Engineering systems analysis of mobility in Odawara city: New transportation services impacts on community engagement

Local cities in Japan are struggling with aging and decreasing populations. Elderly people and parents with young children are uniquely challenged when accessing public transportation, a key to increasing their community engagement and improving local cities' sustainability. This research investigates the introduction of new transportation modes and fares on community engagement of both elderly people and parents of young children. An urban systems model which integrates mobility and civic functions is evaluated by agent‐based simulation to analyze various policies’ impacts on community engagement and financial performance. The model is applied to Odawara City, a typical local city of nearly 200,000 people in Japan. For this case study, two policies which strongly subsidize a community bus and partially subsidize a door‐to‐door van were predicted to generate 10% or greater engagement for the elderly without a financial loss compared to the current baseline case. More specifically, community engagement of elderly people and parents with young children are predicted to increase by 11% with positive (0.25 M yen) net present value per person when the fare of community buses is 100 yen and that of door‐to‐door vans is 300 yen. However, no synergistic effect driven by policies favoring elderly people and those favoring parents is found. Still, the measures to support elderly people's transportation accessibility do not harm the parents’ behavior but rather support their daily activities. The method is demonstrated to be useful for designing new mobility policy in light of a specific population with demographic residential distribution and existing transportation network.

To address population decrease, the Ministry of Internal Affairs and Communications (MIC) and local governments have introduced measures including marketing campaigns promoting visits and subsidized budgets for relocation to local cities.MIC subsidizes the budget to compensate for purchasing houses in a local city and supports involvement in the local community to promote retention of residence for up to ten years.At the same time, MIC has introduced step-by-step marketing approaches to attract people who are interested in local cities.For example, MIC exempts local specialties from taxes, supports engagement in local festivals, and promotes dual residence. 1COVID-19 has also promoted the movement from large cities to local cities.
Many companies have introduced remote work policies in response, allowing employees to choose where they live more flexibly.Recent survey data 2 shows that people moving out from Tokyo increased by 4.7% and was the most significant increase since 2014.
The isolation of elderly citizens is another urgent issue in local cities in Japan.Kobayashi and Fukaya 3 find that, among men, complete isolation increased from about 2% in 1987 to 5% in 2012.Much research shows the importance of elderly people's social interactions and communication with others to their health condition.Holt-Lunstad et al. 4 find that social relationships are comparable with stopping smoking, obesity, and physical inactivity with regard to mortality risk.These factors increase the likelihood of survival by 50%.Anme et al. 5 investigate the relationship between social interaction and the mortality rate of elderly people by comparing the 7-year mortality rate through questionnaire results answered by those 65 years or older living in a rural area around a large city in Japan.They find that more significant social interaction is positively related to the lower mortality rate.Several practical programs and projects have been applied to overcome these social isolation issues by increasing elderly people's community engagement.Kanamori et al. 6 find that increased exercise frequency with others has essential health benefits regardless of the total frequency of exercise.Brady et al. 7 find that SilverSneakers, a national exercise program, decreases social isolation through increased physical activity and also decreases loneliness due to increased social connections.
Transportation is a key to increasing community engagement of elderly people.In fact, local cities in Japan have relied heavily on private cars for decades, even for elderly people.Japan's statistical data 8 shows that the average number of trips per day of 75 years of age or older who have driver's licenses has increased over 25 years and was 2.24 trips per day in 2015, while those without driver's licenses is only 1.19 trips per day.Those people with decrease in cognitive functions or physical ability are unable to drive, resulting in less frequent leaving the home and eventual social isolation.
Local municipalities in depopulated areas have introduced new transportation modes to improve access to public transportation.1][12][13] However, at the same time, investment and operation costs of the new transportation modes have been burdens for some municipalities. 11,14,15ese two main objectives are concurrently targeted: relocation to local cities for younger people and increased accessibility geared to elderly people.There has been little consideration of the synergy of these measures with one another.Regarding mobility options, parents with young children also have difficulties using public transportation as they pay attention to their children's behavior. 16Omori et al. 17 categorize their difficulties into six types of hurdles, including those using public transportation services.By analyzing their survey, Izumi et al. 18 identify that parents thus try to minimize their transport time.Even though those living in Tokyo, which has one of the highest public transportation service levels globally, take their children to kindergartens mainly by bicycles (45%) or by walking (46%), even though their car ownership ratio is high.
The ultimate objective of these measures is to achieve viability and sustainability of local cities, including citizens' active engagement.
Significantly, the introduction of new transportation modes has the potential to increase community engagement.Not only elderly people but all the citizens, including parents taking care of young children, can benefit from the new modes.Specific research questions addressed in this paper and approach are shown in Section 2. In Section 3, the system concept and the basic model in this research are described.In Section 4, the overview of the simulation model and function modules are shown, with some results of simulations and validations.In Section 5, the case study of Odawara City, a local city in Japan, is performed.Discussion and conclusion are shown in Section 6.

RESEARCH QUESTION AND APPROACH
This research aims to investigate the synergetic effect of introducing a new transportation system in a local city on both elderly people and parents raising young children in terms of their community engagement, thus contributing to the sustainability of the local city.
Since building social relationships is critical to a city's sustainability, in this research community engagement is defined as activity that enables social interactions with others in a common space, such as going to a playground or a public facility.
In order to identify the synergistic effect of new transportation services, this research aims to explore the impact of the new modes on the behavior of citizens, especially the community engagement of elderly people and parents taking care of their children.Additionally, the fares of the new transportation modes are a driver in citizens' transportation mode selection and the resulting financial performance of municipalities.The specific research question is as follows: What fares of newly-introduced transportation modes lead to increased engagement, lower cost, and equity, that is, balanced benefit between elderly people and parents?
Research has been conducted for many aspects of public transportation.Related to the investment and operation costs of the transportation network, the dial-a-ride and similar solutions have been explored to reduce the financial cost, cast from a mathematical optimization point of view 19 as a derivative of the traveling salesman problem.
Noda et al. 20 evaluate the scalability of the new on-demand bus transportation services using several demand models on a grid-based road structure.
Applying agent-based modeling and simulation (ABMS) to urban planning and transportation policy has been vigorously researched.

System concept
Figure 1 shows the concept of the system described with the Object-Process Methodology.There are two beneficiaries, elderly people and couples with children attending preschool.Their needs include location changes, such as travel to hospitals, kindergartens, and grocery stores.The system provides a transporting function to satisfy the needs of changing locations.The function's attributes include "safely" and "affordably," the essential and core needs of elderly people and parents taking care of their children for transportation.
Transporting's attributes for elderly people include "independently," and that of parents includes "less physical stress."According to Arita et al., 10 elderly people want to transport with "less physical stress" because of their physical weakness.They also want to transport "independently" since they tend to rely on their families or partners when going out because they cannot drive due to a decrease in cognitive ability.For parents taking care of their children, they want to go out with less physical stress given physical hurdles in using current public transportations, such as conveying strollers without elevators and crowding in trains and buses, according to the survey results by Ohmori et al. 30 From a local city's viewpoint, public transportation is the mechanism to meet the above needs.There have been many projects introducing community buses and/or door-to-door vans in Japan to help elderly people have accessibility to public transportation, especially in rural areas.MLIT also prepared policy guidance for the new transportation's introduction. 31In this research, community buses and door-to-door vans are considered to provide transporting functions.

Basic model
To investigate the synergetic effect, the basic model of the system consists of two major components, Population and Infrastructure, as shown in Figure 2. The population includes Elderly people and Parents who mainly take care of their children.Population's instance is a Person who wants to move from one location to another.The infrastructure consists of locations with activities as their attributes and transportation modes that connect the locations.
In the value path shown in the middle two columns in the figure, first, Person selects an activity that s/he wants based on his/her Activity Sensitivity.Second, according to the selected activity, s/he chooses a location from locations that can be reached by checking the transportation modes.Third, s/he determines which transportation modes she/he uses to move to the selected location based on Mode Preference.Finally, using the selected mode, s/he changes her/his location from the current location to the selected location.

Overview of simulation model
The overall simulation model is shown in Figure 3

Mode preference calculator
The second essential function is the distance-dependent transportation mode selection function.The transportation mode preference is generally dependent on the distance between two locations.When the distance is short, the preferred mode is walking or bicycle, and as the distance increases, train/bus is preferred.According to the statistical data by MLIT, 8 the shape of mode preference can be described using a sigmoid function, while its parameters depend on cities and types of population.In this research, the preference of each transportation mode is dependent on the distance between two locations and represented the dependence with the normalized logistic function When a new transportation mode is introduced, people tend to use the new mode depending on many factors, such as their accessibility and fares, compared with the other transportation modes.
Using a survey, Fukumoto et al. 11 researched the preference for the new community bus in Tahara City in Aichi Prefecture.They find that the preference to use it strongly depends on its fare.The preference to use the new transportation mode can be described using a sigmoid function from their survey.Then, in this research, the new transportation mode preference depends on its fare, and a sigmoid function describes the dependency.Concretely speaking, the preference is calculated by multiplying the conversion ratio from an existing mode to the new mode and the preference of the existing mode.For example, the conversion ratio, c t→c , from trains/buses to community buses, can be defined as follows: c t→c (f c ; f m, t→c , f s,t→c ) = c 0,t→c (1 − g(f c ; f m, t→c , f s,t→c )), because the preference to use community buses decreases as their fare increases.Then, the preference of the community buses p c is calculated as

Activity sensitivity calculator
Activity sensitivity is an attribute of a person in a population.It represents his/her probability of selecting to engage in an activity.
The sensitivity of each activity satisfies the following equation: 1 = ∑ i s(t, a i ) , for all t, where s(t, a i ) means the sensitivity, a i means an activity, and t means time.
It is considered that the activity sensitivity can be affected by the introduction of the new transportation modes.From the surveys of the introduction of the new community buses and door-to-door vans, such as Chatan Town in Okinawa Prefecture 13 and Bunkyo City in The simulator approach.The bold box shows the full simulator, and each thin-lined box inside the bold box shows its function within the simulation model.The upper-left five data are the input data, and the lower-right three data are the output data of the simulation model.
Tokyo, 12 it is found that people using community buses tend to go out more than before.To consider this effect, introducing new transportation modes increases the ratio of going out from the home to be engaged in any activity.It is considered that the increase in the activity sensitivity is proportional to the conversion ratio from existing transportation modes to the new transportation modes, as represented in the following equation: where s(t, a i ) represents the activity sensitivity when there are only existing transportation modes, s(t, a i ) the sensitivity when new transportation modes are introduced, p j the mode preference of a new transportation mode, D the distance between residential location and the location that provides the activity a i , and A the adaption rate.F is the normalization factor to keep the summation of the sensitivities as 1.

Agent-based simulator
The

Capacity calculator
Let us set n i→j (t k , t k+1 ) as the number of passengers moving from location i to location j using a new transportation mode, that is, community bus or door-to-door van, between time t k and time t k+1 .The number of vehicles necessary to transport these passengers, N i→j (t k , t k+1 ), should satisfy the following equation: Then, the total number of vehicles of a new mode, N v , is calculated by the following equation:

Activity engagement ratio and engagement index calculator
The activity engagement ratio, A, is the ratio of duration people are engaged in any activity except Residential to the total time, as calculated with the equation: where N is the number of people.This value can be calculated for each demographic.
The engagement index, EI, is calculated by the weighted summation of the number of hours engaging in an activity.It is represented by the following equation: where N is the number of people, T(a i ) the number of hours engaging in the activity a i , and w a i the weight of engagement of the activity.This index can also be calculated for each population type.

4.2.6
Cost and revenue calculator The cost of the new transportation modes consists of the capital expenditure and the operation cost, while the revenue comes from the fare paid by citizens using the new modes.The capital expenditure CE = , where r is the discount rate, usually 7% in the U.S, and g the growth rate or inflation rate, usually 2-3% in the U.S. 32

Simulation
The agent-based simulator made by Carroll and McDonough 22,23 is used to capture the daily behavior of elderly people and parents taking care of their children in a city.As the detailed descriptions of the simulator are shown in, 22,23 the software probabilistically simulates the decision-making process, shown in Figure 2, of an agent who moves from one location to another.First, all the agents in a city are allocated to their residential locations at 0:00 am.
Secondly, depending on his/her activity sensitivity, each agent selects an activity that s/he wants to engage.The activity sensitivity of each agent is calculated probabilistically by a sensitivity matrix defined in each population type.The sensitivity matrix has Interval, Activity Type, and Sensitivities, as the matrix in the Odawara City model is described in Table 8.Interval is a coupled start and end time to define the period.Activity Type describes what type of activity an agent is willing to engage in during the Interval.Sensitivities define an agent's preference for each activity and his/her time capacity to engage in it, that is, to what extent s/he can engage in it.The preference and capacity are defined with probability distributions initiated at the beginning of each simulation, according to the mean and standard deviation of each activity's preference and capacity.Note that the summation of the mean value of each activity's preference is normalized to 100%.
Thirdly, the agent selects a location s/he wants to engage in his/her selected activity.Since each location provides at least one activity, the agent searches his/her perceptive locations, including his/her current location and its adjacent locations, and s/he finds an appropriate one that satisfies his/her selected activity.
The last decision-making process of the agent is to select a transportation mode for moving from his/her current location to the selected location.As the selection process of modes is similar to that of activities, the agent's mode is probabilistically selected according to his/her transportation mode preference.The transportation mode preference is defined as a probability distribution and is also dependent on the distance between two locations.The summation of the mean value of each mode preference is normalized to 100%, and when limited transportation modes are offered between two locations, his/her likelihood of choosing an offered mode is renormalized to 100% proportional to the original mode preference.
Then, the agent goes to the location that satisfies his/her needed activity with his/her preferred transportation mode.When finishing the engagement to the activity, s/he attempts to select another one according to his/her preference and the engagement time within the time capacity.

Validation
Step-by-step validation was performed to develop confidence in four added functions implemented in the agent-based simulator.In each validation case, the number of agents is fixed to 1000, and the number of simulations is 10.The results are summarized in Table 1, where Exp shows the expected percentage and Avg shows the average number of agents per 1000 agents in percentage.As shown in Table 1, in every case, the average ratio of agents is close to the expected value, which confirms the validity of all four additional functions.
The detailed settings of each step are as follows.The first is to validate the mode selection function when there are only two spaces with no new mode.In 1-a, an agent randomly chooses a mode; that is, the preference of every mode is equal.In the case of 1-b, the preference is skewed, as shown in Table 1.The second step is to validate the distance dependence of the mode selection function.In 2-a, the distance between two locations is 300 m, and Walking/Bicycle dominates the other modes.In the 2-b case, the distance is 2500 m, and Private Car is the dominant transportation mode.
The third and fourth steps include door-to-door vans as a new transportation mode.The third step is to validate the fare sensitivity to the new transportation mode's preference.Two locations exist with 2500 m distance, same as the case 2-b.In the case of 3-a, the new mode's fare is 200 yen, relatively low compared with other transportation services, making agents prefer the new mode.In the case of 3-b, it is 700 yen, higher than others, making agents less preferable to the new mode.
The final step is to validate the increase in the activity sensitivity due to the new mode.There are also two locations with a 2500-meter zero as a base case.In 4-b, the increase is set to 0.5 as a low impact case, while the sensitivity increase is 2.0 in the case of 4-c as a high impact case.

Characteristics of Odawara City
Odawara City is a typical local city in Japan, with approximately 190,000 population in 2021 33 and six train lines, 34 shown in Figure 4.
Since the frequencies of these train lines in Odawara City vary, in some districts, citizens can use public transportation easily, while access is not easy in other districts.Of 6 districts in the city, Chuo, Tomizu Sakurai, and Kawa Tonanbu districts have high public transportation accessibility.Except for the low population density areas, citizens in these three districts can access Tokaido Main Line, Odawara Line, Daiyuzan Line, Hakone Tozan Railway, Tokaido Shinkansen, and bus services easily.
On the other hand, Kawa Touhokubu and Tachibana districts have lower accessibility than the above three districts.Bus services do not cover well some areas in the two districts.Also, although Gotemba Line is in the Kawa Touhoukbu district, it is not convenient for public transportation because its frequency is only once per hour during the daytime.Table 2 shows the utilization ratio of each transportation mode in the above five districts.It is found that, in Kawa Touhokubu district and Tachibana district, the utilization ratio of private cars is at least 10 points higher than that in Chuo, Tomizu Sakurai, and Kawa Tounanbu districts, regardless of the activities people want to be engaged in.
Note that the Kataura district has a much smaller population than the other districts and is not considered in this research scope.
The district has started an outing-support project to introduce new transportation services provided by volunteers who can get only gasoline costs because the bus and taxi companies do not have enough incentives to provide transportation services. 35 The average speed of each transportation mode is shown in Table 4.

Odawara City model
A community bus's capacity c c is set to 30 persons/vehicle, its averaged utilization rate R c is 0.3, and its average transport time  c is 1 hour.
A door-to-door van's capacity c v is 10 persons/vehicle, its averaged utilization rate R v is 0.5, and its average transport time  v is 0.5 h.
The parameters for cost consist of investment cost and operation cost, which are described in Table 5 with assumptions and reference data for each parameter.The discount rate r is set to 7%, and the growth rate 2% to calculate the net present value of the investment in the new transportation modes.
The activity engagement weights of Playground and Public Space are set to 1.0, where the main objectives include social interactions.
Those of Grocery Store and Shopping Mall, where there are some social interactions among visitors, are set to 0.5.Hospital, Kindergarten, and Office have zero weight of activity engagement because there is little social interaction with others living in the same community in those spaces.The parameters of the transportation mode preference and activity sensitivities are described in the next section with calibration.

Model calibration
Each

Community bus
Capital Expenditure C i,c = 6, 000, 000 [yen/vehicle] The average price of microbuses or small buses is around 6,000,000 yen. 40,41eration Cost C o,c = 20, 000 [yen/vehicle/day] The daily operation cost of a community bus is between 11,000 and 20,000, 42 and it is assumed that the cost is the maximum of this range.
Door-to-door van Capital Expenditure C i,c = 3, 000, 000 [yen/vehicle] The price of the HIACE van produced by Toyota, a typical van, is around 3,000,000 yen. 43eration Cost C o,c = 15, 000 [yen/vehicle/day] The daily operation cost of a community bus is between 11,000 and 20,000. 42so, the operation cost of a door-to-door van is lower than that of a community bus, 20,000 yen/vehicle/day.Then, it is assumed that the cost is in the middle of the range described by Matsuzaki.
stuff is considered lower than going to buy food.In the simulation results, the transportation mode utilization ratio to go to a Grocery Store is very close to the statistical data.This consistency matches the expectation that the main customers of the grocery stores are elderly people and parents who mainly take care of their children, and their transportation mode preference can dominate the others' mode preference in the statistical sense.On the other hand, the mode utilization ratio of walk/bicycle to go to the Shopping Mall is lower than the statistical data, while the ratio of trains/buses is close to the statistical one.This difference is consistent with the fact that elderly people tend not to walk or pedal but ride cars.
To go to a Hospital, the mode utilization ratio for walk/bicycle is significantly lower than the statistical data, while that of train/bus is similar.This can be caused by the fact that the site model in this research only includes the hospital located at a distance of 2 or 2.5 km from residential spaces, although, in reality, there are some small hospitals near the residential places or not so far from them.
More granular site models, including hospitals near residential spaces, would be needed to capture the mode utilization ratio for hospitals.
However, in this research, since the number of people going to a Hospital is not significantly large, it is assumed that the difference in the mode utilization ratio does not affect community engagement, and the effect of this mismatch on the output of the simulation would not be so significant.
There are discrepancies between the statistical data and the simulation results to go to an Office.Since the statistical data include workers and students who cannot drive cars in general, the statistical data should be skewed to a higher Walk/Bicycle and Train/Bus ratios.However, the commuting modes include only Private Car and Train/Bus, and the engagement weight of Office is zero, so it is considered that these discrepancies do not affect the output of the simulation.

TA B L E 7
The transportation mode utilization ratio to go to each place in the simulation results compared with the statistical data of Odawara City.

Simulation results and Tradespace analysis
The simulation results for elderly people only, parents only, and both population types are shown, respectively, to identify whether there is a synergetic effect.After these results are compared, the impact of the adoption ratio of each population type is investigated by comparing the simulation results with different adoption ratios.Ten combinations of the fares of community buses and door-to-door vans are investigated, which are shown in Table 11.Note that each combination only includes fares which can be paid with a small number of coins, which is important especially for elderly people because they still tend to pay for cash.In Table 11, "Ideal" means the fares of two modes are both zero, which achieves the highest engagement ratio because the mode preference of each new mode is maximized, and the increasing ratio of the activity sensitivity is maximized."Base" means that the situation is exactly the same as the base case because no one uses any new mode when its fare is beyond acceptable.Also, CnVm means that Community Bus's fare is 100*n yen, and Door-to-Door Van's fare is 100*m yen; for example, C3V5 means that Community Bus's fare is 100*3 = 300 yen and Door-to-Door Van's fare is 100*5 = 500 yen.
Since Door-to-Door Van is more convenient than Community Bus, it is

TA B L E 8
The calibrated activity sensitivity for elderly people and parents in Central District.18:00-24:00

Playground
The sensitivity for each period is normalized to 1, and the time capacity (duration, described as D) is in hours.The standard deviation of each time capacity for each person in simulations is set to a quarter of the time capacity itself.Each activity sensitivity in each period for those in Suburban District is only 86% of the above value, while the time capacity is the same.

TA B L E 9
The calibrated transportation mode preference of each population type.(*1): Fukumoto et al. 11 found that 50% are willing to use it when the fare is 200 yen, and the ratio decreases with the fare increases.(*2): It is assumed that those who use private cars are willing to pay more than those who use trains and/or buses do, that is, f m,p→c > f m,t→c = 200 yen.Also, f m,p→c is lower than half of the daily cost of a private car, 700 yen per day, 45 because a round-trip requires twice its fare, that is, f m,p→c < 350 yen.(*3): It is expected that c t→v > c t→c because door-to-door van is faster and has higher accessibility than community bus.(*4): The average fare of door-to-door vans introduced in rural areas is 300 yen. 15(*5): This value is assumed to be lower than that for Elderly People because parents weigh more on shorter travel times, and community buses have lower speeds than trains/buses.(*6): It is assumed that Parents can afford more fares than Elderly People because the average income of Parents is higher than that of Elderly people.(*7): These are assumed to be smaller than those for Elderly People because Parents weigh more on shorter travel time.Also, these values are assumed to be larger than those of Community Bus because Door-to-Door Van is faster and has higher accessibility than Community Bus.

Mode Parameters
not analyzed when Community Bus's fare is larger than Door-to-Door Van's.
The adoption ratio of each population type can depend only on the population type and the usual trips per day.It is assumed that elderly people are more willing to go out by using the new transportation modes than parents because elderly people have more free time than parents do.Also, it is assumed that Elderly People in Suburban District In the case E_Ideal, the fares are zero, the NPV per person is negative and minimized, while the engagement index is maximized.As fares increase, NPV per person increases and becomes positive on average when the fare of Door-to-Door Van is at least 300 yen.On the other hand, the engagement index decreases as fares increase.
The negative NPV in the ideal case is evident because the revenue from the new modes is zero since the fares are zero.As the fares increase, the revenue increases, and NPV becomes positive when the revenue exceeds the operation cost and compensates the The combination of fares of the new transportation modes.The behavior of the engagement index in the simulations with parents only is different from that in simulations with elderly people.The average engagement index keeps between 0.45 and 0.47, and its range also keeps from 0.39 to 0.53 even as the fares increase.

Fare of community bus (yen)
There are three cases on the pareto-front; P _C1V3, P_C1V5, and P_C3V5.In these three cases, only P_C3V5 has a slightly positive average of NPV per person, while the other two cases have a negative average of NPV per person.However, the variance of P_C3V5 is still significant, and NPV is negative in some simulations.It means that P_C1V3 would be financially more preferable than the other cases, but its advantage is small.Also, from the viewpoint of community engagement, no case has the 10% or more increased level from the base case, although some simulation results surpass the dotted blue line.Then, when the target population is only parents taking care of their children, for each simulation result of all the population types, where the prefix "B_" means the target population is both elderly people and parents.
The resulting NPV per person and the engagement index is almost the same as one with elderly people only.In the ideal case, B_Ideal, the NPV per person is negative and minimized, while the engagement index is maximized.As fares increase, NPV per person increases and becomes positive on average when Door-to-Door Van's fare is at least 300 yen.
On the other hand, the engagement index decreases as fares increase.
There are five cases on the pareto-front; B_ideal, B_C1V1, B_C1V3, B_C3V3, and B_C3V5.In these five cases, all the simulation results of The impact of new transportation modes on the engagement index and NPV per person of all the population types, where "B_" means the target population is both Elderly People and Parents.Same as Figure 6, the large-filled circles and the small points show the averaged value and the simulation results, respectively.The star means the utopia point of this tradespace, and the solid black lines are the pareto-front lines.The dotted black line shows the 10% increased engagement index compared with the base case (B_Base).line on average.Therefore, taking the economics and the community engagement into account, B_C1V3 is the only preferable choice.Note that, the engagement index of B_C1V3 is 0.558, 94% of B_Ideal (0.624), and 111% of B_Base (0.529).
To investigate the synergetic effect of the new modes on the elderly people and parents, Figure 9 shows the comparison of the simulation results that only elderly people are considered, those that only parents are considered, and those that both are considered.It is found that the simulation results that include only elderly people dominate the tradespace over the other cases.This means that the performance of the new modes is maximized when only elderly people are considered as the beneficiaries of the new modes.
Three factors can explain that elderly people's behavior is the dominant factor of this model.The first is that elderly people exist about four times more than parents.The number of elderly people in Odawara City is 25,177, while families with small children are 6412.
The second factor is that the adoption ratio of elderly people is larger than that of parents because elderly people have more free time than parents do.Elderly people are more willing to visit places to be engaged in activities than parents, which increases elderly people's engagement index.The third factor is that parents' conversion ratio to the new transportation modes is set lower than elderly people's, although However, it is also considered that there is a benefit for parents to consider when the beneficiaries of the new modes are parents and elderly people.Based on the simulation results, when the beneficiary is only parents, the new transportation modes would not be introduced because the engagement index has not reached the 10% increased level from the base case and the NPV per person is not so high even in P_C3V5.However, when parents and elderly people are considered beneficiaries of the new transportation modes, there is a case, B_C1V3, that the engagement index exceeds the 10% increased level from the base case NPV per person is positive.
The effect of the adoption ratio on the new transportation modes' performance is investigated by focusing on the cases in which both  shown above.By comparing the same combination of fares with different adoption ratios, such as B_C1V1, BH_C1V1, and BZ_C1V1, the engagement index decreases as the adoption ratio decreases.When the adoption ratio is zero, the engagement index ranges between 0.52 and 0.54.This phenomenon can be understood that when the ratio zero, even though people are willing to use the new modes, the increase in their activity sensitivities is zero, and there is no increase in the engagement index.
Note that there is only one case, B_C1V3, with a positive NPV per person and increases the engagement index by at least 10% more than the base case.It is also revealed that the new modes increase parents' community engagement level, but their performance is only at most 3%, much smaller than that of elderly people.Two factors are considered to explain this small impact on parents' community engagement.The first is that parents have less free time to spare in community engagement, and the second is that the new modes cause longer travel time because their speeds are in general slower than the existing trains/buses and/or private cars, though the core needs of the parents include shorter travel time.
To address the main research question, "What fares of newlyintroduced transportation modes lead to increased engagement, lower cost, and equity, that is, balancing between elderly people and parents?" the synergetic effect of the two modes is investigated by simultaneously simulating the behavior of elderly people and parents.The simulation results show that the new modes increase the total community engagement of both elderly people and parents by at most 18%, and there is one preferable combination of fares in terms of both the community engagement and the financial sustainability.
It is also shown that there is no synergetic effect on the behavior of elderly people and parents.In the tradespace analysis, the cases where the beneficiary is only the elderly people dominate the other cases.On the other hand, the results can be interpreted as that the measures to support elderly people's public transportation accessibility do not harm the parents' behavior but mildly support their daily activities.
The community engagement index can be used to measure the performance of introducing new modes and those of the other related measures, such as attractive events and projects for elderly people and/or parents.Furthermore, since the ideal case shows the upper bound of the community engagement index, excessive marketing of the attractive events can be eliminated by carefully monitoring the adoption ratio and the theoretical upper bounds of the community engagement based on this simulation model.
The unique combination of population demographics, existing mobility options, infrastructure topology, and public venues lead to specific systemic outcomes and thus unique policy design depending on the location.The research results can be generalized that some transportation policies focusing on a specific population might not perform well (in this case for parents), but these policies might perform well if the beneficiaries include other population types.Some transportation measures are focused on a particular demographic in Japan, often elderly people, because their needs are urgent.For example, many municipalities introduce a transportation fares discount program for elderly people since their income is lower than the other generations, though these programs often cause financial deficits.An immediate response to the urgent needs is required from the viewpoint of municipalities' responsibilities, yet, based on this research, the policies for a newly introduced transportation measures focusing on elderly people might not perform well or not capture all the benefits when other beneficiaries are considered.More careful and comprehensive considerations of beneficiaries with broader municipal objectives, such as community engagement, should be critical in planning new transportation policies.
This research results can be adapted to other areas in Japan and in other countries, especially where the population of elderly people is larger than that of parents taking care of their children.Further investigation is merited of the impact of new transportation modes on parents when the ratio of the population of parents to that of elderly people is larger than the ratio in Odawara City, that is, 0.25.Since a larger impact could be attractive to younger generations, a more effective policy in those cases would support the growth of the population of parents after introducing the new modes, contributing to the sustainability of local cities, suffering from aging population.
In this research, only 10 combinations of the fares of community buses and door-to-door vans are investigated.Each combination only includes fares which can be paid with a small number of coins, which is important especially for elderly people because they still tend to pay with cash.However, digital payment methods have been gradually spread for elderly people, better combinations of fares would be suitable for better performance by using multi-objective optimization methods, for example.
Historically, local municipalities have not actively engaged in public transportation services in their regions because the primary providers are companies and the Japanese government regulated the companies directly.Generally speaking, local cities lack human resources who can plan transportation-related development roadmaps.However, due to aging and population decrease in local cities, transportation has become more critical for community sustainability.Therefore, we believe that the method developed in this research can be a useful tool for local cities to understand their citizens' basic transportation behavior and plan a more comprehensive transportation policy direction.
by discussing the additional functions in their agent-based simulation platform and supporting their implementation.
and MATSim 27 are amongst the open-source simulation software that are able to simulate a transportation system.These ABMS based analyses allow consideration of traffic patterns beyond efficiencies captured in ODM (Organization Destination Matrixes), with inclusion of more realistic human behavior during travel and analysis of other performance characteristics.Recently, Integrated Land-Use and Transportations and its derivative model, the Transit-Oriented Development model, have been researched to progress the integrated approach mentioned above by leveraging increased computational power.Waddell et al. 28 utilize UrbanSim to simulate the effects of transportation changes on land use and its feedback on the transportation system performance and show the quantitative importance of the feedback using the data in Greater Wasatch Front Region of Utah.Motieyan et al. 21propose agent-based modeling to balance the high-level Transit-Oriented Development perspectives and granular-level public transit infrastructure perspectives and confirm their model's validity using Tehran municipality's reference data.Beyond a focus solely on citizens' journeys on a transportation network but also capturing their experience and engagement on certain spaces while mobile are needed in this research.Moser et al., Carroll, and McDonough[22][23][24] develop an agent-based modeling and simulation platform, called "Oragamachi," to capture people's engagement behavior in heterogeneous functions of sites and connections, or activities, provided by the infrastructure.This platform's user interface helps those who are not familiar with agent-based modeling utilize the simulations, which is critical in practice because the capabilities for planning public transportation in local cities in Japan is limited and usually not capable of using detailed simulation models.Therefore, in this research, by adding an agent's transportation mode preferences, this platform is used to evaluate the impact of new transportation modes on the community engagement and finance in a local city.

F I G U R E 1
. A local city's architectural decisions consider introducing new transportation modes.The decision input data related to demographics include the fares of the new transportation modes and population statistics, including population types.The input data related to infrastructure include the existing transportation modes, geographical knowledge of the city, and the new transportation modes the city wants to introduce.The city's figures of merit include engagement indices, activity engagement ratio, and cost and revenue; these are the output data.The engagement indices capture the engagement level of the population in the city.The activity engagement ratio shows the ratio of time engaged The system concept of this research is described by the Object-Process Methodology. 29in activities to the total time, including transportation time.The financial performance of the investment in the new transportation modes is captured by the cost and revenue, including the net present value.An agent-based simulator provided by Moser et al., Carroll, and McDonough 22-24 is used as the central part of this simulation to capture the behavior of elderly people and parents in a city, including activity engagement and transportation mode selection between locations.The agent-based simulator requires two main elements; a site model that reflects the geographical infrastructure of the city and a population model representing the demographics of a city.The site model is generated by Site Model Builder, which defines locations in the city with activities they provide to citizens and connections between locations, based on the city's geographical knowledge, existing transportation modes, and new transportation modes the city wants to introduce.The population model generated by Population Model Builder includes the number of persons and the type of each demographic, their activity sensitivity, and their transportation mode preferences.A Mode Preference Calculator calculates the transportation mode preference with respect to the fare of new transportation modes.Note that each agent decides his/her transportation mode in a probabilistic way based on this transportation mode preference.Activity Sensitivity Calculator calculates the activity sensitivity with respect to the transportation mode preference.
agent-based modeling and simulation platform provided by Carroll and McDonough 22,23 is used to capture people's daily activity behavior and their transportation mode preference.The platform consists of two components: infrastructure described by a site model and a population model.The site model includes spaces with activities peo-ple can be engaged in, connections that connect between two spaces, and flows, each of which defines a transportation mode, such as walk, train, and car.The population model includes a person's sensitivity to each activity that the spaces provide and his/her transportation mode preference.Probability distributions describe the sensitivity and mode preference, and, in the simulation, any person can select any activity and mode based on the distributions with constraints by any factors, such as space capacity and transportation capacity.Since the mode preference and the activity sensitivity are affected by the fares of new modes, people's engagement and mode utilization change as the fares change.
where the right-hand side means the number of passengers and the left-hand side means the capacity of the transportation mode, which should be larger than the right-hand side.Here, c v is the capacity of a vehicle, R v the averaged utilization rate of a vehicle, and  v the averaged transport time.

A
simplified site and population models are built to capture the essential behavior of elderly people and parents taking care of their small children.Since Chuo, Tomizu Sakurai, and Kawa Tonanbu districts have a lower utilization rate of private cars than Kawa Tohokubu and Tachibana districts, the first three districts are abstracted as a Central District and the second two districts as Suburban District.According to the census data in,36 the number of elderly people in Central District is 20,304, that in Suburban District is 4873, the number of families with F I G U R E 4 The geography of six districts and six train lines in Odawara City. 34Chuo district, Tomizu Sakurai district, and Kawa Tonanbu district, shown in bold red fonts, have high public transportation accessibility, while Kawa Tohokubu district and Tachibana district, shown in bold blue fonts, have low accessibility.Katakura district is a depopulated area in Odawara City.children younger than six years old in Central District is 5085, and that in Suburban District is 1327.

Figure 5
Figure 5 shows the simple visual overview of the Odawara City site model.The site model includes two residential spaces, Residences in the Central District, where elderly people and parents in the Central District live, and Residences in a Suburban District.The site model also includes the following seven spaces to consider the typical daily behavior of elderly people and parents: population type's activity sensitivity and transportation mode preference for each transportation mode are the essential parameters to be calibrated in the simulation model.Here, the statistical data of daily activities in Japan and daily transportation modes in Odawara City are used to calibrate the activity sensitivity for each space and the mode preference for each transportation mode in the site model.Also, the transportation mode utilization ratio provided by Odawara City is considered the calibration data for the transportation mode preference in this model.Several attempts of changing these essential parameters are done to minimize the difference between the simulation results and the calibration data in Odawara City.The calibration data of Elderly People in the Central District are defined by interpreting the statistical trip-per-day data of elderly people who are 75 years or older in Japan.The statistical trip-per-day data of mothers in their 30 s are used as the calibration data of Parents in the Central District, since mothers in general mainly care for their children in Japan.Elderly people, as well as mothers in their 30 s, have 0.4 trip for going shopping or going to take a meal per day, which is interpreted as 0.3 trip/day for Grocery Store and 0.1 trip/day for Shopping Mall, since the frequency to going to buy clothes and other TA B L E 5 The capital expenditure and operation cost of community bus and door-to-door van.
tend to go out more frequently than those in Central District because, as the statistical data show, the number of trips per day for those in Suburban District is lower than in Central District due to the lack of public transportation.Then, the adoption ratio of Elderly People in the Central District is set to be 1.0, that of Elderly People in Suburban District is 1.5, and those of Parents in the Central District and in the Suburban District are both 0.5, respectively.Later in this section, the simulation results with different adoption ratios are shown to measure the effect of the adoption ratio on the engagement index and NPV.First, only two population types, Elderly People in the Central District and Elderly People in a Suburban District, are assessed to find the effect of the new transportation modes on the behavior of elderly people only.Figure 6 shows the engagement index and NPV per person for each simulation result of Elderly People in Central District and Elderly People in Suburban District, where the prefix "E_" means the target population is only Elderly People.

F I G U R E 6
Community Bus's fare is 100*n yen and Door-to-Door Van's fare is 100*m yen.A blank cell means there is no corresponding simulation result.The impact of new transportation modes on the engagement index and NPV per person of Elderly People in Central District and Elderly People in Suburban District, where "E_" means the target population is only Elderly People.Each large filled circle shows the average of the simulation results of the same combination of fares, and each small point with the same color as that of a large filled circle shows each simulation result.The star indicates the utopia point of this tradespace, and the solid red lines indicate the pareto-front.The dotted red line shows the 10% increased engagement index compared with the base case (E_Base).E_C1V3 and E_C3V3 are enlarged because these two are above the 10% line and also have positive NPV / person.investment cost.Also, the maximized engagement index in the ideal case can be understood as follows: people's preference to use the new modes is maximized because fares are zero, so people's activity sensitivities are maximized.On the other hand, the engagement index decreases as the fares increase because they are less willing to use the new modes, resulting in less frequently engaging in the activities.There are five cases on the pareto-front line; E_ideal, E_C1V1, E_C1V3, E_C3V3, and E_C1V5.In these five cases, all the simulation runs of E_ideal and E_C1V1 have negative NPV per person.However, almost all the simulation runs of E_C1V3, E_C1V5, and E_C3V3 are in the region where NPV is positive.It means that E_C1V3, E_C1V5, and E_C3V3 have advantages from the economic point of view.In addition, it is considered better in terms of community engagement when the engagement index is increased by more than 10% from the base case, shown in the red dashed line.E_Ideal, E_C1V1, E_C1V3, and E_C3V3 are better since these 4 cases are above the red-dashed line on average.Therefore, considering economics and community engagement, E_C1V3 and E_C3V3 are the preferable choices.Note that, the engagement index of E_C1V3 is 0.62, 94% of B_Ideal (0.66) and 113% of B_Base (0.547), and that of E_C3V3 is 0.614, 93% of B_Ideal and 112% of B_Base.Secondly, it is considered that there are only two population types, Parents in Central District and Parents in Suburban District, to assess the effect of the new transportation modes on the behavior of parents taking care of their children only.Figure7shows the engagement index and NPV per person for each simulation result of Parents in Central District and Parents in Suburban District, where the prefix "P_" means the target population is only parents.Similar to E_Ideal, the NPV per person in P_Ideal is negative and minimized, but the engagement index is maximized.NPV per person increases as fares increase generally, but the averaged NPV per person in each case is negative except P_C3V5.In the case P_C3V5, there are nine runs with negative NPV per person out of 40 simulation runs.

F I G U R E 7
The impact of new transportation modes on the engagement index and NPV per person for Parents in Central District and Parents in Suburban District, where "P_" means the target population is only Parents.Same as Figure6, the large, filled circles and the small points show the averaged value and the simulation results, respectively.The star means the utopia point of this tradespace, and the solid blue lines are the pareto-front lines.The dotted blue line shows the 10% increased engagement index compared with the base case (P_Base).there is no significant financial advantage or has a high engagement index.Third, all the population types in combination -Elderly People in Central District, Elderly People in Suburban District, Parents in Central District, and Parents in Suburban District -are considered in the simulation to investigate the synergetic effect of the new transportation modes.Figure 8 shows the engagement index and NPV per person B_ideal and B_C1V1 have negative NPV per person.On the other hand, almost all the simulation results of B_C1V3, B_C3V3, and B_C3V5 are in the region where NPV is positive.This means that B_C1V3, B_C3V3, and B_C3V5 have advantages from the economic point of view.In addition, it is considered better in terms of community engagement when the engagement index is increased by more than 10% from the base case, which is shown in the black dashed line.B_Ideal, B_C1V1, and B_C1V3 are better since these 3 cases are above the black-dashed

F I G U R E 9
parents are less sensitive to the fares of the modes.This is because the speeds of Community Bus and Door-to-Door Van are lower than the existing trains, buses, and private cars but parents' core needs include Comparison of simulation results when elderly people only, parents only, and both are considered.Each cross mark shows the averaged simulation result of each combination of fares when only elderly people are in the simulation.Each diamond mark shows that when only parents are considered.Each filled circle shows when both types of the population are involved in the simulation.Marks with the same color show the same combination of fares.Solid red lines show the pareto-front lines of the cases when only elderly people are considered, while solid blue lines do when only parents are considered, and the solid black lines do when both are considered.E_C1V3, E_C3V3, and B_C1V3 are marked with enlarged filled circles because they have positive NPV per person and their engagement indices surpass the 10% increased level from their base cases.shorter transportation time.The parameter settings based on these three factors make the simulation results dominated by elderly people.
elderly people and parents are considered the beneficiaries.So far, the adoption ratio has been fixed as follows: the adoption ratio of Elderly People in Central District is 1.0, that of Elderly People in Suburban District is 1.5, and those of Parents in Central District and Suburban F I G U R E 1 0 Comparison of simulation results with different adoption ratios.Each filled circle shows the averaged simulation result of each combination of fares described in Figure8, and the solid black lines show the pareto-front.Each cross mark shows when the adoption ratio of each population type is half of that in the previous case, and solid red lines show the pareto-front.Each diamond mark shows when the adoption ratio is zero for every population type.Solid blue lines show the isoline of the diamond-marked cases.Marks with the same color show the same combination of fares.B_C1V3 is marked with an enlarged filled circle because it has a positive NPV per person and its engagement index surpasses the 10% increased level from its base case.District are both 0.5, respectively.Here, two additional cases are considered.The adoption ratio is set to half of the previous simulations for the first case.The adoption ratio of Elderly People in Central District is 0.5, that of Elderly People in Suburban District is 0.75, and those of Parents in Central District and Suburban District are both 0.25, respectively.As the second case, the adoption ratio is zero for all the population types.

Figure 10 compares
Figure 10 compares these two cases with the simulation results Toward establishing a sustainable local city under the aging trend in Japan, this research's objective is to investigate the synergetic effect of introducing new transportation modes on community engagement of elderly people and parents taking care of small children, especially those who live in a local city in Japan where its accessibility to public transportation is lower than metropolitan areas.An agentbased simulation model is introduced to quantify the effect of the fares of community buses and door-to-door vans on the community engagement of elderly people and parents in Odawara City as a case study.The effect of the two modes on the behavior of elderly people and parents is evaluated separately.The simulation results predict that the introduction of the two modes increases elderly people's duration of engaging in community-related activities by at most 21%, and there are at least two preferable combinations of the fares of the new modes that satisfy more than 10% increase in the community engagement and positive net present value per person of the investment in the new modes.
The representation of the basic model with Object-Process Methodology (ISO/PAS 19450: 2015, 2015).
Summary of validation results.The transportation mode utilization ratio when citizens visit hospitals and buy clothes and others 34.
distance, one is a Residence, and the other is a Shopping Mall, and doorto-door vans with a low fare (200 yen) exist as a new mode.In this scenario an agent staying at Residence is 20% sensitive to the Shopping Mall.In 4-a, the coefficient of the increase in the sensitivity is set toTA B L E 1 Distance matrix of each space in the site model.The average speed of each transportation mode and its reference data.
TA B L E 3Door-to-door van 4.2 m/s (15 km/h) The speed is assumed to be between train/bus and private car and faster than that of Community Bus.distance between each space and Residences in the Central District is equal to or shorter than that between each space and Residences in the Suburban District.Also, the distance between residential spaces and Office is assumed as 10 km.Those who commute to outside Odawara City are around 50%, which drives the travel distance between their residence and offices.All the spaces are mutually connected, except between Residence in Central District and Residence in Suburban District, by five transportation modes; walk/bicycle, train/bus, private car, community bus, and door-to-door van, where the last two modes are the new modes considered in this research.Note that any connection between Office and each space has only Train/Bus and Private Car because community buses and door-to-door vans are introduced to support citizens' daily movement within Odawara City, which are not suitable for commuting.

Table 9
The number of trips for each place per day in the simulation results compared with the statistical data in Odawara City.44 8he fact that elderly people go sightseeing and other activities with 0.4 trip/day can be interpreted as 0.1 trip/day for Playground and 0.3 for Public Space, because Public Space includes entertainment and they would prefer going to Public Space to Playground.They go to the hospital with 0.14 trip/day, which can be interpreted as follows: Elderly People in Central District has 0.14 trip/day to Hospital.The fact that mothers in their 30s commute with 0.25 trip/day is interpreted as 0.25 trip/day to Office of Parents in Central District.The mothers are engaged in transferring with 0.5 trip per day, which is interpreted as 0.5 trip/day for Kindergarten.They are engaged in other activities with 0.4 trip per day, which can be considered as 0.2 trip/day for Playground and 0.2 for Public Space.The number of trips to Hospital of Parents In Central is 0.05 per day since they go to the hospital with 0.05 trip per day.The calibration data of Elderly People in Suburban District and that of Parents in Suburban District are calculated by the statistical data in Odawara City.The average number of trips to grocery stores and hospitals is 0.29 and 0.14, respectively, in Central District-consistent with the statistical trip data in Japan.On the other hand, those numbers are 0.25 and 0.12, respectively in Suburban District.Then, in this research, the number of trips in Suburban District is 86% of the number of trips for each space.The overall summary of these numbers of trips is shown in the fourth column of Table 6, named interpreted calibration data [trip/day].forgoing to offices or schools is set to the ratio for going to Office.Note that the utilization ratio for going to offices or schools includes students going from elementary schools to universities, which increases the ratio of walk/bicycle because they cannot drive cars.The fourth column in Table 7, named statistical data, shows the summary of the calibration target of the transportation mode utilization ratio.The activity sensitivity of each population type is calibrated with several trials to match the simulation results of the number of people visiting each space and the expected results.The expected results are estimated based on the interpreted calibration data: a population's 0.1 trip/day to a place means 10% of its total population visits the place.Here, the full site model is used with nine spaces and three modes, that is, Walk/Bicycle, Train/Bus, Private Car.The number of agents in Elderly People in Central District, that in Parents in Central District, that in Elderly People in Suburban District, and that in Parents in Suburban District are 2030, 509, 487, 133, respectively.The simulations run 40 times, and the simulation time range is 24 h, from 0 a.m. to 11:59:59 p.m.The simulation results with the tuned activity sensitivity in Table8are shown in Table6compared with the expected results calculated by the interpreted calibration data.The table shows that the simulation results are so close to the expected results that each activity sensitivity is considered to describe each population type's daily behavior well.The transportation mode preference parameters of Walk/Bicycle and Train/Bus in each population type are also calibrated within the reasonable range8to minimize the difference between the simulation results and the actual mode preference.The calibrated parameters are shown in Walk/Bicycle is close to the statistical data.However, the ratio of Train/Bus is about half of the statistical data, and the ratio of Private Car is more than the statistical one.This discrepancy can be interpreted as the target population in this simulation, being elderly and parents, tend to avoid physically stressful mobility, leading them to use trains and buses less frequently.TA B L E 6 44 Transportation parameters for community bus and door-to-door van.