Assessing the Effectiveness of Managed Lane Strategies for the Rapid Deployment of Cooperative Adaptive Cruise Control Technology

Size: px

Start display at page:

Download "Assessing the Effectiveness of Managed Lane Strategies for the Rapid Deployment of Cooperative Adaptive Cruise Control Technology"

Berenice Cain
5 years ago
Views:

1 Assessing the Effectiveness of Managed Lane Strategies for the Rapid Deployment of Cooperative Adaptive Cruise Control Technology Zijia Zhong Ph.D. Candidate Dissertation Advisor (Prof. Joyoung Lee) New Jersey Institute Of Technology

2 Biographic Sketch B.Eng. in Environmental Engineering, Jinan University, China Journal Papers Z. Zhong, J. Lee, and L. Zhao. "Multi-objective Optimization Framework for Cooperative Adaptive Cruise Control Vehicles in the Automated Vehicle Platooning Environment." Transportation Research Record: Journal of the Transportation Research Board : pp. 3-4 J. Lee, Z. Zhong, Bo Du, Slobodan Gutesa, and Kitae Kim. "Low-Cost and Energy-Saving Wireless Sensor Network for Real-Time Urban Mobility Monitoring System." Journal of Sensors, vol M.S. in Civil Engineering, NJIT Peer-review Conference Papers Z. Zhong and J. Lee. Simulation Framework for Impacts Assessment of Cooperative Adaptive Cruise Control Managed Lane Strategies. the 97th Transportation Research Board Annual Meeting, 018 (Submitted) Z. Zhong, J. Lee, B. Singh, B. Dimitrijevic, S. Chien, and L. Spasovic. "Evaluation of Freeway Merging Assistance System Using Driving Simulator." the Intelligent Transportation System World Congress 017, Montreal, Canada Z. Zhong and J. Lee. "Development of CID-free Hardware-in-the-Loop Simulation Framework." the 96th Transportation Research Board Annual Meeting, LMW Engineering Group, LLC Z. Zhong, J. Lee, and L. Zhao. "Evaluations of Managed Lane Strategies for the Arterial Deployment of Cooperative Adaptive Cruise Control." the 96th Transportation Research Board Annual Meeting, 017 J. Lee, Z. Zhong and L. Zhao. "Multi-objective Optimization Controller for Cooperative Adaptive Cruise Control." the 95th Transportation Research Board Annual Meeting, Present Ph.D. Candidate, NJIT 017 Outstanding Graduate Student Award 015 The Future of ITS-NJ Scholarship Award J. Lee, Z. Zhong, K. Kim, B. Dimitrijevic, B. Du, and S. Gutesa. "Examining the Applicability of Small Quadcopter Drone for Traffic Surveillance and Roadway Incident Monitoring." the 94th Transportation Research Board Annual Meeting, 015 J. Lee, Z. Zhong, B. Singh, B. Dimitrijevic, K. Kim, B. Du, and S. Chien. "WIMAP: Work Zone Interactive Monitoring Application." the 94th Transportation Research Board Annual Meeting, 015 J. Lee and Z. Zhong. "Estimation of Real-time Origin-destination Flow Using Mobile Sensor Network." Intelligent Transport Systems World Congress 014, Detroit, MI. 014

Research Activities C/AV Research Projects Technical

Deployment (NJDOT Sponsored), 017~present Dedicating Lanes

Vehicles (NCHRP Sponsored) 017~present Introduction of

Harmonization (FHWA Sponsored), 013-014 Saxton Laboratory

for Research on Automated Longitudinal Vehicle Control

3 Research Activities C/AV Research Projects Technical Support for Connected Vehicle Technology Evaluation and Deployment (NJDOT Sponsored), 017~present Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles (NCHRP Sponsored) 017~present Introduction of Cooperative IV (Vehicle- Highway) Systems to Improve Speed Harmonization (FHWA Sponsored), Saxton Laboratory Simulation Capability (FHWA Sponsored), Simulation for Research on Automated Longitudinal Vehicle Control (FHWA Sponsored, source code available on OADP), Research Activities

optimization-based (MOOP) control algorithm Wireless

4 Presentation Outline Introduction Literature Review Integrated Simulation Framework Multi-objective optimization-based (MOOP) control algorithm Wireless communication (WC) module Managed lane (ML) strategy Conclusion & Future Research

5 INTRODUCTION

Introduction Literature Operation Framework Optimization Case Study Framework Conclusion Conclusion The Scale of

6 million crashes in 015 * leading cause of death for ages 1-44** Mobility*** 6.

1 billion gallon of fuel wasted 60 billion pounds of additional CO *DOT HS 81 318, Traffic Safety Facts, National

6 Introduction Literature Operation Framework Optimization Case Study Framework Conclusion Conclusion The Scale of the Problem Safety 35 thousand highway deaths & 3.6 million crashes in 015 * leading cause of death for ages 1-44** Mobility*** 6.9 billion hours of travel delay $160 billion congestion cost Environment*** 3.1 billion gallon of fuel wasted 60 billion pounds of additional CO *DOT HS , Traffic Safety Facts, National Highway Traffic Safety Administration (August 016) **10 Leading Causes of Death by Age Group, United States 014, Centers for Disease Control and Prevention ***015 Urban Mobility Scorecard, Texas A&M Transportation Institute and INRIX (August 015)

7 Introduction Literature Operation Framework Optimization Case Study Framework Conclusion Conclusion Public Road Mileage & Lane Miles Vehicle Miles of Travel Vehicle Miles of Travel (VMT) Demand vs. Supply Million 9.0 Public Road Mileage, Lane-Miles, and VMT Trillion Lane Miles Vehicle Mile of Travel Public Road Mileage Year

human error Fundamentally increases highway carrying capacity

8 The Opportunity Cooperative Adaptive Cruise Control () Enhances driver safety considerably by reducing or even eliminating human error Fundamentally increases highway carrying capacity Offers high-performance driving even under restricted driving condition

Near-term Deployment Here Today The USDOT has initiated the CV

C/AV technologies Level 1 Level Phase 1 Phase Phase 3 In Testing

Evaluate(018-01) Level 3 Level 4 Level 5 Someday(?

9 Near-term Deployment Here Today The USDOT has initiated the CV Pilot Deployment Program with 3 major pilot sites for testing C/AV technologies Level 1 Level Phase 1 Phase Phase 3 In Testing Pre-deployment ( ) Develop and Deploy( ) Operate and Evaluate(018-01) Level 3 Level 4 Level 5 Someday(?) The near-term large-scale deployment of can be facilitated by using managed lane strategy

10 Research Landscape

11 LETERATURE REVIEW

12 Field Experiment 1996 KONVOI 1997 Automated Highway System Demonstration 008 Energy ITS Project KONVOI 009 Safe Road Train for the Environment Energy ITS Project GCDC 011 Grand Cooperative Driving Challenge 015 USDOT CV Pilot Deployment Program 016 European Truck Platooning Challenge AHS Demo ETPC

13 Operation Evaluation via Simulation Most of the studies only used hypothetical networks The longitudinal control of vehicles do not include a platoonlevel cooperation where all the vehicles intelligently negotiate, and make necessary concessions to diverse platoonlevel goals Few simulations implemented local coordination where stand-alone vehicle actively seeking for platoon opportunity Research Zhong et al. (017) Li et al. (017) Songchitr uksa et al. (016) Lee & Park (014) Arnaout and Bowling (014) Transportation Network 5-km -lane signalized arterial, Fairfax Co. Parkway, VA Hypothetical 15- km freeway w/ 1 on-ramp 4-km 3-lane, I-30 segment, TX (uncalibrated) Hypothetical15- km, 4-lane freeway w/ 1 onramp Hypothetical 6- km 4-lane U- shape highway w/ 1 on-ramp Longitudinal Control Predetermined accelerations with conditional switching IDM factoring in the leader and nd vehicle Transfer function based on the preceding vehicle(s) Predetermined accelerations with conditional switching Intelligent Driver Model (IDM) Lateral Control Rear-, mid, front joining No implemented Rear-joining Rear-, mid, front joining MOBIL(Kesting et al. 007)

14 under Imperfect Communication Traffic simulator provide realistic vehicle behavior & interactions, whereas network simulator deals with packet-level simulation with various communication protocol Some studies only focus on the packetlevel communication simulation and did no deal with the vehicle movements. Some studies post-process the wireless simulation and as such the wireless communication has no influence on vehicle movement Synchronizer has been used in some of the studies, but the simulation scale are limited due to computational burden Research Akhtar et al. (015) Chen et al. (009) ElBatt et al. (006) Network Simulator Mathmath cal models Traffic Simulator SUMO Synchronizer N/A (Post process) Scale 6000 vph ns- N/A N/A 400 vehicles QualNet N/A N/A (Post process) Xu et al. (004) ns- SHIFT N/A (Post process) Ucar et al. (016) Motro et al. (016) Segata et al. (014) Veeraraghavan and Miloslavov (01) Eichler et al. (005) 190 vehicles 1,000 vehicles OMNeT++ SUMO TCP/IP 10 vehicles OMNeT++ SUMO TCP/IP 3 vehicles OMNeT++ SUMO TCP/IP 4 vehicles NCTUns Vissim TCP/IP 300 vph ns- CARISMA TCP/IP 900 vehicles

15 Research Goal & Objectives The primary goal of this dissertation is to evaluate the impacts of introducing vehicles into the existing roadway and the effectiveness of using managed lane strategies. 1. To formulate a platoon-wide multi-objective based control algorithm that is able to accommodate various operational objectives.. To develop a realistic microscopic simulation test bed for testing under mixed traffic condition 3. To assess a variety of available managed lane strategies for under imperfect wireless communication. 4. To investigate the suitable performance measures for traffic.

16 Integrated Simulation Framework

17 1 Simulation Framework 3 0 Vissim Wiedemann Carfollowing model Non- Vehicles COM: Common Objective Model Interface API: Application Programming Interface Route Choice Decision Transportation Network Vehicles Driver Model API Wireless Communication Module COM O-D Monitoring Switch-off Vehicle Control Platooning/ Clustering algorithm MOOP- Controlling Algorithm Initialize Simulation Network Set Network Parameters

18 Integrated Simulation Framework Multi-objective optimizationbased (MOOP) control algorithm Wireless communication (WC) module Managed lane (ML) strategy

19 1 Multi-objective Optimization 3 0 Multi-objective optimization (MOOP) seeks to optimize multiple conflicting objectives The MOOP framework allows vehicles to conduct cooperative driving

20 1 SOOP vs MOOP 3 0 Difference SOOP MOOP Goals Search for an optimal solution Search for a Pareto frontier Maintain diversity of solutions in the Pareto frontier Search Space Artificial Approximation Decision variable space Need to construct a composite objective by fix-ups methods (e.g., weighted sum, scaling) Decision variable space and objective space the set of optimal solution can be obtained without applying fix-ups

21 Non-dominated Sorting Genetic Algorithm (NSGA-II) Initial Parent Evaluation Next Iteration Non-dominated sorting Crowding distance sorting Ranking/Fitne ss Assignment Selection Parent Population Pt F1 F Pt+1 Pt+1 Pt+1 Crossover and Mutation F3 Offspring Evaluation Offspring Population Qt F4 Ranking/Fitne ss Assignment Rejected

22 1 MOOP Control Algorithm 3 0

23 1 MOOP Control Algorithm (Cont d) min{(headway) H h(t 1), n i 1 n xi( t 1) xi( t) (Jitter), x i e ( L 0 0 i, j x i x )) i j j e for x 0 (CO & Fuel) } 3 3 e (M 0 0 i, j xi xj ) i j e for x 0 subject to (min gap) (acceleration) (speed limit) xi ( t 1) x ( t 1) i i,min min i i i 0, i i i i h x x x x x x (E-IDM) f ( T, s ) x comfort i,min i,max max H i i target headway for vehicle i within a platoon, s h( t 1) headway of vehicle i at time interval t+1, s i n the total number of vehicles within a platoon adjustment coefficient for speed x ( t 1) acceleration of vehicle i at t+1,m s x ( t) acceleration of vehicle i at t, x i L e i, j M h comfort e i, j i,min comfortable threshold for acceleration,m s regression coefficients for emission and fuel consumption regression coefficients for emission and fuel consumption user-defined minimum headway for vehicle i, s x ( t 1) front distance for vehicle i to preceding vehicle i i at time interval t+1, m x ( t 1) speed for vehicle I at time interval t+1, m/s x x x i,min i,max min max minimum acceleration of vehicle i, m s maximum acceleration of vehicle i, m s minimum allowable speed on a particular roadway, m/s x maximum allowable speed on a particular roadway, m/s 0 3

24 1 Enhanced Intelligent Driver Model 3 0 The Enhanced Intelligent Driver Model (Kesting, 010) x x x( x xlead ) s0 xt x 1 ab a, xidm xcah x s xidm xcah (1 c) xidm c xcah b tanh( ), oth b erwise des 0 CAH x min( xlead, x), x ( ) min(, ) lead x xlead x xlead x xlead x min( xlead, x) ( x xlead ) ( x xlead ) min( xlead, x), otherwise x x acceleration of ego vehicle, m/s x x x x IDM CAH lead lead 0 * acceleration calculated by IDM model, m/s acceleration calculated by CAH component, m/s accleration of leading vehicle, m/s current speed of leading vehicle, m/s x current speed of ego vehicle, m/s x desired speed of ego vehicle, m/s x gap between the ego and leading vehicle, m s s des minimum distance, m effective minimum gap, s Heaviside step function a maximum accleration, m/s b desired deceleration, m/s c coolness factor T desired time gap, s free accelration exponent

25 Integrated Simulation Framework Multi-objective optimizationbased (MOOP) control algorithm Wireless communication (WC) module Managed lane (ML) strategy

1 Wireless Communication Module 3 0 Pr obability of Sucessful Reception: 4 3( x/ ) i R(,,, ) (1 i(, )( ) ) i 1 P x f e h Communication Density: f where, h (, ) two-demensional polynomial i f

26 1 Wireless Communication Module 3 0 Pr obability of Sucessful Reception: 4 3( x/ ) i R(,,, ) (1 i(, )( ) ) i 1 P x f e h Communication Density: f where, h (, ) two-demensional polynomial i f communication density transmission power, m equipped vehicle density, vehicle/km transmission frequency, Hz x The analytical model is derived from ns- simulator for 5.9GHz DSRC communication by Killat et al. (009) Channel access delay is assumed zero Same level of countering noise and interferences are assumed for different data rates

27 Integrated Simulation Framework Multi-objective optimizationbased (MOOP) control algorithm Wireless communication (WC) module Managed lane (ML) strategy

multifaceted managed lane facilities Access express lanes reversible

28 Managed Lane Pricing value priced lanes toll lanes HOT lanes Eligibility HOV lanes, truck lane, HOV used by other vehicle groups, multifaceted managed lane facilities Access express lanes reversible lanes busways, exclusive truck facility Increase complexity with active management

29 String Operations 1. Ad hoc clustering HOV HOV Mid joining A vehicle decelerates to create larger intra-platoon gap Rear joining Front joining is taking the exiting ramp, not seeking clustering. Local coordination 3. Global coordination HOV Traffic flow direction HOV

30 Ad Hoc Clustering of 30 40% MPR 50% MPR 60% MPR Simulation result for a 4-lane highway Managed lane(s), if implemented, start(s) from the leftmost lane The distributions of headway collected for either ML-1 or ML- become narrower

31 Traffic Volume Introduction Literature Framework Case Study Conclusion 100 % Suitable Managed Lane Strategy High Volume Dedicated Lane for Mixed Users (e.g., + HOV) Deployment Scenario Deployment Scenario 31 Managed lane can facilitate clustering to harness the short following distance enabled by VV communication The boundary (as pointed) under various traffic conditions could be determined via simulation No Dedicated Lane for Dedicated Lane for Market Penetration Rate

32 1 Managed Lane Strategy 3 0 Phase 1 Provide free SOV (single occupancy vehicle) use of the managed lane to encourage customers to purchase/retrofit vehicles until reaching unacceptable operating condition in managed lane Phase Allow platooning with safety gaps to share the managed lane with non- vehicles. Phase 3 Transitioning to dedicated lane once MPR warrants Allow automated high-performance driving (with higher cursing speed) Travel Lane ML Strategy 4 th (leftmost) 3rd nd 1st (rightmost) 1 HOV + HOV HOV + 5 HOV HOV 8 HOV

33 Case Study

1 Case Study: MOOP Algorithm 3 0 vehicle with predetermined trajectory vehicles controlled by one of the eight algorithms Non-/ vehicle Controlling Algorithm Short Designation Note Human H Vehicles

E-IDM E The vehicles are controlled by the E-IDM, which is a representation of ACC vehicle behaviors. SOOP-HW S-1 The vehicles are controlled by SOOP for Headway objective only.

34 1 Case Study: MOOP Algorithm 3 0 vehicle with predetermined trajectory vehicles controlled by one of the eight algorithms Non-/ vehicle Controlling Algorithm Short Designation Note Human H Vehicles are controlled by the Vissim Wiedemann car-following model. It yields realistic for human driving behaviors. E-IDM E The vehicles are controlled by the E-IDM, which is a representation of ACC vehicle behaviors. SOOP-HW S-1 The vehicles are controlled by SOOP for Headway objective only. SOOP-Jitter S- The vehicles are controlled by SOOP for Jittering objective only. SOOP-CO S-3 The vehicles are controlled by SOOP for CO emission objective only. MOOP-HW M-1 The vehicles are controlled by MOOP with Headway objective preferred. MOOP-Jitter M- The vehicles are controlled by MOOP with Jittering objective preferred. MOOP-CO M-3 The vehicles are controlled by MOOP with CO emission objective preferred.

35 1 Results The MOOP approach achieves better balance by accounting for the optimality of the remaining three objectives. M-1 also produces much less variation of the objective value. 3 0 The S-1 and M-1 can drastically increase highway carrying capacity due to more compact platoons.

36 1 Sensitivity Analysis 3 0 MOOP w/ Headway Preferred MOOP w/ Jitter Preferred MOOP w/ CO Preferred

37 1 Case Study: Managed Lane in Imperfect Wireless Communication 3 0 A 9.3-km 4-lane hypothetical network was created in Vissim with two interchanges located at mile marker km and 6 km respectively. The traffic demand for the network is higher than it typically is. No managed lane, 1 lane, and lane were tested. Wireless communication is evaluated at 10Hz

38 1 Speed-flow Diagram with 3 0 NML-30% NML-70% NML-90% ML1-70% ML1-90% Base: no vehicles NML: non managed lane ML1: one managed lane ML: two managed lane

39 1 Wireless communication module 3 0 This wireless communication module is able to provide communication density as well as probability of successfully communication for each individual vehicle The managed lane strategies impact for communication density can be observed

14-mile with 4 lanes in each direction With recurring congestion during

40 1 Realistic Large-Scale Network 3 0 The I-66 Segment, VA A major commuter corridor out of the beltway of Washington D.C. 14-mile with 4 lanes in each direction With recurring congestion during weekdays HOV lane implemented in the leftmost lane Has 6 interchanges and HOV exclusive exits

41 1 I-66 Network Calibration 3 0 Circle: field Cross: simulation Circle: field Cross: simulation Circle: field Cross: simulation Circle: field Cross: simulation Circle: field Cross: simulation Circle: field Cross: simulation Speed-flow Diagram (RTMS Data) Travel Time (INRIX)

42 Conclusions Within the proposed simulation network: The proposed MOOP control algorithm have been tested numerically. The algorithm offers a good balance and adaptive framework for platoon-wide cooperation The DSRC wireless communication module has been integrated to the Driver Model API and tested for its suitability for large-scale traffic simulation The effectiveness of certain managed lane strategies of deploying vehicles has been investigated The I-66 network has been properly calibrated for evaluation

43 Future Research To integrate the MOOP- algorithm in Vissim Driver Model API To conduct evaluation on the I-66 network To derive policy recommendations for using managed strategies To fine tune the simulation test bed (e.g., distributed computing, more efficient GA solver)

44 Contributions Development of an integrated simulation framework capable of Performing true cooperative driving based on multi-objective optimization Implementing high-definition longitudinal and lateral movements Simulating DSRC communication and its impact in simulation runtime in mixed traffic operation at a large-scale network Evaluation of deployment of vehicles Impact of managed lane strategy at network-level under mixed traffic condition Near-term deployment recommendations for stakeholders

45 Proposed Schedule

46 Thank you for your time Zijia Zhong Ph.D. Candidate Web:

Modeling Driver Behavior in a Connected Environment Integration of Microscopic Traffic Simulation and Telecommunication Systems.

Modeling Driver Behavior in a Connected Environment Integration of Microscopic Traffic Simulation and Telecommunication Systems Alireza Talebpour Information Level Connectivity in the Modern Age Sensor