0LAUK0 2015 01 Design Report

Quick Links

Home
Content	Plan	Report	Simulation
Log	Week1	Week2	Week3	Week4	Week5	Week6	Week7

Autonomous Bus Scheduling System Design Report

The following report is a simple version of the final report. For the final report you can download the following file [1].

Introduction

Background

Public transportation has been around for a couple of decades now. It started with the stagecoach in the nineteenth century. Later on came the steam trains which were capable of carry large capacity of people at the same time. The trains were too big to travel in city themselfs, so the tram was invented. However not city are capable of creating a tramrail network. So with the increasing demands of mobility in the sixties, busses started to be used for public transportation in cities around the Netherlands and the rest of europe.

Bus stops were created around cities and a static scheduling system was implemented. This static scheduling system uses a fixed time table around the city. The busses follow fixed routes each time and do not take into account if there are any people waiting at a bus stop or not. Meaning that the busses do not take a shorter route when nobody is willing to stop of is waiting a certain bus stop. This means it is time for a dynamic system, this will be the main focus in this report. The new scheduling system for busses aims to create a more efficient passenger flow. It will be able to send extra busses to bus stops that are or are getting too crowded and also the system will choose a shorter route to it’s destination when no passengers are waiting at the coming bus stops. Users can check in at the bus stop or with their mobile phone apps. The scheduling system will then optimize the routes by taking this information into account. A small part of the report will discusses the possibility of using autonomous busses in the new dynamic scheduling system, analyzing the impact of user, society and enterprise.

Different approaches can be used to optimize this schedule. Those will be analyzed and take feedback data from a survey. This data will be inputted in a simulation in order to analyze the consequence of the different needs of the users. Furthermore in this report the infrastructure will be redesigned. The most important infrastructure alteration will be the placement of the bus depot. The simulation will also be used to validate the new dynamic scheduling system.

The validation will have with different requirement like waiting time and crowdiness but also with happiness. This is a design report for dynamic scheduling bus system, incorporating the needs of the User, Society and Enterprise.

Objectives

Objective	Short description
1.2.1	A potential maximum waiting time of 20% longer than it currently is.
1.2.2	An average decrease in travel time of 20% (not including during rush hours)
1.2.3	An increase in happiness of the travelers

1.2.1

When a bus skips a few bus stops, it will arrive sooner than expected at the next stops, which means that people might have to wait longer than usual for next bus if that one does not skip any. This will lead to potential longer waiting times. However, this also means that there are also potential shorter waiting times between busses, if the first bus does not skip any stops, but the second one does.

1.2.2

When there are stops that can be skipped, more efficient bus routes can be used to reduce the travel time. This will not be possible during rush hours, because the bus stops will be too crowded, which means that there will barely be any stops the bus can skip. In the following picture shows a fragment of the bus route between Eindhoven and Veldhoven.

1.2.3

Happiness will be “quantified” in the simulation, and the decrease in travel time should be weighing more heavily than the potential increase in waiting time, and perhaps the troubles caused by the use of the app to sign up for bus rides.

Requirements

Requirements	Short description
1.3.1	Automatically optimize the bus routes by skipping several stops according to check-in info from bus stops and web interface.
1.3.2	Arrange more buses to certain routes to make sure every passenger has a seat when passenger flow increases (determined by check-in info).
1.3.3	Keep the schedule static during heavy load hours. This schedule is predetermined by adapting from the traditional scheduling systems.
1.3.4	The scheduling system must be able to switch to manual control mode.
1.3.5	The scheduling system must be able to collect feedback data from user through various methods( web interface, phone apps, etc...).
1.3.6	The system must be able to track locations of each bus and this info will be provided to user through various ways ( sms notifications, web interface, mail, phone apps, etc…)

1.3.1

When there’s no one checked in in the coming bus stops, the scheduling system should try to search for a shortcut to skip them by taking current traffic loads into account. The system should only search within the available routes (without disturbing the neighborhoods) and send the new route to the bus.

1.3.2

The scheduling system will arrange more busses to certain routes to control the crowdiness. By crowdiness we mean outside the heavy load hours every passenger should have a seat.

1.3.3

The scheduling system should not optimize the routes dynamically during heavy load hour. When doing so the system won’t be conflict with other transport systems.

1.3.4

The scheduling system provides the interface to manually control the routes and schedule for special situations ( fire, natural disasters, etc…)

1.3.5

The scheduling system provides the interface for collecting user feedbacks like “too late” “too earlier and missed it” “5 min later than expected”. This data can be used to further optimize the system’s optimization algorithm.

1.3.6

The system will send notifications to user who has pre checked-in for route changing, reminds and other info. User will be able to choose the way of how this notification sent to them. User who have not pre checked-in can found these data on web interface/phone apps.

Proposed Plan of Action

As can be seen, we have already clarified the background, objectives, and the requirements regarding the autonomous bus scheduling system. Now, we will we explain how we are going to investigate whether such a system could replace the current scheduling system.

We have three main cases we should distinguish:

1. Will the ABSS decrease the average waiting and traveling time of the passengers?

2. Will the ABSS enhance the beneficial looking at the three USE-aspects?

3. Should autonomous busses replace the bus drivers?

In order to investigate whether or not the ABSS will decrease the average waiting and traveling time, we will create a simulation with as well as possible real life features. With real life features a Poisson Process for passenger arriving time for example is meant. We use Unity3D (Frank should update it) to make this simulation.

For the second and last case we will conduct a survey and distribute them among at least 70 bus passengers, the passengers are not entirely randomly selected, but are selected so that in each age and gender category we have about the same amount of people.

Besides the survey we will study research that already has been done in scientific journals or informative magazines, analyze those findings and apply them in our own investigation.

Past attempts

Time	Description	Author
1967	The scheduling problem has been formulated as a constrained optimization problem for frequency determination. The objective was to minimize the total travel time for a given fleet size constraint and a random search procedure was used for solution.	Lampkin, W. and Saalmans, P. D.
1972	A model was proposed to solve the optimization problem. This model searches for an optimum bus network by adjusting iteratively the frequencies and type of buses on each link to correspond to the link flow level, such that the service on some links is enhanced whilst on others it is depleted. The optimum situation is reached when no further change in link service levels is detected.	Rea, J. C.
1974	A system has been proposed that can determine optimum frequencies for a set of bus routes and fleet size, which could minimize the total travel time and discomfort (travelling without a seat) by a gradient method procedure.	Silman, L. A., Barzily,Z. and Passy, U.
1977	A mathematical programming algorithm of the compound minimization type for bus traffic model has been developed. The problem of optimal bus frequencies is solved by a gradient projection method.	Scheele, S.
1980	A detailed simulation model for studying the effect of frequencies on various route level and network level measures of effectiveness.	Dhingra, S. L.
1981	They developed a model that allocates the available buses between time period and between routes so as to maximize net social benefit subject to constraints on total subsidy, fleet size and level of vehicle loading.	Furth, P. G. and Wilson, N. M. H.
1982	Another model that recognizes passenger route choice behaviour and seeks to minimize a function of passenger wait time and bus crowding subject to constraints on number of buses available and the provision of enough capacity on each route.	Han, A. F and Wilson, N. M. H
1994	The model used the previous concepts of optimal vehicle size, frequency adjustment for co-ordinated routes, timed transfer and transit centers.	Shih, M. and Mahmassani, H. S.

Current attempts

The best known current attempts regarding autonomous bus transport are actually more aimed at providing autonomous public transport and they are all based on autonomously driving vehicles. As an example for this we can take the Google Self-Driving Car Project, it is a project that was started by Google X, the experimental sub-division of Google, and it involved developing technology such that the car is able to operate autonomously. The most important sensor is a LIDAR system which is used to create a 3D scan of its environment.

In order for our autonomous bus transport system to be fully autonomous we need to have the buses drive autonomously on top of the autonomous scheduling we are already intending, this will help prevent human errors that could impact our autonomous scheduling system. The Google Self-Driving Car Project shows that this will be possible right now and may be realistic by 2020 once relevant regulations and laws are updated such that autonomously driving vehicles are allowed.

Not all attempts are successful however, in Eindhoven there has been the Phileas bus for example, which should be driving autonomously. It would drive autonomous over bus lanes such that it did not have to interact with other traffic and there on the bus lanes all intersections have traffic lights such that there was no need to deal with other traffic. The project however failed with reasons not exactly known but there have been some small accidents, the only autonomous part right now is that it stops autonomously at bus stops attempting to provide a nicer experience to passengers compared to bus drivers occasionally braking late and hard.

As a conclusion we think that autonomously driving vehicles are definitely going to be possible, most likely within five years already and thus autonomous public transport is looking to be achievable as well and may even arrive earlier due to less constraints. Parallel to this autonomous scheduling systems are also very likely to be developed, a reason for this is that you want a central that can control the public transport as a whole; while you want every vehicle by itself to be able to ensure safety to its passengers and surroundings, for example by performing an emergency stop as needed, there are also bigger issues such as roads that are temporarily blocked, etc. that are better solved by a central system, and in order to have immediate feedback and also a better experience for the passengers these decisions will need to be made very fast and that is exactly where some kind of artificial intelligence to support the autonomous scheduling system is very useful for. Lastly, if we look at the past and current attempts we can see that it is really not clear that a bus is the best choice. If you want an autonomous system where a bus can reach a passenger as fast as possible while you also don't want to waste the capacity of the bus, then maybe a bus is not the solution but instead you want a much smaller vehicle that is capable of delivering a high throughput through the network while still being economically feasible.

Survey analysis

This survey has two goals. First of all it wants to determine the weight of, waiting time, travel timeand the busyness of the bus for the simulation. These weights will be used in the simulation to calculate a good overall happiness level for the passengers. Secondly it we want to find out what the social impact could be of a bus scheduling system. We want to know how open people are to the possible change coming. Do they think improvement is needed at all? Are they prepared to use an app? How do people think about bus routes going through their streets? All these questions will help us understand what the public expects from such a system and how open they are towards this change. The survey that was used can be found in the wiki page called survey. The survey was handed out on the bus station of Eindhoven central station and via Facebook. Because not all questions are about traveling in a bus, and we also wanted participants that took the bus less, we used Facebook and handouts to distribute the surveys.

In total there were 102 responses to the survey. According to the Centraal Bureau voor de Statistiek(CBS) 6.7% of the total population of the Netherlands is using the public transport. This means the total of people using public transport in the Netherlands is 1125600. To get a desired margin of error of 5% 387 responses were needed. Since there were only 102 responses to the survey, the margin of error will be 10% with a confidence interval of 95%. This margin of error is pretty high which does not boost well for the validity of this study. However with the lack of time we decided to continue instead of gathering data since the results from this survey will be used for the simulation.

Before we preform statistical tests on the data the outliers had to be removed first. All data entrieslooked normal except one entry. This entry had all extreme values that deviated from the mean by a large amount (1 versus M=6.1, 6.5, and 6.4). This entry also had non-serious remarks on the open questions. We decided to remove this entry. With this removed the sample consisted of 60 man and 41 women with an average age of 26 (M = 26.26, SD = 12.26). When we looked closer at the data it became apparent that at question 6-8 there were a lot of answers around 5. When we looked at the survey on a mobile phone it became apprehend why this was the case. If one looked at the survey with a mobile phone the possible answers shown were 1-5 and one had to use the scrollbar to reveal the answers 6-10. This caused a lot of people to view it asa 1-5 scale instead of a 1-10 scale. To resolve this problem we decided to multiply every entry that had all values lower or equal to 5 at questions 6-8 by 2. This is of course not perfect since it is possible that a participant rated the questions this low while using the full scale. However the variables did show normality after this multiplication while this was not the case before this. With the low margin of error and the changing of the variables for question 6-8 it is no longer possible to get valid conclusions from this survey. However as stated before we will continue with the data analysis while acknowledging that the results from this analysis will not be scientifically valid. We will assume that the changes made and the small sample size will not influence the validity of this analysis.

First let us look at the means for, waiting time (M = 7.84, SD = 1.36), travel time (M = 7.27, SD = 1.61), and busyness (M = 7.55, SD = 1.65). From this we can conclude waiting time is most important for passengers to experience a pleasant journey as this has the highest average of all factors. The means will now be used in the simulation to see how effective our system is. It is interesting to see that all these values are not too far apart, but this is probably due to the multiplying of the low values.

To see whether there is a difference in responses from people traveling often with the bus and people traveling less with the bus, we ran a chi2 test on whether people travel often with the bus and how important they rated waiting time, travel time and busyness. For waiting time and busyness there was a difference, χ2(1, N = 101) = 32, p = .024 and χ2(1, N = 101) = 33, p = .05. Travel time did not show a significant difference, χ2(1, N = 101) = 18, p = .64. From this we conclude, people rate waiting time and the busyness of a bus higher when they travel more often with the bus.

To distinguish we diced to split the participants up into two groups, those who take the bus regularly (at least once a week), and those who take the bus less (less than once a week). For the group traveling often with the bus the values are: waiting time (M = 8.05, SD = 1.18), travel time (M = 7.29, SD = 1.69), and busyness (M = 7.97, SD = 1.54). For people who travel less with the bus the values are: waiting time (M = 7.46, SD = 1.60), travel time (M = 7.23, SD = 1.50), and busyness (M = 6.77, SD = 1.57). Waiting time is still the most important factor for comfort according the participants.

Now let us look at the social impact. Currently the people are neutral about the current system (M = 3.32, SD =.82) for a score between 1 and 5. A chi2 test learned us that there is significant difference in these answers whether people travel often or less often with the bus χ2(1, N = 101) = 9, p = .42. Although people are not dissatisfied whit the current system there is still a lot of room for improvement. Of all the participants asked 55% would travel more often with the bus if the system would get improved this is again not depend on how often you travel with the bus. The 55% is a substantial amount and could convince investors to invest in this system since it would generate more and happier passengers. The data shows that 55% of the participants are prepared to use an app to register for a bus stop. This is a low value considering the young average age of the participants (M = 26.26, SD = 12.26). Younger people are easier in using smartphones in day to day life and people should use the app for the system to be most effective. A Wilcoxon Signed-Ranks test was conducted to see if indeed older people would be hesitant to use an app, and this was indeed the case Z=-2.79, p<0.0053. Our system functions best when all people use an app to register for a bus stop and only 55% was willing to use an app. This is something which is not nice and is something that should be resolved otherwise the system won’t function as well. It is good to notice that there is no one that has trouble with a bus route going through their street if the bus route is already there. However there are people that say that they would mind it if there would be a bus route going through their street. This has the implication that a part of the public the public (47%) will not like the idea of bus routes going through their street. However this could change over time since the people with currently bus routes through their street don’t mind.

Concluding we can see that for the simulation waiting time will be the most important to determine happiness. Furthermore there is a difference in results when people travel more often with the bus. To get the most accurate representation it is important to use the average values of the people traveling often with the bus. These values of the important factors for a comfortable journey are: waiting time (M = 8.05), travel time (M = 7.29), and busyness (M = 7.97). For the social impact the survey gave us a better understanding of how the users think. Currently the users are ok with the current system but there is room for improvement. Improving the system would make people take the bus more often. This is of course positive and makes the new system interesting for investors, as it would create more passengers. It is however the case that using an app to register is not that popular with the users. Only 55% wants to use this feature and older people are even more likely to not use the app. To get our system optimized the app should be used. There should be work done to increase the amount of users of the app. Lastly 47% of the participants would mind it if there would come a bus route through their street. However of all the participants with a bus route going through their street, no one finds it unpleasant.

Simulation

Simulation A large part of designing a bus system is simulating how the system is going to work. This simulation gives you a better idea of how the system is going to work and how effective it is. It is also possible to see where the weaknesses are in our system. The simulation we created consists of some bus lines where the busses can drive over. Each piece of route between two bus stops has a travel time which is based on the real travel time of the current system. There will be a random flow of passengers generated that are distributed over the bus stops. The happiness level of these passengers is dependent on waiting time, travel time and crowdedness. The weight of these variables is determent via the survey and its analysis. With these weights and the information the simulation gives us, we can calculate the happiness of each passengers and the average happiness. We can run the simulation with our new system and the old system and compare the happiness. Furthermore we will be able to see how the flow of passengers will influence our system. It could be possible that in rush hours the new system will be the same as the old system since it is not possible anymore to skip bus stops. This information can help us to make conclusive remarks on our system. For more information about the simulation look at the simulation tab.

References

1.Lampkin, W. and Saalmans, P. D. (1967). The Design of Routes, Service Frequencies and Schedules for a Municipal Bus Undertaking: A Case Study, Operation Research Quarterly 18, pp 375 - 397.

2.Rea, J. C. (1972).Designing Urban Transit Systems: An Approach to the Route Technology Selection Problem, Highway Research Record 417, Highway Research Board, Washington, D. C., pp 48 - 58.

3.Silman, L. A., Barzily,Z. and Passy, U. (1974).Planning the Route System for Urban Buses, Computers and Operations Research, Vol. 1, pp 201 - 211.

4.Hsu, J. and Surti, V. H. (1977). Decomposition Approach to Bus Network Design, ASCE Journal of Transportation Engineering, Vol. 103, pp 447-459.

5.Scheele, S. (1977). A Mathematical Programming Algorithm for Optimal Bus Frequencies, Ph.D. thesis, Department of Mathematics, Linkoping University, Linkoping, Sweden

6.Dubois, D., Bell, G. and Llibre, M. (1979). A Set of Methods in Transportation Network Synthesis and Analysis, Journal of Operations Research Society, Vol. 30, No.9, pp 797-808.

7.Dhingra, S. L. (1980). Simulation of Routing and Scheduling of City Bus Transit Network, Ph.D. thesis, Department of Civil Engineering, lIT Kanpur, INDIA.

8.Furth, P. G. and Wilson, N. M. H.(1981). Setting Frequencies on B-us Routes: Theory and Practice, Transportation Research Record 818, Transportation Research Board, Washington, D. C., pp 1 - 7.

9.Han, A. F and Wilson, N. M. H (1982). The Allocation of Buses in Heavily Utilized Networks With Overlapping Routes, Transportation Research B, Vol. 16, No.3, pp 221 -232.

10.Baaj, M. H. (1990). The Transit Network Design Problem: An AI-Based Approach, Ph.D. thesis, Department of Civil Engineering, University of Texas, Austin, Texas.

11.Shih, M. and Mahmassani, H. S. (1994). A Design Methodology for Bus Transit Networks with Coordinated Operations, Research Report 60016-1,Center for Transportation Research, University of Texas at Austin, Austin, Texas.

12.Dashora, M. (1994). Development of an Expert System for Routing and Scheduling of Urban Bus Services, Ph.D. thesis, Department of Civil Engineering, lIT Bombay, INDIA.

13.“A survey of public opinion about autonomous and self-driving vehicles in the U.S., the U.K. and Australia”, Brandon Schoettle and Michael Sivak, July 2014

14.“Ethical Decision Making During Automated Vehicle Crashes” in Transportation Research Record: Journal of the Transportation Research Board, Noah J. Noah.

15. U.S. Census Bureau. Vehicles Involved in Crashes by Vehicle Type, Rollover Occurrence, and Crash Severity: 2009. Statistical Abstract of the United States. Publication Table 1107. U.S. Department of Commerce, 2012.

16. Office of Freight Management and Operations. Freight Facts and Figures 2011. Publication FHWA-HOP-12-002. FHWA, U.S. Department of Transportation, 2011.

17. NHTSA. Traffic Safety Facts 2009: A Compilation of Motor Vehicle Crash Data from the Fatality Analysis Reporting System and the General Estimates System. Publication DOT HS 811 402. U.S. Department of Transportation, 2009.