You are on page 1of 46

THE MOVEMENT PLANNER SYSTEM FOR 23rd Feb 2019 @

IIM Lucknow
DISPATCHING TRAINS IN RAILROAD
NETWORKS 1
AGENDA

Introduction to Data Motivation for rail Movement Planner Simulation


Science and movement planning and its features Framework
Business Analytics in the Railroad
How do you
Industry The Movement Planning
demonstrate the value
Problem
of the Movement
Movement Planner Planner to a railroad
Objectives network?
Movement Planner
Constraints
A peek into the
algorithm
Business Impact
INTRODUCTION TO DATA SCIENCE
AND BUSINESS ANALYTICS
4
5
What is “DATA SCIENCE”?
Data Science is an interdisciplinary field that uses scientific methods,
processes, algorithms and systems to extract knowledge and insights
from data in multiple forms, both structured and unstructured
 Data Science can also be looked upon as a "concept to unify statistics,
data analysis, machine learning and their related methods" in order to
"understand and analyze actual phenomena" with data
 Data Science helps decision making by basing decisions on insights
extracted from large data sets
 Data Science employs tools, techniques and theories drawn from
many fields within the context of mathematics, statistics, information
science, and computer science
Data
Science is
Multi-
Disciplinary
Data Science
is a
combination
of multiple
related skills
Why NOW?
Data is being produced and consumed at an increasingly phenomenal
rate.
Technology trends
 Improved data creation and collection technologies
 Computing has become faster but cheaper
 Data storage and retrieval has become faster but cheaper
 New tools and techniques to analyze data

Business trends
 Technology has created a level-playing field for organizations
 Economies of Scale and Scope no longer give businesses the key
competitive advantage
 Firm size does not have the same competitive advantage as before
 Increasingly “Data” has been looked upon as a source of competitive
advantage
THE DATA DELUGE
CATEGORIZATION
BASED ON BUSINESS
QUESTIONS
FROM DATA TO ACTIONS

Source: https://blog.dcrworkforce.com/hindsight-foresight-prescriptive-analytics-work
MOVEMENT PLANNER
DISPATCHERS IN A RAILROAD NETWORK

 Dispatchers control railroad operations in the section of track that they are responsible
for.
 Plan all activities associated with moving trains
 Meet-pass decisions
Clear signals with the help of CAD (Computer Aided Dispatch) systems
WHY DO YOU NEED A MOVEMENT PLANNER?
 Seven major railroads in America – Class I
20,000 to 50,000 miles of track (Indian railways - 67,312 km (41,826 mi))
 1,000 to 2,500 trains run on each per day
 Handle mixed type of freight traffic with trains having varying priorities
 Manual dispatching leads to sub-optimal decisions because each dispatcher
focuses only on the section of track that he/she controls.
 Lack a global view
 Speed of freight trains – 20-25 mph
 1 mph increase would lead to $2.5 billion savings in operational and capital expenses
 Crew changes at unplanned locations result in major expenses
 Trains wait for long periods of time at meet-pass locations leading to network
inefficiency
BNSF TRAIN SYMBOLS WITH PRIORITIES
THE MOVEMENT PLANNER PROBLEM
 A movement plan is a detailed plan for dispatching trains on a
railroad network
 Contains details of:
 Meets and passes (Crossings)
 Train movements
 Schedules all activities – crew change, fuel, inspection, train stops, train
originations and train terminations
 Delays (Dwells)
 A movement plan can be visualized on a train graph
TRAIN GRAPH – VISUAL DEPICTION OF A MOVEMENT PLAN
MOVEMENT PLANNER
Currently available methods solve small and approximate versions of the
problem encountered in practice
The GE Transportation solution is the only successful implementation of a MP
solution that is capable of dispatching hundreds of trains in practice
Complex optimization problem requiring many decisions
 Path from Origin to Destination for each train
 Meet-Pass locations
 Activity Locations
Hundreds of constraints and variables
Near-real time
 Low run-time of algorithm (< 90 seconds)
 Total time = Preprocessing + Plan Generation + Post Processing
21
DATA REQUIREMENTS: NETWORK DATA
Static Network Data
 Topology
Track section lengths (GUID)
Track Name, milepost , direction
Track characteristics – Siding, Mainline, Yard leads
Switching constraints (directional)
Height, weight and width constraints
Track gradient – used for planning purposes
Left Hand running or Right Hand running
Dynamic Network Data
Track Blocks
Speed limits – temporary and permanent
DATA REQUIREMENTS – TRAIN DATA
 Movement Reports
 Real-time location of trains
 Location reports as the train enters the track
 Train Schedules
 Very important for planning purposes
 Operation costs in BOFs, business objective functions
 Train Priorities
 Train Consist Data
 Number of engines, HP, weight, etc.
 Number of rakes, type, weight, etc.
 Train Crew
 Crew entry and expiration times
DATA REQUIREMENTS – ACTIVITY DATA
Activity Data
Activity states
Activity time – scheduled time
Activity locations
Activity duration
Activity preferences
Overlapping activities
DATA REQUIREMENTS - CONFIGURATION DATA
Station Activity Profiles (SAPs)
 Guide the MP on which location to use at a station for performing a specific
activity
 Some activities can be performed at only specific locations
 [ Location (Track #], Activity Type, Direction of Arrival, Direction of Departure,
Preference]
Track Designation Constraints (TDCs)
 Right-hand or Left-hand running preferences on tracks
Planning Track Restrictions (PTRs)
 Track restrictions for specific train categories e.g. Passenger trains will not be
allowed to move through tracks specifically used by freight trains for performing
activities
MOVEMENT PLANNER OBJECTIVES [1]
The movement planner attempts to simultaneously optimize several objectives:
 Average Train Speed – maximized  minimizing time that train spend in
waiting for meets-passes and activities
 increases throughput, reduces crew costs and decreases number of locomotives
needed to support a given train schedule
 Average Train Speed by Train Group
 High priority trains vs Low priority trains
 Algorithm Run time
 Fast response time to generate plans quickly
 State of the railroad network changes dynamically

26
MOVEMENT PLANNER OBJECTIVES [2]
 Number of Crew Expirations should be minimized
 Government regulations – max of 12 continuous hours
 Expensive – tracks blocked, reduced network capacity, crews need to be
transported to difficult to reach locations, etc.
 Percentage of meets-passes retained between plans
 A measure of algorithm robustness and stability
 Higher the better
 Average number of switches per train
 Minimize this metric in order to enable better auto-routing of trains
 Equal value passes
 Number of such passes should be minimized

27
MOVEMENT PLANNER CONSTRAINTS [1]
Track-based Requirements
 Track Blocks
 Speed Restrictions
 Switching constraints
 Height, weight and width constraints
 Preferred routes
 Siding constraints – length, multiple train per siding
MOVEMENT PLANNER CONSTRAINTS [2]
Activity-Scheduling Requirements
The MP should schedule different types of activities on the rail network at the appropriate
locations
 Meets
 Passes
 Crew changes
 Combo-train activities
 Linked-train activities
 Helper Locomotives
 Station activities
 Off-track activities
 Locked meets and passes
MOVEMENT PLANNER CONSTRAINTS [3]
Train-based Requirements
The MP should satisfy many constraints imposed by train schedules, priorities and other
train characteristics
 Schedule Adherence
 Train priorities
 Order constraints
 Key Trains
 Unplannable Trains – L, R, Y and K trains in BNSF
 Unplanned Events
 Early or Late Trains – Staging locations – EAT
 Changes in speed restrictions and track blocks
 Train Cancellations
ALGORITHM DESCRIPTION [1]
 The MP algorithm is very complex containing many components
 One simple formulation
minimize (sum of weighted costs incurred in moving trains over
planning horizon)
subject to Departure constraints;
Arrival constraints;
Flow-balance constraints;
Track-segment capacity constraints;
Travel-time constraints; and
Train activity constraints.
ALGORITHM DESCRIPTION [2]
 The GET MP algorithm exploits the structure of the problem to develop a
good solution for the entire rail network.
 Detailed model of the rail network used to simulate the movement of trains
on the network
 Iterative – systematically builds schedules for each train by moving trains at
the start of the planning cycles towards their destinations
 Systematically searches the feasible space of solutions to build a movement
plan
 Techniques used include:
 Heuristic methods (Greedy Search)
 local optimization techniques
 Backtracking methods
ALGORITHM DESCRIPTION [3]  A preferred route is the best
route – to achieve all objectives
 Start with current state of the
network
 Incrementally move trains one at
a time until the train reaches a
decision point
 Evaluate move and decide on
further movement
 Move along preferred route
 Delay train
 Choose alternate route
 Choose another train to move
 Resolve infeasible situations using
other techniques
 Repeat until the end of the
planning horizon
ADDITIONAL ALGORITHM FEATURES (1)
 Algorithm has many components that help in improving the overall efficiency and
effectiveness of the solution
 Priority Inheritance
 Low priority train in front of a high-priority train on a single track will see it’s priority
boosted till it reaches a place where the high-priority train can pass it.
 Congested Area Routing
 Special logic to handle routing of trains in busy terminals to reduce delays
 Track Designation Constraints (TDCs)
 Tracks marked right hand or left-hand running to guide the algorithm
 Business Objective Functions (BOFs)
 Each train has a BOF that controls its priority.
 The value of the BOF depends on the train category and dynamic attributes like train
delay, time to crew expiration, etc.
ADDITIONAL ALGORITHM FEATURES (2)
Sorting Advantage
 While planning, high priority trains are planned earlier than low priority trains so
that critical resources are allocated to them first followed by other trains
MOVEMENT PLANNER METRICS
THE MOVEMENT PLANNER SYSTEM IN PRACTICE
Railroad Live Train Message
Info Network
Translation Auto
Messages Signals
Systems Layer Router

Topology
Platform and Movement Dispatcher
Scenario Views Plans
Configuration
Data
SAPs, etc RTMP (Real
Time Movement Plan
Planner Planner) Quality
Configuration Analytics
Data Core Movement
Planner
37
38
IMPACT – BENEFITS TO RAILROADS (NS)
 Improved Dispatching
Increase in Average Velocity
 2 mph  $200 million savings in operational and capital expenses
 Capacity Improvement
 Travel Time decrease
 Decrease in crew expirations by around 58%
 Improved schedule adherence
 Accurate Estimated Time of Arrivals
Train arrivals estimated with an accuracy of plus or minus five minutes, two hours in
advance
 Maintenance Planning – Window of Opportunity for maintenance

39
HOW DO YOU DEMONSTRATE THE
VALUE OF MOVEMENT PLANNER TO A
RAILROAD NETWORK
Simulation
SIMULATION FRAMEWORK Analytics
Simulator
(Orchestra) Simulated
Network
Customer Data Customer Customer
Location Reports, Dependent Independent
schedules, consist,
crew, track
restrictions, etc.
Platform and
Message Translation Views Movement
Topology Planner Plans
Configuration RTMP (Real
Scenario Data
Configuration
Time Movement
Data Planner) Plan
SAPs, etc
Manual Core Movement Quality
Corrections Correction Planner Analytics
Framework 41
ACTUAL TRAIN MOVEMENTS
SIMULATED PLAN
IMPACT OF BAD DATA ON MP
 Movement Planner needs good quality data to create feasible plans.
Gap in Movement Reports:
Discontinuities in the plans for trains - Trains jumping from one point to
another in the movement plan
 Infeasibilities
 Perturbations
 Activity duration changes
 Track reservations by MP would be incorrect if it does not receive accurate data on
a real-time basis
 Trains scheduled to perform activities at incorrect locations
EXAMPLE OF CONSEQUENCES OF LOW DATA QUALITY
ANY QUESTIONS?
THANKS

You might also like