IMTs to the fleet --- could result in EUC savings of approximately \$3,700 per incident.

As UDOT evaluates the effectiveness of their IMT program and contemplates potential expansion, this study's insights become a crucial asset for informed decision-making. Our [results](#sec-results) suggest that increasing the IMT fleet has the potential to decrease VHD and enhance the speed of incident response.

The model we developed for this project addresses two significant 'what-if' scenarios, applicable to UDOT or any other transportation agency aiming to optimize their IMT program:

1.  What if the incidence of traffic disruptions in our region surges? How might this impact our incident management program's effectiveness?
2.  What if we decide to increase the size of our IMT fleet? What are the potential advantages and potential challenges associated with this expansion?

While our primary objective was to address these scenarios, the model's versatility enables future exploration of additional questions, such as:

-   Identifying the optimal deployment locations for IMTs in our system.
-   Assessing whether the current operational hours of the IMT program are optimal.

## Limitations {#sec-limitations}

The limitations of this study are mainly attributed to the model we developed and the aspects that can be enhanced. As outlined in the [incident modeling](#sec-inc_modeling) section, we opted for MATSim to simulate IMT performance due to its ability to handle large-scale networks, incorporate real-world data, and mimic authentic driver behavior. While MATSim can realistically portray driver behavior, employing additional methods could have further enhanced its performance in our simulations.

In the [scoring and replanning](#sec-MATSim_Score) section, we describe how agents in MATSim alter their routes through an iterative scoring process influenced by various factors including late arrival, mode choice, and activity type. Although this method is generally effective, it may present challenges when simulating unforeseen events like traffic incidents.

As @dobler2012 highlight in their chapter of the MATSim manual, employing a within-day replanning tool is helpful within the MATSim framework, especially for handling unexpected situations. They note that while the iterative modeling approach of MATSim is adept at reaching user equilibrium under typical conditions, it tends to falter during sudden events. This can result in seemingly irrational behaviors, such as agents changing routes before an incident actually takes place.

The scenario depicted in Figure \ref{fig-within-day} serves as an illustrative example of the potential routing issues that can arise in MATSim when within-day replanning is not applied. In this instance, an agent is navigating from a start point marked by the red dot to a destination marked by the green dot. At 14:02, a traffic incident disrupts the agent's original route. Nevertheless, due to the anticipatory nature of MATSim's iterative approach, the agent already reroutes at 14:00---2 minutes before the incident occurs. This early route change showcases the challenges of solely relying on an iterative modeling strategy for responding to sudden events. It also emphasizes the benefit of incorporating a within-day replanning feature, which relies on a single iteration to provide more accurate and realistic navigational adjustments.

```{=latex}
\begin{figure}
\centering
\includegraphics{figures/within_day.png}
\caption[Within-day replanning approach for a MATSim routing problem.]{Within-day replanning approach for a MATSim routing problem (Dobler et al., 2012).}
\label{fig-within-day}
\end{figure}
```

Unfortunately, the problem showcased in Figure \ref{fig-within-day} surfaced at times throughout our MATSim simulations. In certain scenarios, agents would have preemptively avoided specific road links anticipated to have incidents, rather than adhering to more realistic behavior patterns observed on real roads. Typically, drivers maintain their intended course until directly confronted with congestion or delays, at which point they may choose to reroute. This variance between the simulated and actual driver responses to unforeseen road incidents underscores a key area for enhancement in our simulation model, ensuring it more accurately reflects realistic driving behavior.

## Next Steps {#sec-next_steps}

Moving forward, enhancing the simulation of IMT response and performance could be achieved by implementing the adjustments mentioned in the [limitations](#sec-limitations), specifically concerning the replanning and configuration settings of the MATSim model. Future steps could also involve exploring additional methods for simulated analysis of IMT performance measures. Depending on the specific requirements of UDOT or other transportation agencies, the simulation could be modified to evaluate optimal IMT starting locations, hours of operation, or their overall cost-benefit ratio.

Overall, the simulation we developed largely corroborates previous findings regarding the efficacy of IMTs. While it necessitates certain adjustments to enhance its reliability, it demonstrates potential as a tool for examining IMT performance in ways that previous studies have not.