-
Notifications
You must be signed in to change notification settings - Fork 1
/
01-intro.Rmd
70 lines (49 loc) · 13.2 KB
/
01-intro.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
```{r, include = FALSE}
# Outline
#Here is my outline for the entire thesis. Hopefully this will give me a rough idea of what to write about.
# 1. Introduction *(5pg)*
# a. First/Last mile transit
# 2. Literature Review *(20-30pg)*
# a. Discus Mode Choice (4-step model, ABMs, microsimulations)
# b. Discuss ActivitySim (modechoice)
# c. Discuss MATSim (modechoice)
# d. Discuss BEAM (modechoice)
# e. Figure explaining tour type and first trip difference?
# f. Variables showing path,person,location?
# 3. Methods
# a. What did we add to BEAM?
# b. Algorithms
# c. How did we make scenario?
# d. What scenarios did we run? SLC + SL / Data & Calibration
# e. What tests did we run? (path, person, location effects)
# 4. Results
# a. Describe tests and findings
# 5. Discussion
# a. What does it mean?
# b. Limitations --> Future research recommendation
# 6. Conclusions
# a. Wider implications
# b. First mile last mile
# c. climate change lol
```
# Introduction
## Problem Statement
Public transit is a sustainable mode of transportation that helps improve the overall quality of life. It's sustainability is shown in the effects it has on the environment, on the economy, and with the community. Environmentally, public transit helps metropolitan areas meet national ambient air quality standards (NAAQS) by lowering overall vehicle emissions and pollutants [@fta22]. Public transit also helps conserve land and reduce the need to construct new roadways. This helps decrease travel demand and degraded water runoff. Fuel use and greenhouse gas emissions are also lowered. Economically, public transit has shown to easily return more to the national economy than it costs. For example, in 1997 public transit returned $26 billion. The economic benefits seemed to originate from households that prefer not to drive, people who wanted to avoid congestion, people who lived near transit, a decrease in auto emissions, and an increase in property tax among others [@fta15]. Lastly, public transit also benefits the community by improving overall equity among individuals. Usually a community is composed of a wide range of people with varying socioeconomic status. Public transit gives opportunities to those people who otherwise would not have been able to have. Public transit can decrease travel time and costs, increase accessibility, improve property value, improve health, and increase employment [@venter18]. Overall, these environmental, economic, and social benefits make public transit seem like the be-all and end-all. Public transit may be sustainable, but it comes with its own limitations.
The limitations of public transit deal with problems relating to cost, time, and distance among other reasons. For example, an overpriced transit ticket can deter travelers from using public transit and vice versa. According to @todd, transit ridership has an average elasticity with respect to fares of -0.4, meaning that for every 1.0% increase in fare, transit ridership will decrease by 0.4%. Transit ridership depends heavily on correct pricing. Along with fares, travel time and travel distance are also significant limitations. The overall travel time when using transit is often longer than other modes of travel. Longer travel times often push away potential transit users. In addition, since transit uses fixed routes, the distance to and from transit locations can be hard to cover. This extra distance added to the trip can increase travel times and travel costs. Trips to and from transit locations are referred to as "first mile" and "last mile" trips although they can vary in length and mode. The success of transit depends heavily on providing sufficient access and making first mile and last mile trips manageable for as many passengers as possible. Travel time, distance, and cost are significant limitations for public transit and they may have contributed to the recent transit ridership decline.
According to @greg, the most recent steepest decline in transit ridership occurred between the years 2015 to 2018. A variety of factors may have caused this decline, among them being the introduction of bike share and Transportation Network Companies (TNCs). Bike share is a shared transportation service where bikes are available to share among individuals at a small price. TNCs are ridesharing companies that match passengers to drivers with vehicles to provide a ride at a price (e.i. Uber or Lyft). @greg estimate that the introduction of bike share has caused a decrease in bus ridership of 1.8% and the introduction of TNCs will decrease heavy rail ridership by 1.3%. This decline in transit ridership is not sustainable and if it continues, will cause economic and environmental problems across the United States. Therefore, reversing the decline of public transit ridership is essential to creating a more sustainable future.
In 2020, a project from the Transit - Serving Communities Optimally, Responsively, and Efficiently (T-SCORE) Center was proposed with the goal of keeping transit sustainable and resilient into the future. Along with keeping transit sustainable, T-SCORE hoped to discover how to increase transit ridership. As a result T-SCORE was divided up into two tracks: the community analysis track and the multi-modal optimization and simulation (MMOS) analysis track. The community analysis track aimed to statistically determine why ridership was declining and identify possible solutions. The MMOS analysis track had the purpose of using modeling techniques to determine the effectiveness of on-demand transit vehicles. On-demand transit vehicles serve as ridehailing vehicles with the sole purpose of bringing users to and from transit locations (i.e. using Uber to get to the train). The hypothesis was that instead of having ridership of public transit compete with TNCs, to instead pair together ride hail and transit services. If put at a fair price, on-demand transit vehicles could possibly help many users minimize travel time and cost of the first-mile and last-mile trips while also increasing overall transit ridership.
Modeling on-demand transit vehicles was an effective avenue toward understanding their impact on transit ridership. Figure \@ref(fig:mmos) shows the entire modeling and optimization process used in the MMOS track to analyze on-demand transit vehicles. The inputs are represented in green whereas the outputs are represented in yellow (with the exception of the mode choice model). Since the research revolved around understanding the impact of a new mode option, using an accurate mode choice model was essentail. Mode choice models are used to analyze and predict which modes of transportation individuals will use on their trips. Therefore, the mode choice model would determine which individuals would choose to use on-demand transit vehicles. A problem arose however when we noticed that the different modeling frameworks within the MMOS track each used a different mode choice model!
```{r mmos, fig.cap='Overview of the T-SCORE MMOS tract process.', out.width='75%', fig.asp=1, fig.align='center', echo=FALSE}
knitr::include_graphics("pics/mmos_mode.png")
```
Figure \@ref(fig:mmos) displays in blue all the parts that use its own mode choice model. First, the activity-based model used to create the daily activities and trips uses its own mode choice model. Then the multi-agent simulation has a different mode choice model as well as the multi-modal optimization having its own mode choice. We wondered if using three separate mode choice models would have an adverse effect in modeling the on-demand transit vehicles. Therefore, in an effort to minimize variability between models in predicting the usage of on-demand transit vehicles, a consistent mode choice structure was designed.
## Purpose of Research
The main objective of this research was to create a consistent mode choice structure across multiple model frameworks. Although the T-SCORE project presented three frameworks each with their own mode choice model, this research focuses on the effects of a consistent mode choice model between two frameworks: an activity-based model and a microsimulation tool. This consistency is needed because the mode choice models within activity-based models and microsimulation tools are vastly different. For example, activity-based models often use a variety of utility parameters to predict mode choices whereas microsimulation tools use less. In addition, microsimulation tools are better at predicting ridehailing services more realistically. Usually, microsimulation tools use varied wait times and varied vehicle availability whereas activity-based models use average wait times and constant vehicle availability. Overall, some key differences do exist between these two modeling frameworks, so creating a consistent mode choice model could result in constant predictions of new transportation modes between frameworks.
In order to determine if a consistent mode choice model between frameworks results in more consistent predictions, three measures of effectiveness were completed. First, a comparison of final modal distributions were analyzed: the final distribution using inconsistent modeling frameworks was compared with the final distribution using consistent modeling frameworks. The second measure of effectiveness analyzed the importance of using multiple utility parameters in the mode choice function. Specifically, the weight of importance of each path, person, and location type variables were described. Lastly, a comparison was done showing how mode choice was modeled at the sub tour trip level between modeling frameworks. With these three measures of effectiveness, the importance of a consistent mode choice model will be discussed.
Two important software were used to create consistent mode choice structures. ActivitySim was used as the activity-based model and BEAM was used as the microsimulation tool. ActivitySim is an open-source advanced activity-based travel model maintained by a group of Metropolitan Planning Organizations (MPOs), Department of Transportations (DOTs), and other transportation agencies [@asim]. Its primary purpose it to use a synthetic population to generate daily activity patterns (DAPs) for each individual. Using ActivitySim and a synthetic population, a set of DAPs for agents in the Salt Lake City, Utah region were generated. (The Salt Lake City, Utah region was the location of the test scenario used in this research.) BEAM, standing for Behavior, Energy, Autonomy, and Mobility, is an agent-based microsimulation tool developed by Lawrence Berkeley National Laboratory and the UC Berkeley Institute for Transportation studies [@beam]. BEAM's primary purpose is to model the DAPs of individual angents within a region to create a model that accurately displays the region's current transportation system. BEAM has the capability of accomplishing within-day planning where agents are able to dynamically respond to new transportation services. Using BEAM, the DAPs generated from ActivitySim for the Salt Lake City, Utah region were modeled.
This project's main objective was carried out by changing BEAM's default mode choice model to more closely align with that of ActivitySim's. Then, the results generated by BEAM were compared to those by ActivitySim to determine the effects of mode choice consistency. The various differences that exist between ActivitySim's and BEAM's mode choice models are described in Chapter 3 and the outline of the entire research project is described in Section \@ref(outline).
## Outline of Research {#outline}
Chapter 1 introduces the objectives and motives behind the research. This chapter also gives the outline to the thesis. In Chapter 2 a literature review on mode choice modeling is conducted. Chapter 2 begins by explaining mode choice in the four step model as well as giving a brief history of mode choice in discrete choice modeling. Afterwards, the broad range of mode choice models used in activity-based models are explained. Then, the mode choice models of ActivitySim, MATSim, and BEAM are described in detail.
Chapter 3 describes the methods that were taken in order to answer the research questions presented. First, how the test scenario was created is explained. This also involves understanding the set up of ActivitySim and BEAM for the Salt Lake City, Utah region. After explaining the setup process, the methods toward creating an improved mode choice model in BEAM are described. An explanation on data validation and calibration follows.
Chapter 4 describes the results of having a consistent mode choice model within the BEAM model. Specifically, this chapter explains three tests used to verify the effects of mode choice consistency. The first test compares total modal distributions between ActivitySim and BEAM with and without a consistnent mode choice model. The second test aims to show the effects of different types of mode choice utility parameters. The third test shows the differences in subtour mode choice between the BEAM and ActivitySim models.
Chapter 5 has the purpose of reviewing the results and determining the effects of the research as a whole. It also describes what the next steps moving forward are as well as the limitations in this research. A conclusion is then given in Chapter 6 where the wider implications are explained and the first-mile/last-mile problem is restated.