Traffic Management for Smart Cities TNK115 SMART CITIES DAVID GUNDLEGÅRD DIVISION OF COMMUNICATION AND TRANSPORT SYSTEMS
Outline Introduction Traffic sensors Traffic models Frameworks Information VS Control 2
Introduction Only in Stockholm congestion costs roughly 5-8 billions per year Depending on how you value lost time and environmental impact Build new roads? Expensive No room Generates new traffic > environmental problem! Demand management road/congestion charging? Unresonably expensive to make the problem go away Inform users and manage traffic To improve utilization of existing infrastructure 3
Introduction Selected components in Smart City Traffic Management (our definition) Efficient use of the traffic data explosion Multimodal monitoring as well as management strategies Less infrastructure heavy Dynamic mobility/travel pattern analytics on city scale Real-time OD estimation 4
Estimation and Prediction What? velocity (kmph) for nid 42 with cid 416 time between 2013-03-21 06:00 and 2013-03-21 22:00 100 4.07239 90 3.57239 80 3.07239 70 60 2.57239 position (km) 2.07239 50 40 1.57239 30 1.07239 20 0.572391 10 6AM 7AM 8AM 9AM 10AM 11AM 12PM 1PM 2PM 3PM 4PM 5PM 6PM 7PM 8PM 9PM 10PM time 0
Traffic Data What to estimate? Flow Speed Density Travel time O/D-matrix Why? Traffic information Queue warning Route choice Traffic control Ramp metering Lane metering Travel patterns Statistics Average daily traffic Travel time trends Environmental effects 6
Main sensor characteristics Fixed sensors Penetration level Number of lanes Accuracy Position Speed Occupancy Aggregation level Installation procedure Placement Floating sensors (probes) Penetration level Sampling frequency Space VS time Accuracy Position Speed David Gundlegård, ITN 10-07-30
Standard sensors Radar Flow Speed Loop detector Occupancy Speed License plate cameras Travel time 8
Radar Number of sensors: ~1800 Data: Flow and speed Aggregation period: 1 min
Radar Number of sensors: ~1800 Data: Flow and speed Aggregation period: 1 min
Alternative sensors Opposite to standard sensors High installation cost High maintenance cost Reliable Quality for low speeds Alternative sensors Traffic signal detectors Radar towers Bluetooth- / WiFisensors Toll data GPS-probes Cellular network data David Gundlegård, ITN 10-07-30
Traffic signal detectors Control traffic signals Large potential ~ 600 traffic signals in Stockholm Currently does not support sharing of data centrally Next generation traffic signals will be possible to use for traffic estimation David Gundlegård, ITN
Highway surveillance Pedestrians Standing vehicles Lost cargo Radar tower Detects movement 360 within a radius of ~ 800 m Possible to use as traffic detector? Speed and flow Travel time Vehicle tracking enables detailed behaviour observations Sid 13 David Gundlegård, ITN 2019-01-29
Raw Radar Tower Data
Radar Tracking
Bluetooth / WiFi Detects Bluetooth / WiFi devices in cars Simple and cost efficient installation ~20% penetration rate Travel time observations David Gundlegård, ITN 10-07-30
GPS-probes Smartphones Navigation devices Fleet management systems Taxi Buses Truck fleets Pay-as-you-drive Point speed and travel time observations Sid 17 David Gundlegård, ITN 2019-01-29
Taxi Number of taxis: ~1500 Data: Time, location, hired Sampling period: 1-2 min 8:00-8:05
Taxi Number of taxis: ~1500 Data: Time, location, hired Sampling period: 1-2 min 8:00-8:15
Taxi Number of taxis: ~1500 Data: Time, location, hired Sampling period: 1-2 min 8:00-9:00
Participatory sensing Main drivers for massive participatory sensing applications - Incentive - Privacy - Battery efficiency - Scalability For smart transportation applications privacy, battery efficiency and scalability is typically traded against traffic estimation accuracy and coverage Minimize privacy issues, battery efficiency and data overhead while maintaining traffic estimation accuracy for primary road network
Personal Spatial Footprint
Simple Data Processing Work place 1 Work place 2 Daycare Home
Distributed Geofencing
Cellular network data Traffic data possible to estimate Travel times Traffic flow Travel demand (O/D-matrix and activity patterns) Incident detection Steps Location data collection Transport mode classification Signalling to location Map matching Trip definition or trajectory extraction Route classification Traffic model assimilation, filtering and fusion Sid 25 David Gundlegård, ITN 2019-01-29
Cellular network data System dependent - GSM MS MS - GPRS/EDGE - UMTS U m - LTE RSS BTS MS BTS A bis A BSC BSC BSC BSC - Measurement reports - Handover (HO) NSS VLR MSC VLR MSC - Handover - Location Updates HLR GMSC O IWF OSS EIR AUC OMC Billing - Call Detail Records (CDR) Sid 26 David Gundlegård, ITN 2019-01-29
Travel demand estimation Dakar, Senegal
People density estimation Dakar, Senegal
Travel flow estimation Dakar, Senegal
Travel patterns from mobile network data Norrköping
Trip distribution over time Distribution of trips per hour within the municipality
Origin-destination flows Flow profile for trips between Åby (O) and Zone 1 (Central-zone) (D).
Traffic models For assimilation and prediction during non-recurring events we need traffic models Static macroscopic Strategic planning Volume-delay functions Static traffic assignment Macroscopic Microscopic Mesoscopic 33
Microscopic traffic flow models Modeling single vehicle behavior, with vehicle specific variables Vehicle position Speed Acceleration Sub models to describe the reaction of each driver: accelerating, breaking, lane changing Ability to model heterogeneous drivers Possible to model how single vehicles equipped with ADAD, ACC, I2V, V2V systems affect the surrounding traffic 34
Macroscopic traffic flow models Analogously to liquids in motion (hydrodynamic models) Locally aggregated quantities Density Flow Speed Used for describing the evolution of congested regions and propagation of shock-waves Based on hydrodynamic flow-density relationships Require discretization in time and space 35
Framework Assimilation of macro model Sid 36 David Gundlegård, ITN 2019-01-29
Framework Micro/meso model HISTORICAL TRAFFIC DATABASE (PROFILES/ BEHAVIORAL PATTERNS) INITIALIZATION HISTORICAL OD MATRIX (A) OFF-LINE GENERATION OF CANDIDATE TARGET OD MATRICES TARGET MATRIX GENERATION FOR SELECTED PROFILE NETWORK MODEL OD MATRIX ADJUSTMENT PROCESS REAL TIME DATABASE OF TIME-SLICED OD SEED MATRICES (B) ON-LINE SELECTION OF TARGET OD MATRIX REAL-TIME TRAFFIC DATA SELECTION OF THE OD TARGET MATRIX FOR TIME SLICE k SELECTED TARGET FOR TIME SLICE k REAL-TIME TRAFFIC DATA FOR TIME SLICE k REAL-TIME OD KALMAN FILTER ESTIMATOR PREDICTED OD MATRIX FOR TIME SLICE k+1 DYNAMIC TRAFFIC MODEL COMPLETE NETWORK INFORMATION Link Velocidad Speed en los Map arcos MANAGEMENT STRATEGIES DATABASE DEFINITION OF KEY PERFORMANCE INDICATORS Link Tiempo Travel de viaje de Times los arcos ONLINE EVENT DETECTION IMPACT EVALUATION PROCESS DECISION SUPPORT PORCESS EVALUATION OUTPUT CURRENT VS. FORECASTED STATES Alternative paths and forecasted path travel times (C) REAL-TIME ESTIMATION AND PREDICTION OF THE OD MATRIX FOR TIME SLICE k+1, ESTIMATION OF CURRENT AND PREDICTED STATE, STRATEGY SELECTION AND IMPACT EVALUATION 37
Framework Macrosopic fundamental diagram Origin r LARGE URBAN OR METROPOLITAN AREA Alternative recommended route GATE-IN Congestion Destination s QUEUE URBAN AREA TO MANAGE GATE-OUT Input flow rates (k) B C Critical Point in the managed area Real-time Traffic Data Measurements from sensors Output flows n(k-1) A Allow access Restrict access Estimation algorithm for n k ADAPTIVE FLOW CONTROL STRATEGY 38 Figure 6 Potential use of the Network Fundamental Diagram to support Active Area WideTraffic Management Strategies
OD Estimation UPPER LEVEL (Objective function evaluation) v g LOWER LEVEL (Traffic Assignment ) F ~ k ~ k ~ k k ~ k k ~ k k ~ k g,v,tt γ1f1 g,g γ2f2 v,v γ3f3 tt,tt.. ~ k ~ k s t v assignmt g g~ k 0 39
Preprocessing Raw Traffic Data Base Data Filtering and Completion Filtered & Complete Data Base Data Fusion & Traffic Modeling Processes 40
Assimilation and fusion
Assimilation and fusion - example Nonlinear measurement model for fusing travel times and point speed measurements using CTM-v as traffic model Ensemble Kalman Filtering for assimilation and data fusion Inherent travel time delay Travel time decomposition v y n n n n M ( v h( v n n 1 ) (0, Q (0, R n n ) ) ) n n
Assimilation and fusion - example Predict using CTM-v Correct using point speeds Correct using travel times
Traffic information VS Control Information Variable Message Signs Traffic maps Control Lane closures Ramp metering Demand management 44 Congestion charging Dynamic departure time and mode shift incentives
Building a traffic estimation system Sensor data Android GPS probes Preprocessing of measurements Map matching Road network model Needs to be routable Spatial and temporal aggregation Fusion Model assimilation Visualization 45
46 Norrköping: Privacy preserving participatory sensing
47 Norrköping: Privacy preserving participatory sensing
www.liu.se