Next Generation Traffic Control with Connected and Automated Vehicles

Transcription

1 Next Generation Traffic Control with Connected and Automated Vehicles Henry Liu Department of Civil and Environmental Engineering University of Michigan Transportation Research Institute University of Michigan, Ann Arbor July 19, 2016

2 Current Traffic Signal Systems An open loop control system. Majority of transportation agencies DO NOT monitor or archive traffic signal data. Benefit/Cost ratio of signal re-timing is about 40:1; but usually traffic signal systems will be re-timed every 2 ~ 5 years.

3 SMART Signal System Development ----Funded by USDOT/MnDOT, Total ~$2.5M, :09:15.012, D8 on, 7.902s 08:09:15.481, D8 off, 0.468s 08:09:16.761, G3 off, s 08:09:16.761, Y3 on, s 08:09:17.620, D9 on, 2.686s 08:09:18.151, D10 on, 2.593s 08:09:18.307, D9 off, 0.687s 08:09:18.823, D10 off, 0.671s 08:09:20.244, Y3 off, 3.482s 08:09:21.649, D22 on, s 08:09:22.008, D22 off, 0.359s 08:09:23.242, G1 on, s Detector #8 on at 08:09:15.012; Vacant time is 7.902s Green Phase #3 off at 08:09:16.761; Green duration time is s Detector #9 off at 08:09:18.307; Occupy time is 0.687s Yellow Phase #3 off at 08:09:20.244; Yellow duration time is 3.482s Green Phase #1 on at 08:09:23.242; Red duration time is s Event-based high resolution data TS-1 type cabinet MnDOT Implementation

4 Traffic Signal Performance Measurement Queue Estimation Arterial Travel Time Estimation Distance H n L max v 3 Loop Detector A B C v 2 D v 5 L d n L min v 1 v 4 n T g TA T n r TB n n 1 n T max T C T g T min n 1 T r Time Liu, Wu, Ma, and Hu. (2009). Real-time queue length estimation for congested signalized intersections. Transp. Res. Part C, 17(4), Liu, H.* and Ma, W. (2009) A Virtual Vehicle Probe Model for Time-dependent Travel Time Estimation on Signalized Arterials, Transportation Research Part C, 17(1),

5 Field Implementation at Pasadena February type cabinet

6 Connected Vehicles A connected vehicle system is based on wireless communication among vehicles of all types and the infrastructure. The wireless communications technology could include: 5.9 GHz DSRC LTE-V and 5G cellular networks Other wireless technologies such as Wi-Fi, satellite, and HD radio Source: USDOT

7 Connected and Automated Vehicles The path toward connected vehicles will ultimately lead to automated vehicles. Source: USDOT

8 Transition to Next Generation Traffic Control Systems Signal-free intersection Infrastructure Adaption Connected and Automated Vehicles Connected Vehicles Spatial and temporal signal control Detector-free signal operation Lane reassignment Regular Vehicles Current Practice - Fixed time/actuated/adaptive Signal

9 Safety Pilot Model Deployment at Ann Arbor Funded by USDOT (August 2012 May 2015) $31M 2843 vehicles equipped Passenger cars, trucks, buses, motorcycles, and a bike 73 lane-miles of roadway 27 roadside installations Collected over 110 Billion DSRC basic safety messages over 38 Million miles of driving 9

10 10 Vehicle-to-Infrastructure (V2I) 19 Intersections 3 Curve-related sites 3 Freeway sites All DSRC communications logged

11 Traffic Control with Connected Vehicles Data Sever MAP/SPAT Control MAP/ SPAT BSM Data Collection Device RSE MAP/ SPAT BSM Vehicle with OBE DSRC BSM Vehicle with OBE RSE: Roadside Equipment OBE: Onboard Equipment 11

12 CV Data Collection Devices Signal Cabinet at Plymouth & Barton Data Hub

13 Why Detector-Free is Important? Many traffic signals in the US are fixed-time. To retime these signals, manual data collection has to be conducted. For vehicle-actuated or adaptive signals, vehicle detectors have to be maintained properly, which is also costly. Connected vehicles are mobile sensors. Potentially we can use connected vehicle data to evaluate traffic signal performance, retime traffic signal, or control traffic signal in real time.

14 Traffic Signal Evaluation Using CV Trajectory Sample CV data Plymouth Green 3 Arrival on Green: 25% Green Green Start Start Green End

15 Key Problem: Traffic Volume Estimation If traffic volumes are known, then there are known optimization methodologies to retime the traffic signals. How to estimate arrivals using CV data with low penetration? Regular Vehicle Connected Vehicle

16 Case Study-Int. Plymouth & Green Int. Plymouth & Green Date: 04/25/16-05/13/16 Plymouth Green 11:00 AM 18:00

17 Vol (Veh/h/l) Validation of Estimation Observed data collected on 04/25/16 and 04/26/ % Overall MAPE :30 11:30 12:30 13:30 14:30 15:30 16:30 17:30 18:30 Observed Data Estimated Data

18 Transition to Next Generation Traffic Control Systems Signal-free intersection Infrastructure Adaption Connected and Automated Vehicles Connected Vehicles Spatial and temporal signal control Detector-free signal operation Lane reassignment Regular Vehicles Current Practice - Fixed time/actuated/adaptive Signal

19 Formulation Bi-level optimization: Lower level: trajectory control - Objective: minimize fuel consumption/emission - Generate compact platoon - Control platoon leading vehicle speed Upper level: signal optimization - Objective: minimize delay/maximize throughput - Determine signal parameters - Decide platoon length

20 From Temporal Control to Spatial-Temporal Control

21 Cooperative Driving on Dedicated Road for CAV Platoon control Through cars Left-turn cars Signal optimization

22 Cooperative Driving on Dedicated Road for CAV Platoon control Through cars Left-turn cars Signal optimization

23 Signal-free Intersection Signalized Intersection Signal-free Intersection Source: Self Driving Car Simulator

24 Slot-based (reservation-based) Algorithm Source: Self Driving Car Simulator

25 Mcity Safe, repeatable, off-roadway test environment for AVs: simulated city Technology research, development, testing, and teaching $6.5M project; $3.0M funding from MDOT Construction commenced July Grand opening: July 20, 2015

26 Conclusion Connected and automated vehicle technology will transform the surface transportation system and significantly impact on our society. It will also transform the traffic control industry. It brings a set of completely new research questions during the transitional process from human driven vehicles to autonomous vehicles. An interesting time for transportation research

27 Contact Information Henry Liu, Ph.D. Professor, Department of Civil and Environmental Engineering, Research Professor, Transportation Research Institute UMTRI University of Michigan, Ann Arbor 2320 G.G. Brown, 2350 Hayward Street Phone: Fax:

28 Multi-Modal Considerations in CAV Traffic Control Larry Head, Mehdi Zamanipour Systems and Industrial Engineering University of Arizona July 19 th, 2016

29 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion The fundamental questions. Will we even need traffic signals in the future? What happens when the volume increases? Do we see emergent behavior that mimics traffic signals? How will we transition during market adoption? Fajardo, D. T. Au, T.Waller, P. Stone, and D.Yang, Automated Intersection Control: Performance of Future Innovation Versus Current Traffic Signal Control, Transportation Research Record: Journal of the Transportation Research Board, No. 2259, Transportation Research Board of the National Academies, Washington, D.C., 2011, pp

30 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion (Circa 2007) Control for Safety and Efficiency in a CAV Environment Information Richness Vehicle Control Vehicle Trajectories Vehicle Dynamics Vehicle Performance Detection 1 Today Complex Networked Control Systems 2 CICAS-V Priority Cooperative Control Possible Early CV Application Vehicle Based Safety Actuated Adaptive Optimizing Structured Aware Active Efficiency Safety

31 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Multi Modal Intelligent Traffic Signal System (MMITSS) USDOT Connected Vehicle Pooled Fund project Sponsors: FHWA, MCDOT, Caltrans, VDOT Team: UA, UC PATH, Econolite, Savari Dynamic Mobility Applications (DMA) Components: Intelligent Traffic Signal Control (I-SIG) Signal Priority (TSP, FSP, PREEMP) Mobile Accessible Pedestrian Signal System (PED-SIG) Real-time Performance Observer (PERF-OBS) 4

32 Connected Vehicles and Infrastructure Systems DSRC 5.9 GHz Radio BSM/SRM Signal Phase and Timing (SPaT) MAP Vehicle(s) + Connected Vehicle Equipment Cooperative Applications: Transit Priority Truck Priority Emergency Vehicle Priority Connected Vehicle Infrastructure Equipment Road Side Unit (RSU) On Board Unit (OBU) After Market Safety Device (ASD) MAP Data Digital Description of Roadway (D. Kelley, 2012) 5

33 Introduction System Architecture MMITSS Simulation Environment Signal Priority Basic Algorithm Coordination Concepts Priority Extended Model Weight Analysis Conclusion Priority Hierarchy Rail Crossings Emergency Vehicles Freight Coordination Transit BRT Express Local (Late) Passenger Vehicles Pedestrians A Traffic Control System Section 1 Priority for Freight July 19,

34 Introduction System Architecture MMITSS Simulation Environment Signal Priority Basic Algorithm Coordination Concepts Priority Extended Model Weight Analysis Conclusion Priority Hierarchy Rail Crossings Emergency Vehicles Transit BRT Express Local (Late) Pedestrians Passenger Vehicles Freight A Traffic Control System Section 2 Priority for Transit Pedestrians July 19,

35 Introduction System Architecture MMITSS Simulation Environment Signal Priority Basic Algorithm Coordination Concepts Priority Extended Model Weight Analysis Conclusion Real-Time Performance Measures by mode, by movement Transportation Management Volume (mean, variance) Delay (mean, variance) Travel Time (mean, variance) Throughput (mean, variance) Stops (mean, variance) Connected Vehicle System Management Channel Utilization (b/s) Market Penetration A Traffic Control System Infrastructure Device Reliability and Availability July 19,

36 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Considerations in Traffic Control (aka Objectives) Manage Queues for Regular Vehicles (minimize delay) Phase actuation/adaptation (Service Opportunities) Provide Priority for Special Vehicles (minimize delay) Coordinate Signals for Progression (throughput, smooth flow) Constraints Traffic Movement Conflicts (dual-ring/phase controller, approach based not Lane Based) Safety (Human Drivers minimum green, yellow change, red clearance) 9

37 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Hierarchical Priority Priority Hierarchy Rail Crossings Emergency Vehicles Freight Transit BRT Express Local (Late) Pedestrians w m is the weight assigned to mode m α, β are weights assigned to priority vehicles and coordination Zamanipour, M., Head, K.L., Ding, J. A Priority System for Multi-Modal Traffic Signal Control. In Proceeding of Mobil.TUM 2013 Conference, Munich, Germany. 10

38 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion An Integrated Priority Traffic Signal Control Model Minimize (Connected Vehicles Delay (Priority and Regular) + Coordination Priority Request) (Link) s.t. Precedence Constraints Phase Duration & Interval Constraints Coordination Priority Request Delay Evaluation Constraints Connected Vehicle Delay Evaluation Constraints (Link) 11 Zamanipour, M., A Unified Decision Framework for Multi-Modal Traffic Signal Control Optimization in a Connected Vehicle Environment, Ph.D. Dissertation, University of Arizona, July 2016 CP p K k k p TL l J j TM m m l j m c J d w m,,,, min (5.1) (5.21), (6.9) (6.19), (7.22) (7.33). [Zamanipour, M.] s.t. 0, 0,,,,, l j k p k p k p k p k p k p d c f y x g t

39 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Modal Preferences and Weight Analysis Seconds 39 Transit Vehicles and Trucks Average Delay (1,0) (128,1) (64,1) (32,1) (16,1) (8,1) (4,1) (3,1) (2,1) (1,1) (1,2) (1,3) (1,4) (1,8) (1,16) (1,32) (1,64) (1, 128) (Transit Weight, Truck Weight) (1,0) Transit With Priority Truck With Priority Transit Without Priority Truck Without Priority 12

40 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Modal Preferences and Weight Analysis Average Truck Delay Modal Delay Space (seconds) Average Transit Vehicle Delay 13

41 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Coordination Priority Model Provide smooth flow of traffic along streets by synchronizing two or more adjacent intersections <- Back 14

42 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Coordination Priority Model Coordination virtual request in time-phase diagram Non- Coordinated phases Coordinated phase Non- Coordinated phases Coordinated phase Cycle Permissive Period Cycle ø2 ø2 Non-Coordinated Phase Ø1 Local Time Coordinated Phase Ø2 Clock Time 12:00:00 12:01:40 12:03:20 Non-Coordinated Phase Ø4 Non-Coordinated Phase Ø3 Coordinated Phase Yield Point Coordinated Phase Force-Off Point Coordinated Phase Permissive Period 15

43 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Coordination Priority Model Minimize Priority Request Delay + Coordination Delay s.t. Precedence Constraints Phase Duration & Interval Constraints Priority Vehicles Delay Evaluation Constraints Coordination Delay Evaluation Constraints rl rl t t rl t rl ru ru t + g ru t + g ru t + g delay = (t-rl) + (ru-(t+g)) delay = (ru-(t+g)) delay = 0 (a) (b) (c) (d) <- Back 16

44 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Coordination Priority Model <- Back 17

45 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Coordination Priority Model <- Back 18

46 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Coordination Priority Model Experiment Specification: Gavilan - Daisy Mountain, Anthem, AZ 8-phase intersection Simulation Length: 1 Hour Main Street (3lanes) Demand: 500 veh/h/l Side Street (2 lanes) Demand: 300 veh/h/l Cycle: 100 seconds Coordinated Phase Split: 30 Seconds Coordinated point is start of phase Both phase 2 and 6 are coordinated Coordinated phase 6 timing comparison Time Fully Actuated Cycle Number Coordinated Phase Red Time Coordinated Phase Green Time Time ASC/3 with Coordination Time Coordination Priority Back Cycle Number Coordinatd Phase Red Time Coordinated Phase Green Time Cycle Number Coordinatd Phase Red Time Coordinated Phase Green Time

47 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Coordination Priority Model Effect of increasing side street demand on coordinated phase 6 using Coordination Priority Time Side Street Demand: 300 veh/h/l Cycle Number Coordinatd Phase Red Time Coordinated Phase Green Time Time 120 Side Street Demand: 450 veh/h/l Cycle Number Coordinatd Phase Red Time Coordinated Phase Green Time <- Back Time Side Street Demand: 750 veh/h/l Cycle Number Coordinatd Phase Red Time Coordinated Phase Green Time 20

48 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Extended Priority Model with Queue Consideration Assumptions: The acceleration rate, the deceleration rate, and the speed limit for all vehicles are the same. The vehicle composition includes equipped and non-equipped vehicles. Some of the equipped vehicles are priority eligible vehicles and the rest are regular passenger vehicles. The saturation flow rate, s, is assumed to be constant. The shockwave queue, q, is assumed to be constant based on historical data if there is no connected vehicle in the queue. It is also assumed that the average vehicle length (L) is known <- Back 21

49 Cumulative Arrival-Departure Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Extended Priority Model with Queue Consideration Cumulative Arrival-Departure (a) First Vehicle Is Detected by Loop Detector at Stop Bar, (b) Connected Vehicle Joins The Queue VQ j Cumulative Arrival-Departure Connected Vehicle Is in the Queue and Another Connected Vehicle: (a) Approaches the Queue (b) Joins the Queue q ~ d j s Cumulative VQ j Arrival-Departure q j d j d j 1 s T 0 Current Time T j (a) 1 t p LT j T max Time T j T 0 T j+1 LT j LT j+1 T max Current Time (a) 1 t p Time q ~ q j 1 s d j 1 VQ j <- Back T 0 q j T j Current Time (b) d j 1 t p LT j s T max Time VQ j+1 22 VQ j q j d j (b) 1 t p T 0 T j T j+1 LT j LT j+1 Current Time T max Time

50 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Extended Priority Model with Queue Consideration Travel Time comparison of Priority Model (Ch. 5), Coordination Model (Ch. 6) and Extended Priority Model (Ch. 7) Transit Vehicle Frequency Demand Level Measurement ASC/3 Free Priority without Queue (Chapter 5) Free Actuated Extended Priority with Queue (Chapter 7) Coordinated Actuated Extended Priority with Queue (Chapter 7) Coordination ASC/3 Coord. Priority without Queue 25% PR 50% PR (Chapter 6) 25% PR 50% PR 5 minutes 10 minutes <- Back All CP NCP Transit All CP NCP Transit All CP NCP Transit All CP NCP Transit

51 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Extended Priority Model with Queue Consideration Percent improvement of applying Extended Priority Model (Chapter 7) when compared to Priority Model (Chapter 5), Coordination Model (Chapter 6) and Transit Vehicle Frequency Demand Level Measurement Fully Actuated Actuated Coordinated 25% PR 50% PR 25% PR 50% PR 5 minutes 10 minutes All CP NCP Bus All CP NCP Bus All CP NCP Bus All CP NCP Bus <- Back 24

52 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Weight Distribution: Modal Preferences and Weight Analysis Coordination=5.0 Truck=1.0 Transit= TSP and FSP and Coordination Priority Coordinatd Phase Red Time Coordinated Phase Green Time Coordination=1.0 Truck=1.0 Transit= TSP and FSP and Coordination Priority Coordinatd Phase Red Time Coordinated Phase Green Time Coordination=1.0 Truck=5.0 Transit= TSP and FSP and Coordination Priority Coordinatd Phase Red Time Coordinated Phase Green Time

53 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Considerations in Traffic Control (aka Objectives) Manage Queues for Regular Vehicles (minimize delay) Phase actuation/adaptation (Service Opportunities) Provide Priority for Special Vehicles (minimize delay) Coordinate Signals for Progression (throughput, smooth flow) Constraints Traffic Movement Conflicts (dual-ring/phase controller, approach based not Lane Based) Safety (Human Drivers minimum green, yellow change, red clearance) 26

54 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Opportunities for Traffic Management in a CAV Environment Explore the role of traffic signals Transition period (20++ years) to Connected and Automated Vehicles Role of Infrastructure in the Future? Safety Redundancy Assuming we still need signals (I'll retire before the assumption expires) Relief from Human Control Parameters (Min, Yellow, Red) Lane based control Mode based control 27

55 Introduction System Architecture Simulation Environment Signal Priority Algorithm Coordination Priority Extended Model Weight Analysis Conclusion Questions? Discussion Larry Head Systems and Industrial Engineering University of Arizona (520) Zamanipour, M., A Unified Decision Framework for Multi-Modal Traffic Signal Control Optimization in a Connected Vehicle Environment, Ph.D. Dissertation, University of Arizona, July 2016 July 19,

56 Intersection Vehicle Infrastructure Cooperative Traffic Management and CAVs Wei-Bin Zhang University of California at Berkeley Automated Vehicles Symposium San Francisco July 19, 2016

57 Evolvement of Traffic Control Strategies Detection Inputs to Controller Strategies/imitations Actuated Control Stop-line detectors Passage detectors Baseline control plans based on historical traffic data Loop actuations Baseline plan does not count real-time traffic demand Gap-out: waster green time when waiting for the last vehicle passed the intersection Max-out: not enough green time to allow platoon passing through Adaptive Control Stop-line detectors Upstream detectors Predicted flow rate arrived at the intersection using flow and speed measurements at upstream loops The prediction of arrival flow that uses the temporal information measured at upstream detectors could be greatly distorted due to the loss of the spatial information Advanced Traffic Control based on V2X Vehicle location and speed as function of time and space Predicted vehicle and platoon arrival time (1 st vehicle) at the intersection and its occupancy (time duration the platoon passing the stop-line) Signal coordination adjustments for platoon-based green wave Green termination by the end of platoon Accommodate prioritized vehicle passing sequence based needs Human behaviors can only be accommodated to a limited extend Traffic control for automated vehicles Vehicle location and speed as function of time and space Predicted vehicle, platoon and Vehicle flow status, and cooperative vehicle actions potentially to achieve optimization for efficiency and safety improvements from all desirable angles 2

58 Intersection Vehicle Infrastructure Cooperative Traffic Management Estimating performance measures from probe data Better capture traffic at various penetration rates Data fusion with existing traffic detection data Intelligent Traffic signal control for mobility Priority for designed travel groups Queue spill-back avoidance Perimeter control in grid networks to prevent gridlock Traffic signal control for safety Dynamic all-red extension minimizing arrival rate on yellow ECO signal operations In-vehicle driver speed advisory for minimum fuel consumption Integration of adaptive signal priority with driver advisory/assist

59 Fusion of Hybrid Data for Travel Time Estimation

60 Fusion of Hybrid Data for Queue Length Estimation 1: probe based: State equations are based on shockwave speed at time step n and n-1 Shockwave speed is measured as distance between locations of two probe snapshots divided by time difference 2: loop detector based: State equations are based on shockwave speed at time step n and n-1 Shockwave speed is measured using (q, k) diagram as flow difference divided by density difference

61 Mean Absolute Percent Error (Analysis for I-880) Congested periods between 3/2 and 3/17, 2012

62 Estimation Error w.r.t Penetration Rate By J. Q. Li and K. Zhou 7

63 Priority for Specified Travelers

64 CA Affiliated CV Testbed along El Camino Real Major arterial in congested area Caltrans 2070 controller software (similar to LADOT s traffic control software) Interface controller between Caltrans traffic controller and Savari RSE Use Data Manager for centralized data control, timestamping, and synchronization Stanford Cambridge California Page Mill Portage/Hansen Curtner Matadero Ventura Los Robles Maybell Charleston

65 California Connected Vehicle Test Bed 7/19/2016 MMITSS NOCoE Webinar 10

66 DSRC Communication Range CA Testbed (10 intersections) 7/19/2016 MMITSS NOCoE Webinar 11

67 CA MMITSS AZ Site CA Site Test bed traffic Suburban intersections Major arterial in congested intersections Traffic Controller System Architecture Econolite ACS3 traffic controller software MRP integrated with RSE Peer to peer data communication Caltrans 2070 controller software (similar to LADOT s traffic control software) MRP hosted by a separated industrial PC Data Manager for managing data sharing, time stamping, and synchronization MMITSS Algorithm MMITSS traffic and priority control algorithm based on adaptive traffic control system MMITSS traffic and priority control algorithm based on Caltrans TSCP for coordinated-actuated signal control 7/19/2016 MMITSS NOCoE Webinar 12

68 Priority Algorithm: Considering Bus Delay Average delay (sec/veh)

69 Priority Algorithm: Trade off with Traffic Delay on Non-Bus Directions Average delay (sec/veh)

70 Data Comparison (Truck + Bus): Number of Stop and Delay Time due to Stops 7/19/2016 MMITSS NOCoE Webinar 15

71 Summary of Priority Benefits Without Priority With Priority Improved Percentage SB NB Total SB NB Total Number of Trips Number of Stops 3.9 ± ± % Cumulative Time of Stop 127.0s ± 65.3s s ± 53.2s 14.1% Trip Travel Time 386.4s ± 82.1s s ± 67.6s 7.4% Delay due to Stops 132.1s ± 74.2s s ± 54.5s 12.5% 7/19/2016 MMITSS NOCoE Webinar 16

72 Traffic Signal Control for Safety

73 Red Light Running Collision Avoidance ~20% of all intersection crashes occur due to signal violation crashes Connected Vehicles can provide effective detection methods to detect activities near intersections

74 Research Findings Dilemma Zone (Cont d) Study of individual vehicle trajectories revealed that drivers did not make a one time stop-and-go decision Accelerate to beat the yellow were common Change mind during the yellow interval Following a leading vehicle was found to be the most common case Prediction of RLR event needs to consider these behaviors A prediction algorithm has been developed and evaluated with field data Distance@ speed speed (mph) Yellow (mph) onset(ft) (a): East Bound v 2 /2a + v v Y Type I Dilemma zone s to intersection s to intersection Vehicles could have stopped safely Vehicles accelerate to beat the yellow but proceeded following a leading vehicle Velocity@ yellow 100 onset (mph) Distance to 100 Intersection(ft) 50 Distance to Intersection(ft) 0 0

75 Autoscope Virtual Loops Sensors 12 Autoscope cameras, 3 for each approach Software reconfigurable virtual loops Each virtual loop reports speed, timestamp, vehicle length, etc. Using discrete point detectors to lower the system cost and have wider choice of sensors. Outputs Multiple outputs are associated to study the driving behaviors All detectors are used to study driver s decision making during signal phase transition Only 2 nd and 3 rd advance detectors are used for RLR prediction

76 Data Analysis: Findings Characteristics of Red-light running at arterial intersections Time-into-red distribution Over 90% of RLR within first 2s into red Confirmed existing empirical statistics All red extension for 2-3 seconds accounts for most RLR and its related hazardous situations

77 Data Analysis: Findings (Cont d) Characteristics of Red-light running at arterial intersections Headway distribution at advance area 60% of RLR following other vehicles Avg Headway of RLR < Headway yellow through Normalized Frequency Headway of RLR Headway of Yellow through Headway(s) Over 60% of RLR vehicles were in a platoon Possible reason for RLR: car-following behavior

78 Data Analysis: Findings (Cont d) Characteristics of Red-light running at arterial intersections Headway distribution at advance area Headway of RLR Headway of stopping vehicle RLR Yellow through Average headway (following zone) 2s 2.2s Stopping 2.25s Normalized Frequency Headway(s) RLR headway 10% less than that of Yellow through

79 Data Analysis: Findings (Cont d) Review Dilemma zone protection Green-extension system Offline timing Static all red extension Yellow duration Dynamic all red extension Arterial intersections vs high speed intersection Yellow onset(ft) s to intersection Yellow onset distribution Type I Dilemma zone 2s to intersection v 2 /2a v ( t-t 0 ) Velocity@ yellow onset (mph) Over 90% of RLR were NOT trapped in the dilemma zone Different RLR driving behavior at arterial intersection comparing to high speed intersection Dilemma zone protection system may not perform as expected in this case

80 Summary of Findings Data Analysis TSA ConOps Data Analysis based on Autoscope Data (on Calif SR-82, El Camino Real) The characteristics of RLR at arterial intersections are different than that on high speed intersections: Over 90% of RLR vehicles were NOT caused by being trapped in dilemma zone. Dilemma zone protection (or green extension) systems may not perform as expected at arterial intersections. Over 60% of the RLR vehicles were within a platoon, and with a headway that is 10% less than yellow through vehicles. Traffic control algorithms also consider platoons may be effective in reducing RLR frequency

81 Distance to intersection TSA Algorithm Development (Cont d) TSA Algorithm-RLR / hazard prediction Case 1: Last minute all-red extension Case 2: All-red extension requested at fixed location How to determine the minimum distance System requirement: the system should provide protection (or warning) for 85% of the RLR occurrences Percentage (histogram) Percentile (cumulative distribution) Minimum Distance required to detect(feet) Distribution of t tinred v of RLR Page Mill Rd EB Page Mill Rd WB Page Mill Rd SB 60 Page Mill Rd EB Page Mill Rd WB Page Mill Rd SB Minimum Distance required to detect(feet) Last sensor Non-RLR RLR Adaptation with discrete sensor Red onse t All red interval Time into red Adaptation (CICAS-V input) Process time time

82 Infrastructure Based Concepts for Safety 27 Intersection Safety Measure / Crash Surrogate Probabilistic Model of Intersection Safety Traffic Control Adapt signal (On-demand all-red extension) Veh Trajectories Minimize chance of of crash crash (w/o (w/o too much too much comprise comprise on efficiency) on efficiency) Corridor Measurement Measurement of Corridor Corridor Level Control Safety by Connected (adaptive offset) Vehicle data Optimize progression and reduce the yellow arrivals 27

83 ECO Driving

84 Eco-Approach Scenario Diagram Intersection of interest

85 Where are the Differences? CV measurements enable Accurate arrival prediction (through message rely) for advance signal adjustments Better estimation of traffic signal status More effective performance measures Travel time (sec) Sample single run along 11-singlized intersections and 7 bus stops Predicted arrival time Bus trajectory Singalized intersection Bus stop Travel distance (m) 30

86 RICHMOND FIELD STATION TEST SETUP - Driven Lap: Blue Leg + Black&White Leg - Recording Area: Blue Leg (320 m) -Straight Approach: Yellow Leg (Actual GPS-Position faked)

87 RICHMOND FIELD STATION TEST RESULTS Summary APIV Uninformed* Informed* w/priority Uninformed w/priority Informed Number of Laps Stop Frequency (%) % of Change % % % Mean Stopped Time (sec) % of Change % % % Travel Time (sec/trip) % of Change % % % Fuel (l/100km) % of Change % % % 32

88 Fuel Saving for Different Display All the percentages are comparisons with baseline (no DVI) trips Scenario (1) (2) (3) (4) (5) (6) Current signal state Scenario description Green Green Green/Re d Maintai n speed to pass Speed up to pass Have to stop HMI T2C : Time to signal Change HMI VEC : headway to Virtual Eco-driving Car HMI TGTS : advisory Target Speed band Red Reduce speed to pass Red Maintain speed to pass T2C -3.9% 11.3% No data 10.2% 12.5% VEC 1.1% 25.7% No data -10.7% 3.6% TGTS -3.6% 29.3% 26.9% No data 2.8% Overall -2.1% 23.7% 26.9% 0.3% 7.2%

89 Intersection Vehicle Infrastructure Cooperative CAVs

90 Connected Vehicles Infrastructure Contents and Services Overthe-Air (Broadcast) Other Vehicles Connected Vehicles Cloud Applications (Structured)

91 Subject to interference Vision System - Blockage of reference by snow, dirt or water (may be nondetectable) - Interference by wet pavement, sun glare, headlight of on-coming traffic, shadow (may be nondetectable) - Moderate installation cost - High maintenance requirement

92 Challenges for Vision Based System Camera System Reference Lines

93 GPS-based System Subject to interference - Blockage by tunnel, trees, tall buildings (detectable with delays) - The GPS signal is subject to degradation and loss through attaches by hostile interests. Potential attacks cover the range from jamming and spoofing of GPS signals to disruption of GPS ground station and satellites. *** - Possible to be jammed by GPS jammers (may be detectable) - Spoofing (non-detectable) - Moderate to high cost vehicle antennas *** USDOT, Vulnerability assessment of the transportation infrastructure relying on Global Positioning System, 2001

94 Communication System (e.g., DSRC): Technological Challenges Line-of-sight constraint Data collision Subject to interference General RF noises Receiver overload Adjacent channel interference or out of band emissions (OOBE) Cross-talk inter-flow interference Subject to jamming

95 To Close Connected vehicle, infrastructure and travelers will facilitate revolutionary changes of the transportation systems for better mobility, higher capacity, safe, cost effective, environmental friendly, and traveler responsive solutions Greater Challenges lie ahead technical and otherwise

96 Questions and Thoughts? Wei-Bin Zhang

97 Connected Vehicle Adaptive Signal Control Systems (CoVASS) Breakout Session #8, AVS 2016 Jul. 19, 2016 Young-Jun MOON, Ph.D. Chief Director of National Transport Technology R&D Center ISO/TC204 WG17 Convenor

98 I What is CoVASS 2

99 Connected Vehicle Adapted Signal Control System Smart Mobility R&D Projects funded by MoLIT Ministry of Land, Infrastructure and Transport 39km/h V2X-i 43km/h 45km/h 35km/h 3

100 R&D for CoVASS Technologies being Developed V2X-i Communication Infrastructure in n x m Intersections Dynamic Signal Phasing with Cycle-free & Local Network Offsets n x m Network Optimization Collision Avoidance R&D Process Stage I: 2015~2020 Step 1: 2015~2017 Development and Verification of CoVASS Algorithm Step 2: 2017~2020 Test Bed and Validation of CoVASS in nxm Network Stage II: 2021~2025 CoVASS Development & Deployment in Real Networks (nxm + nxm + ) 4

101 II Why CoVASS 5

102 Road Vehicle Policy MoLIT Comprehensive Plan to Reduce Traffic Fatalities 2013~2017 Goal : Reduce road traffic fatality rate by 30% until 2017 Fatality per 10,000 vehicles : 2.5 in 2011 => 1.6 in 2017 Development & Implementation : 5 strategies Ministry of Land, Infrastructure and Transport (MoLIT) 6

103 Road Vehicle Automation Policy MoLIT Introducing Relevant Systems Real Test on Road Standard and/or Regulation for Commercialization Digitalizing Infrastructure Roadway Digital Map Precise Positioning V2X Connectivity Developing T&E Technologies Core Algorithms T&E Sites (K-city) Model Deployment Improving Legislation for Road Test Operation Permission of test operation ( 15) temporary based on requirements Designation for Test Road ('16) Developing Management System Temporal Exemption for Devices ( 15) Automated Driving Function Establishing Legislation for Commercialization Making Regulation & Harmonization Harmonization with UN Regulations New Insurance Policy Compensation for Traffic Accident Induced by Automated Driving Recall & Inspection Legislation ( 18) Legal System for Automated Vehicles 7

104 Cooperative ITS (C-ITS) - MoLIT Focusing Safety, Promoting Mobility & Sustainability (Green Transport) V2V & V2I Communication for Connected Vehicles Next Generation ITS to provide a service on the open platform 8

105 Next Generation for Signal Control System Real-time Signal Control System by COSMOS Cycle, Offset, Split Model of Seoul (COSMOS) Standardized Signal Control Systems from 2000s by National Police Agency (NPA) Controlling Oversaturated Flow Conditions in Signalized Intersections Technology Concerns beyond COSMOS Vehicle Detection Systems based on ICT Digital Signal Controller Dynamic Algorithms according to Variable Flow Conditions in Signalized Intersection Networks 9

106 III CoVASS Technologies 10

107 Signal Control Strategies Traffic Flow Conditions Intersections Low Volume Under-saturated Saturated Isolated CoVASS Green Activated 1 x n CoVASS Green Coordinated m x n CoVASS Queue Growth Equalized 11

108 12 CoVASS Green Coordinated Algorithm

109 CoVASS Queue Growth Equalized Algorithm 13

110 Vehicle Adaptive Detection by V2X-i RSE V2I 14

111 IV Future Works 15

112 Automated Driving R&D Projects (MoLIT) Areas ~ Levels Level 0 Level 1 Level 2 Level 3 Systems Smart Highways (7 years)) Cooperative Automated highway Systems (C-AHS) (5 years) Automated Vehicle & Highway Systems (AVHS) Digital Infrastructure Satellite Based Positioning Technology I (8 years) Satellite Based Positioning Technology II & Digital Map (5 years) CV/AV Technology Signal Controls CoVASS (5 years) CoVASS II Simulation CV/AV Traffic Simulation Model (4 years) CV/AV Traffic Simulation Model II (4 years) Pilot Projects C-ITS Pilot I (3 years) C-ITS Deployment Automated Driving T&E T&E ASV Safety T&E (8 years) K-City (3 years) AV T&E 16

113 Roadmap for C-AHS Integrated Goal (Pan Government) Survey, Analyze and Design 1 st Stage 2 nd Stage 3 rd Stage Improve GPS accuracy (Margin of error: 10~15m 0.5m) Develop core technologies 공용 TB 인프라구축 (1) Broad scale implementation, Produce prototype and assess performance standardization Build common Build common Test bed(3) Maintenance Test bed(2) 기술시연 Ad-hoc and regular demonstration Prepare permission guidelines Pilot testing during for test driving Pyeongchang Winter Olympic Final Demonstration Level 3 Automated vehicle (FY2020) 17

114 Future Works Inter-disciplinary R&D Projects for Automated Driving in C-AHS, CoVASS, K-City R&D by MoLIT Automated Vehicles R&D by MoTIE Collaboration & Cooperation R&DT by National Consortium with Research Institute, Industry, and Academia Evaluation & Deployment by Public Sectors including Local & Central Governments Future Works Automated Vehicles Convergence Forum was established on Jun. 13, 2016 by MoLIT Governments (MoLIT, MoSIP, MoTIE, NPA), Research Institute, Industry, etc. Collaboration and Coordination with Automated Vehicle Symposium from

115 Thank you very much!!! Young-Jun MOON, Ph.D. Chief Director of National Transport Technology R&D Center The Korea Transport Institute (KOTI) ISO/TC204 WG17 Convenor ITS Committee Member in TRB

116 CAV Traffic Signal Research Issues Extension of scalability of intersection control to networks in different demands. Interaction of control with routing and lane choice. Multimodal/heterogeneous considerations. How to partition the network (MFD), dynamically, - control (access control) Revisit traffic flow theory in the CAV environment Integration of applications Eco-Driving, speed harmonization, Managed lanes for CAV (benefits of lane management and control, corridors) where can it be effective? Network level considerations where network management strategies Use better information, e.g. Railroad crossing, to make management decision

117 CAV Traffic Signal Research Issues (cont) Trajectory control, optimization Control objective (delay), path based optimization Integration of network control and signal control Fault tolerance/robustness of control system System failures, weather, adaptability of difficult conditions High resolution data (Purdue) as a start to using high fidelity data How do we integration these two data sets to better utilize the data? Social commitments/attitudes to cooperative system operation. Human factors, behaviors Program that allows the two sides to connect

118 CAV Traffic Signal Research Issues (cont) Workforce education and development Changing the culture of traffic signal technicians/engineers ($) Perspective of planners, administration, operators,.system views/education Explore communication impacts on advanced traffic management E.g. DSRC range/interference versus other communication technologies. Economics of alternative comm technologies, Performance spec What could the communications portfolio for traffic control look like? Resolution (fidelity) for different applications to become a requirements spec for traffic control Level of automation for traffic control (level 1, 2, 3, 4, ) Individual vehicle characteristics or level of vehicle automation in signal control

119 CAV Traffic Signal Research Issues (cont) Geometric and service opportunities in control concepts to serve all the users. Changing the stop bar location to allow dynamics for control? Sorting the traffic (-pre-signal).dynamic map (local intersection map) change in lanes, stop bar. What happens if there is no connectivity, but only automated? What are the performance characteristics? What impact when the vehicles learn about signal operations/characteristics? Consideration of value of time in the control creating a market for priority. Purchased priority. Encouraging shared use.

120 CAV Traffic Signal Research Issues (cont) Establish baseline models/conditions for research analysis Estimation of the gains, improvements, expectations at different levels of control Factors include: market penetration rate, intelligence in the control system, constraints on the system (safety, ) understand trade-offs, Complexity/degrees of freedom of parameters less parameters make the system more usable Control strategies Vehicle control e.g create platoons, speed control, Human acceptance/considerations Driver expectancy, trust (in vehicle, in signal),

121 CAV Traffic Signal Research Issues (cont) Impact of large fleets (that are trying to optimization their own operation) in the overall system and their impact of competing control strategies Transportation network service providers Shared mobility view of the world

122 CAV Traffic Signal Research Issues (cont) Categories - Network User capabilities/characteristics Institutional issues Traffic flow theory Application scenarios Control algorithms/strategies Human factors Infrastructure adaptation Evolution from today to next generation Impact of shared mobility in traffic control