Presentazione tesi magistrale Martino Marangon

36
Analysis of the impact of speed variation dynamics on the optimization of the real-time railway traffic management problem 1 University of Trieste Department of Engineering and Architecture Master Degree in Computer Engineering Master student: Martino Marangon Supevisor: Lorenzo Castelli Co-supervisor: Paola Pellegrini

Transcript of Presentazione tesi magistrale Martino Marangon

Page 1: Presentazione tesi magistrale Martino Marangon

Analysis of the impact of speed variation dynamics on the optimization ofthe real-time railway traffic management problem

1

University of TriesteDepartment of Engineering and Architecture

Master Degree in Computer Engineering Master student: Martino Marangon

Supevisor: Lorenzo Castelli Co-supervisor: Paola Pellegrini

Page 2: Presentazione tesi magistrale Martino Marangon

Introduction

2

Page 3: Presentazione tesi magistrale Martino Marangon

Introduction

3

Railway Infrastructure● Track circuit: portion of track where the presence of a train is

detected automatically● Block section: it is composed by one or more track circuit

○ Delimited by two main signals○ Different block sections may share one or more track circuits

● Route: ordered sequence of block sections, that links an origin-destination pair

Page 4: Presentazione tesi magistrale Martino Marangon

Introduction

4

Block section● When a train enters the first track circuit of a block section, all the

following are reserved to it● If a train uses a track circuit, no other train can use it.● Utilization: both physical occupation and prior reservation● For safety reason, only one train can use the same block section at

the same time.● The aspects of a signal indicate to a train driver the behaviour that

has to be followed○ Green aspect: two following block sections are free ⇨ train can

proceed at designed speed○ Yellow aspect: following block section is free, following one is used by

another train ⇨ the train has to decelerate and reach a complete stop before the next signal

○ Red aspect: following block section is used by another train ⇨ train must reach a complete stop before the signal

Page 5: Presentazione tesi magistrale Martino Marangon

Martino Marangon

Real-time railway traffic management problem

(rtRTMP)

5

Page 6: Presentazione tesi magistrale Martino Marangon

Real-time railway traffic management

6

Train Delay● Primary Delay: when a train suffers of a certain amount of

delay due to unexpected events (technical failure)● Secondary Delay: when a train suffers of a certain amount of

delay caused by the delay of another train

Due to the interaction between trains, these delays may propagate along the whole network generating knock-on delay

rtRTMP● Consists in selecting the train routes and schedules in order to

minimize the propagation of delays● Time constraints: time interval to compute a solution is strictly

limited

Page 7: Presentazione tesi magistrale Martino Marangon

7

Modeling rtRTMPThe modelling approaches to rtRTMP can be divided into two classes:

● Fixed-speed models: speed profile of a train is not updated in case of conflict

Real-time railway traffic management

Page 8: Presentazione tesi magistrale Martino Marangon

8

● Variable-speed models: the speed profile of a train is modelled in order to include possible consequences of conflicts

Real-time railway traffic management

Page 9: Presentazione tesi magistrale Martino Marangon

9

Modeling rtRTMP● Variable-speed model: complete and accurate but difficult to

solve in short computation time● Fixed-speed model: require less computation time● The fixed-speed model is the most used

Objective● Develop a variable-speed program that can be used to

evaluate the conflict solution computed by fixed-speed algorithms○ RECIFE-MILP algorithm

Real-time railway traffic management

Page 10: Presentazione tesi magistrale Martino Marangon

Objectives of the study:

1. Develop a program to model Infrastructure and Rolling Stock

2. Develop a program to compute running times in absence of traffic

3. Develop a program that can be used to evaluate the solutions of the RECIFE-MILP algorithm

10

Page 11: Presentazione tesi magistrale Martino Marangon

Objectives of the study:

1. Develop a program to model Infrastructure and Rolling Stock

2. Develop a program to compute running times in absence of traffic

3. Develop a program that can be used to evaluate the solutions of the RECIFE-MILP algorithm

11

Page 12: Presentazione tesi magistrale Martino Marangon

Modelling Infrastructure and Rolling Stock

Context:● RECIFE Project

○ Rodriguez et al., RECIFE: Models and Tools for Analysing Rail Capacity, 2007

Scenarios:● Information in various data structures in several files● Standard format for data representation

○ Infrastructure, Rolling Stock, Timetable, Maintenance work

○ XML files and schemas○ RECIFE Data Structure (RDS)

Proposal:● Develop a program to generate Infrastructure.XML and

RollingStock.XML according to schema defined by RECIFE Data Structure

12

Page 13: Presentazione tesi magistrale Martino Marangon

Modelling Infrastructure and Rolling Stock

RDS - Infrastructure

13

Page 14: Presentazione tesi magistrale Martino Marangon

Modelling Infrastructure and Rolling Stock

RDS - Rolling Stock

14

Page 15: Presentazione tesi magistrale Martino Marangon

Modelling Infrastructure and Rolling Stock

Program to generate Infrastructure.XML and RollingStock.XML

Input:● Various XML and text files containing several informations

about the Infrastructure and the Rolling Stock○ OpenTrack output○ Handmade Files

Output:● Infrastructure.XML● RollingStock.XML

These files are used as:● Basis for the development of Running Time Computation

Program● Basis for the development of other programs concerning

railway management in real-time15

Page 16: Presentazione tesi magistrale Martino Marangon

Modelling Infrastructure and Rolling Stock

Pierrefitte-Gonesse● Length of the line ~ 16 km● Running Time ~ 7 minutes and 40 seconds● 79 Block Sections and 89 Track Circuits

16

Page 17: Presentazione tesi magistrale Martino Marangon

Modelling Infrastructure and Rolling Stock

Rosny sur Seine - St. Etienne du Rouvray● Length of the line ~ 70 km● Running Time: ~ 23 minutes and 27 seconds high speed trains,

26 minutes and 57 seconds passenger trains, 41 minutes and 3 seconds freight trains

● 10 stations● Presence of gradient

17

Page 18: Presentazione tesi magistrale Martino Marangon

Objectives of the study:

1. Develop a program to model Infrastructure and Rolling Stock

2. Develop a program to compute running times in absence of traffic

3. Develop a program that can be used to evaluate the solutions of the RECIFE-MILP algorithm

18

Page 19: Presentazione tesi magistrale Martino Marangon

Running Time Computation

Running Time

● Amount of time that a train needs for travelling along a track circuit

● Duration calculated in absence of traffic● Depends on:

○ Infrastructure characteristics ○ Rolling Stock characteristics

● In order to compute running time:○ Route partitioned into sections of equal characteristics○ Compute Traction Resistance○ Establish if train is in:

■ Braking Section■ Acceleration Section■ Constant Speed Section

19

Page 20: Presentazione tesi magistrale Martino Marangon

Running Time Computation

Traction Resistance RF:

○ Rolling Stock Resistance RL○ Distance Resistance RD○

Rolling Stock Resistance● Davis mass-independent formula

○ A and B accounts for mass and mechanical resistance○ C refers to air resistance

20

Page 21: Presentazione tesi magistrale Martino Marangon

Running Time Computation

Distance Resistance RD○ Gradient Resistance RS○ Curve Resistance RB○

Gradient Resistance●● Since railway gradient is very slight

I: inclination [‰]

21

Page 22: Presentazione tesi magistrale Martino Marangon

Running Time Computation

Curve Resistance

● Roeckl’s formula

22

Page 23: Presentazione tesi magistrale Martino Marangon

Running Time Computation

23

Braking Section

● Maximum speed of the following section is lower than speed of the current section

● Speed reduction must be completed before the following section

Compute Braking Section2 steps:

1. Calculate braking distance to reduce speed from maximum speed vmax1 of current section to maximum speed vmax2 of following section

If sred is shorter than length of current section:i. First part constant speedii. Second part braking section

Page 24: Presentazione tesi magistrale Martino Marangon

Running Time Computation

24

2. If sred is longer than the length ssec of current sectiona. Compute new maximum speed value for current sectionb.

If the maximum speed of the previous section is greater than the maximum speed of the current section, the braking distance has to be computed for the previous section. For this reason the braking distance is computed in reversed order.

Page 25: Presentazione tesi magistrale Martino Marangon

Running Time Computation

Acceleration Section● Maximum speed of the following section is greater than the

maximum speed of current section● 3 possibilities

○ Maximum speed of the section is reached

25

Page 26: Presentazione tesi magistrale Martino Marangon

Running Time Computation

○ Exit speed lower thanmaximum speed

○ Intersection withbraking curve

26

Page 27: Presentazione tesi magistrale Martino Marangon

Running Time Computation

Compute acceleration value

● FZ = Z(v) - RF(v,s) traction power surplus○ Z(v) tractive effort○ RF(v,s) traction resistance

● ρ mass factor for rotating masses

27

Page 28: Presentazione tesi magistrale Martino Marangon

Running Time Computation

Constant speed section

● Train has reached its maximum speed or the maximum speed of a section

● Describe movement: s = v•t● Gradient and curve section:

○ Traction resistance increases, possibly surmounting tractive effort: train decelerates

28

Page 29: Presentazione tesi magistrale Martino Marangon

Running Time Computation

Running Time Computation Program

● Input:○ Infrastructure.XML○ RollingStock.XML

● Output:○ RunningTimes.txt: running time for each train○ TrainDetails.txt:

■ One for each train■ Detailed information about the train run

○ Gantt.XML

29

Page 30: Presentazione tesi magistrale Martino Marangon

Running Time Computation

Comparison with OpenTrack● 2 different time steps

○ 1 second: simulation faster but less precise○ 0.1 second: simulation slower but more precise

30

Page 31: Presentazione tesi magistrale Martino Marangon

Objectives of the study:

1. Develop a program to model Infrastructure and Rolling Stock

2. Develop a program to compute running times in absence of traffic

3. Develop a program that can be used to evaluate the solutions of the RECIFE-MILP algorithm

31

Page 32: Presentazione tesi magistrale Martino Marangon

Running Time Computation with Traffic

Context● RECIFE-MILP algorithm

○ Pellegrini et al., RECIFE-MILP: An Effective MILP-Based Heuristic for the Real-Time Railway Traffic Management Problem, 2015

○ Tackles the real-time traffic management problem based on the fixed-speed model

Proposal● Given a perturbed situation, use RECIFE-MILP to compute a

set of solutions● Develop a program that can be used to assess the gap

between RECIFE-MILP solutions and solutions computed taking into account the consequences of braking and acceleration phases

32

Page 33: Presentazione tesi magistrale Martino Marangon

Running Time Computation with Traffic

Running Time Computation with Traffic Program

● Input:○ Infrastructure.XML and RollingStock.XML○ TimeTable.XML○ Disruption.XML: amount of delay that the trains suffer when

entering the infrastructure ○ RECIFESolution.XML: information on train routing and priorities

chosen by RECIFE-MILP algorithm● Output:

○ RunningTimes.txt: infrastructure entrance time, infrastructure exit time and running time for each train

○ TrainDetails.txt:■ One for each train■ Detailed information about the train run

○ SolvedGantt.XML33

Page 34: Presentazione tesi magistrale Martino Marangon

Conclusions and Future Work

34

Page 35: Presentazione tesi magistrale Martino Marangon

Conclusions and Future Work

Running Time Computation Program● For both infrastructures modelled we reached very good results:

the average time difference is around/lower than 1 second for the whole trains’ routes

● Program is very flexible: can be easily modified in order to include:○ Human behaviour of train drivers○ Weather conditions○ Track conditions

Running Time Computation Program with Traffic● For all experiment, the program finds conflict-free solutions

○ Some scenarios different solutions than OpenTrack● Evaluate conflict solutions computed by RECIFE-MILP● Evaluate conflict solutions computed by others algorithms

35

Page 36: Presentazione tesi magistrale Martino Marangon

36

Thank you