Smart Cities and Urban Mobility - Digitisation
Urban Mobility Analytics
Planning, Management and control
Data and Monitoring
Emerging Data Sources
Traditional Data Sources
Digitisation of Vehciels
Digitisation of Infrastrcuture
Digitisation of Mobility Services
Loop Detectors
Drone Imaging
Video Tracking
Vehicle re-identification
Manual Counting
Development of Time-Space Diagrams
Geo-tagged social media
Smart Card Data
Floating Car Data
Provided by IoT, Connected vehicles, Smartphones
Driver logs
Fixed Observer
Aerial photos
Statistical Approach (Data drive approach)
Modelling Approach
Simulation Approach
Macroscopic (Traffic flow)
Mesoscopic (Queue)
Microscopic (Vehicle dynamics)
Traffic waves
Flow = desnsity * speed
Fundamental Diagram
Links: Features like roads
Queuing diagrams: Cumulative plot
Pulsed service rate
Constant service rate
Variables
Types of models
Modelling of indivudal vehicle movements within transport system, the vehicle moves according to its physical characteristics, fundamental rules of motion and rules of driver behaviour
Links
Nodes
Network
Individual zones within cities
Whole cities
freeway network
Freeways
Rural highways
urban streets
Roundabouts
Signalised intersections
non-signlaised intersections
Merges/diverges
Car following
Lane changing
Bus Lanes
HOV Lanes
Smart Motorways
Speed Reduction Schemes
Re-Routing
Used as dynamic response to unexpected/uncommon events or to avoid "bottleneck"
Accident
Work zones
Major events
Digitisation
SatNavs/Smartphones
Radio
Variable Message Signs
Considerations:
Space availability
Bus flows
System bottlenecks
Signage
Bus/car interactions
Considersations:
HOV Flows
Required occupancy
Space availability and bottlenecks
Enforcement
Carpooling incentives
HOV/non-HOV interactions
Increase safety (advance driver warnings)
Stabilize traffic flow
Increase driver information
Remind drivers of speed limits
Facilitate Lane closures
Lower speed limit of entire network
Different speed limits within same network
Europeans Citizens initiative (30km/h in urbans areas)
Make city centers safer
Limit emissions
Traffic reduction
Reduce noise
Traffic calming via design
Narrow roads
Roundabouts
Wide sidelwalks
Speed bumps
Smart intersections
Smart Roads
Connected vehicles
Autonomous vehicles
Monitoring services
Mobility services
Parking
Perimeter control
Congestion pricing
Where to implement?
State of existing network
State of existing public transportation system
How to reallocate the revenue?
How to determine the price?
How to determine the boundary?
Effects on the rest of the network
Reducing traffic flow into controlled zone when network starts to become congested to maintain network optimal
Off-Street parking
On-Street parking
Open lots
Parking garages
Limit Parking spaces
Yes
No
Reduce car travel demand
Shift to public transport
Free up space
Harm local business
Cruising vehicles congest traffic
Traffic parameters
Location (km or coords)
Speed (km/hr)
Density (veh/hr)
Flow (veh/hr)
Demand (veh/hr)
Capacity (veh/hr)
Queue length (veh/km)
Travel time (hr)
Waiting time (hr)
Delay (hr)
OD-Matrices
Fundamental diagram (Flow/Density)
Uncongested
Congested
Optimal density
Building Simulation
Model validation
Tailoring Simulation
Calibration
Validation
Problem definitions and model objectives
System definition
Model development
Model calibration
Verification
Documentation
Verify
Geometric representation (streets, intersection, roundabouts)
Traffic control systems (phases, offsets, timings)
Traffic rules (travel directions, banned movements)
Individual vehicle model (car following, lane change, gap accpetance)
Traffic demand (input flor patterns, turning ratios, OD matrices)
Route choice
Determine value of parameters to replicate reality much as possible
Sensitivity analysis to focus on calibration of key parameters
Compare results of simulation to independent data set to determine accuracy of model
Calibrate
Directly measurable physical parameters (vehicle mix)
Physical paratmeters not measurable (desired speed)
Tuning parameters, not physical (sensitivity to speed, aggressiveness)
Model parameters
Attributes of road network
Attributes of vehicles
Attributes of traffic system
Traffic light characterstics (cycle length, green split)
Merging priorities
Length
Width
Number of lanes
Speed limit
Geometric design (straight, horizontal curve with radius, vertical curve, grade)
Type (cars, trucks, cyclists, buses, vans)
Length
Vehicles characteristics (max acceleration/deceleration rates, fuel consumption, emissions)
Diver characteristics (aggressiveness level, desired speed, reaction times, lane changing preference, Sensitivity to speed differences, gap acceptance)
Capacities
Demands
Jam desnity
Wave speed
Node type
Merging priority rules, signal timing
Intersection geometry
Lost time at intersections
Vehicle distribution across lanes / traffic rules
Gap acceptance requirements
Car Sharing
Smart motorways
Agent-based modelling
Scenario Data
Public Trasnport
Network
Population
Demand
Scoring
Re-planning
Scenario: several datasets and parameters describing infrastructure (supply) and demand in a region
Describes a network of nodes and links along which agents/vehicle can move
Attributes: length, max speed, # of lanes, flow capacity
Multimodal: car, walk, bike, bus/train
Describes:
Location of PT stop facilities
Routes of PT lines
Departure times at stops of routes
PT vehicle capacities
transit vehicles operate at capacities and serve stops where agents board and alight
no. persons within a zone with some attributes (age, gender)
Facility data about homes, Workplaces, Schools, Shopping facilities... no. workplaces per zone, no. school places per zone
Behaviour data: travel diary survey, commuter distances, travel time distributions
Describes agents and their plans
Desribes:
Activity types
Activity durations
Activity locations
Trip mides
Trip routes
To compare different plans, a score (utility value) is calculated (Scoring function)
Postitive for performing activities
Negative utility for monetary costs, travelling, arriving late, leaving early
Calibration
Unreliastic modeal split
too/little/much traffic on roads
traffic peaks at wrong times
traffic jams at unexpected places
too many/few agents at certain places
Mostly relies on data processed through learning algorithm to learn the model then apply it
Big data required
Common used by AI, machine learning and deep learning
Machine leanring: Any technique which enables computer to mimic human behaviour
Machine Learning: Subset of AI using statistical methods to improve machine with experience
Deep learning: subset of ML enabling computation of multi-layer neural networks feasible
Adavantages
Models require assumptions and restrictions causing inaccuracy in output data, Simulations avoid restrictions allowing random processes; in some cases simulation is only practical modelling technique
Can study relationships between components in detail and simulate projected consequences of multiple design options before real-world outcome
Possible to compare alternative designs to find optimal
Process of simulation development provide insights into inner workings of network
Nodes: Features like intersections
Disadvantages
Simulation results only as good as model, still estimates/projected outcomes
Optimization only possible with few alternatives as model developed with limited variables
Accurate simulation model requires extensive resource or expensive software
Simulations cost lot of money to build and use