Autonomous Vehicles and Traffic Physics: Framing

1 downloads 0 Views 12MB Size Report
The car-‐following model of Talebpour, Hamdar, and Mahmassani (2011) is used. • Accelera>on Behavior: Probabilis>c. • Percep>on of Surrounding Traffic.
Autonomous Vehicles and Traffic Physics: Framing the Questions Hani Mahmassani Northwestern University

Outline Ø  Motivation: Autonomous Vehicles, Connected Systems Ø  Flow Implications: Freeways §  Research Questions §  Simulation Approach: Traffic, Wireless Communication §  Stability Analysis: §  § 

Analytical Approach Simulation Results Trajectory Processor for particle-based simulators

§  Throughput Analysis: Simulation Results §  Lane Changing in Connected Environment: Game Theory

Ø  Flow Implications: Arterials and Urban Street Junctions Ø  Flow Implications: Networks §  NFD §  Fleet Routing – Objective functions, Moving towards SO using empty returns

Ø  Are we there yet?

2  

Connectivity   Connected  systems   (internet  of  everything)  

Smart   Highways   Cooperative   Driving    

Ad-­‐hoc     networks  

Coordinated  

Peer-­‐to-­‐Peer   (Neighbor)   Receive    only   Isolated  

-­‐  Optimized  flow   -­‐  Routing   -­‐  Speed  harmonization    

  Connected  

-­‐  Real-­‐time  info   -­‐  Asset  tracking   -­‐  Electronic  tolling    

   

Fully  manual     Level  0  

Autonomous   Vehicles    

INTELLIGENCE   RESIDES   ENTIRELY   IN  VEHICLE  

Fully  automated     Level  4  

Automation  

Questions Traffic Flow Implications: Freeways •  What  are  the  implications  of  connectivity  and/or  automated  functions  on   how  we  model  driver  behavior  and  traffic  interactions?   •  How  do  we  model  the  communications  aspects  (of  connected  systems)   jointly  with  the  traffic  flow  (e.g.  to  support  operational  control  design)?   •  What  are  the  implications  of  automation  vs.  connectivity  on  traffic  system   performance  in  terms  of        

SAFETY    THROUGHPUT  (“Capacity”)    STABILITY  (  !  Safety)    FLOW  BREAKDOWN  (Reliability)   SUSTAINABILITY  (Greenhouse  gases,  energy)    

•  What  kinds  of  collective  effects  could  we  expect?  Are  fundamental   diagrams  still  useful  for  this  purpose?   •  What  is  the  sensitivity  to  relative  market  penetration  on  impact  on  mixed   traffic  performance?   •  What  kinds  of  controls  should  agencies  be  contemplating?  

Work  in  collaboration  with  recent  PhD  graduate   Alireza  Talebpour     Currently  Assistant  Professor  at  Texas  A&M   University  

Accelera'on  Framework   No  Automa'on   Not  Connected  

No  Automa'on   Connected  

Self-­‐Driving   Not  Connected  

Accelera'on  Framework   Self-­‐Driving   Not  Connected  

No  Automa'on   Connected  

No  Automa'on   Not  Connected  

• 

AcceleraKon  Behavior:  

ProbabilisKc  

• 

PercepKon  of  Surrounding  Traffic   CondiKon:  

SubjecKve  

• 

ReacKon  Time:  

High  

• 

Safe  Spacing:  

High  

• 

High-­‐Risk  maneuvers:  

Possible  

•  The  car-­‐following  model  of  Talebpour,  Hamdar,  and  Mahmassani  (2011)   is  used.   • 

Probabilistic

• 

Recognizes two different driving regimes: •  Congested •  Uncongested

• 

Consider crashes endogenously

Accelera'on  Framework   No  Automa'on   Not  Connected  

Self-­‐Driving   Not  Connected  

No  Automa'on   Connected  

Ac've  V2V  Communica'ons  

InacKve  V2V  CommunicaKons  

AcKve  V2I  CommunicaKons  

InacKve  V2I  CommunicaKons  

• 

AcceleraKon  Behavior:  

DeterminisKc  

• 

PercepKon  of  Surrounding  Traffic  CondiKon:  

Accurate  

• 

ReacKon  Time:  

Low  

• 

Safe  Spacing:  

Low  

• 

High-­‐Risk  maneuvers:  

Very  Unlikely  

•  The  Intelligent  Driver  Model  (Treiber,  Hennecke,  and  Helbing,  2000)  is   used.  

Accelera'on  Framework   No  Automa'on   Not  Connected  

No  Automa'on   Connected  

Self-­‐Driving   Not  Connected  

AcKve  V2V  CommunicaKons  

Inac've  V2V  Communica'ons  

AcKve  V2I  CommunicaKons  

InacKve  V2I  CommunicaKons  

•  Sources  of  informaKon:  drivers’  percepKon  and  road  signs   •  Behavior  is  modeled  similarly  to  the  “No  AutomaKon  Not  Connected”.  

Accelera'on  Framework   No  Automa'on   Not  Connected  

No  Automa'on   Connected  

Self-­‐Driving   Not  Connected  

AcKve  V2V  CommunicaKons  

InacKve  V2V  CommunicaKons  

Ac've  V2I  Communica'ons  

InacKve  V2I  CommunicaKons  

•  TMC  can  detect  individual  vehicle  trajectories   •  Speed  harmonizaKon   •  Queue  warning  

•  Depending  on  the  availability  of  V2V  CommunicaKons:  

•  AcKve  V2V  CommunicaKons:  IDM   •  InacKve  V2V  CommunicaKons:  Talebpour,  Hamdar,  and  Mahmassani.  

Accelera'on  Framework   No  Automa'on   Not  Connected  

No  Automa'on   Connected  

Self-­‐Driving   Not  Connected  

AcKve  V2V  CommunicaKons  

InacKve  V2V  CommunicaKons  

AcKve  V2I  CommunicaKons  

Inac've  V2I  Communica'ons  

•  No  communicaKon  between  vehicle  and  TMC   •  Depending  on  the  availability  of  V2V  CommunicaKons:  

•  AcKve  V2V  CommunicaKons:  IDM   •  InacKve  V2V  CommunicaKons:  Talebpour  ,  Hamdar,  and  Mahmassani  

Accelera'on  Framework   No  Automa'on   Not  Connected  

No  Automa'on   Connected  

•  On-­‐board  sensors  are  simulated:  

Self-­‐Driving   Not  Connected  

•  SMS  AutomaKon  Radars  (UMRR-­‐00  Type  30)  with  90m±2.5%  detecKon   range  and  ±35  degrees  horizontal  Field  of  View  (FOV).  

Accelera'on  Framework   No  Automa'on   Not  Connected  

No  Automa'on   Connected  

Self-­‐Driving   Not  Connected  

•  Speed  should  be  low  enough  so  that  the  vehicle  can  react  to  any  event   outside  of  the  sensor  range  (v      max            )  (Reece  and  Shafer,  1993  and  Arem,   Driel,  Visser,  2006).  

Connected  Vehicles  Technology Communication  

•  It  is  essential  to  consider  the  V2V/V2I  communications  when  modeling  a   connected  environment.   •  Connectivity  through  the  vehicular  ad  hoc  network  (VANET)  is  a  key  element.   •  Several  studies  focused  on  connectivity  in  a  VANET,   • 

Jin  et  al.  (2011)  

• 

Ajeer  et  al.  (2011)  

• 

Durrani  et  al.  (2010)  

14  

Connected  Vehicles  Technology Communication  

•  Most  of  these  studies,   • 

Assume  homogenous  Poisson  distribution  for  vehicles  along  a  road  segment.  

• 

Consider  road  segments  as  one-­‐dimensional  objects.  

• 

Assume  normal  distribution  for  speed.  

•  It  is  essential  to  study  the  connectivity  of  VANET  by  considering   • 

Non-­‐homogenous  distribution  for  vehicles  along  a  road  segment.  

• 

Road  segments  as  two-­‐dimensional  objects.    

•  Existence  of  a  communication  link  between  two  nodes  depends  on,   • 

Wireless  technology  

• 

Transmission  power  and  rate  

• 

Distance  and  geographical  location  

• 

Signal  propagation  and  interference  

  15  

Communication  Network   Dynamic  Nature  of  Vehicular  Movements  

t  =  0  Sec  

t  =  100  Sec  

t  =  500  Sec  

t  =  1000  Sec   Based  on  NGSIM  Data  

16  

Communication  Network   Percolation

•  There  are  many  instances  in  which      a  fluid  spreads  through  a  medium,    a  disease  spreads  among  people,    information  spreads  in  social  networks,  and    a  liquid  penetrates  into  a  porous  material.  

•  Broadbent  and  Hammersley  (1957)  introduced  the  “percolation  theory”  to  model   these  instances.   •  There  are  two  models,  Discrete  Percolation  and  Continuum  Percolation  

•  Design  question:    how  to  form  clusters  of  communicating  vehicles,  with  a   “leader”  communicating  with  the  infrastructure  (V2I)  and  other  groups,  and   transmitting  information  within  the  group?  

17  

Clustering  Algorithm   What   i s   a   c luster?     •  Each  cluster  consists  of,   o  One  cluster  head   o  Several  cluster  members  

•  Assumption:  cluster  members  can  only  communicate  with  the  cluster  head   (1-­‐hop  communication  between  cluster  members).   •  A  cluster  head  can  communicate  with  cluster  members  and  other  cluster   heads  from  other  clusters.  

Having    stable  clusters  is  the  key  to  reducing  signal  interference.   This  study  incorporated  driving  history  and  driver  heterogeneity,  in  addition   to  the  usual  distance  and  speed  measures  into  VANET  clustering   algorithms.    

V2V  Communica'ons  Model   Clustering  

V2V  Communica'ons  Model   NS3  Implementa'on  

Network  Simulator  3  (NS3)  is  a  discrete-­‐event  communicaKon  network   simulator.     Dedicated  Short-­‐Range  CommunicaKon  (DSRC)  Protocol  is  the  standard   protocol  for  V2V  communicaKons.    DSRC  in  5.9GHz  spectrum.     DSRC  interface  uses  7  non-­‐overlapping  channels  (Xu  et  al.,  2012):   •  A  control  channel  with  1000m  range.   •  Six  service  channels  with  30-­‐400m  range.  

  DSRC  uses  

•  The  control  channel  to  send  safety  packets.   •  Service  channels  to  send  non-­‐safety  packets  (e.g.  Clustering  informaKon)  

V2V  Communica'ons  Model  

NS3  Implementa'on  –  Clustering  Frequency   Packet  size  =  50  byte:  LocaKon,  speed,  acceleraKon   Packet  Forwarding  Overhead  =  10  ms  (Koizumi  et  al.,  2012)  

Communication  Network  

Connectivity:  C onstant  Transmission   Power    

Communication network at t = 0 Sec

EffecKve  Transmission  range  =  5m   Biggest  Cluster  Size  =  8   EffecKve  Transmission  range  =  10m   Biggest  Cluster  Size  =  93  

 DSRC  in  5.9GHz  spectrum.   EffecKve  Transmission  range  =  20m   Biggest  Cluster  Size  =  216  

22  

V2V  Communica'ons  Model  

NS3  Implementa'on  –  Packet  Delivery  

Outline Ø  Motivation: Autonomous Vehicles, Connected Systems Ø  Flow Implications: Freeways §  Research Questions §  Simulation Approach: Traffic, Wireless Communication §  Stability Analysis: §  § 

Analytical Approach Simulation Results Trajectory Processor for particle-based simulators

§  Throughput Analysis: Simulation Results §  Lane Changing in Connected Environment: Game Theory

Ø  Flow Implications: Arterials and Urban Street Junctions Ø  Flow Implications: Networks §  NFD §  Fleet Routing – Objective functions, Moving towards SO using empty returns

Ø  Are we there yet?

24  

Stability  Analysis   •  Local  Stability  vs.  String  Stability   Stability   String  Local   Stability  

Wilson,  R.  E.,  and  J.  A.  Ward.    (2010)   Car-­‐following  models:  fifty  years  of  linear  stability  analysis  –     a  mathematical  perspective.     Transportation  Planning  and  Technology  

Stability  Analysis   A  car-­‐following  model  can  be  formulated  as:         Empirical  observaKons  suggest  that  there  exists  an  equilibrium  speed-­‐spacing   relaKonship:     f ( s * ,0, V ( s * )) = 0, ∀s * > 0     A  platoon  of  infinite  vehicles  is  string  stable  if  a  perturbaKon  from  equilibrium   decays  as  it  propagates  upstream.  

Stability  Analysis  

String Stable Regime

Unstable Oscillatory Regime

Unstable Collision Regime

Stability  Analysis   Following  the  definiKon  of  string  stability,  the  following  criteria  guarantees   the  string  stability  of  a  heterogeneous  traffic  stream  (Ward,  2009):   2   ⎡ f n 2 ⎤ ⎡ ⎤ ⎢ v − f Δnv f vn − f sn ⎥ ⎢ f sm ⎥ < 0   2 ⎥⎦ ⎢⎣ m ≠ n ⎥⎦ n ⎢ ⎣   where    





Stability  Analysis  

Heterogeneous  Traffic  Flow  

•  Parameters  of  regular  vehicles   are  adjusted  to  create  a  very   unstable  traffic  flow.   •  Low  market  penetraKon  rates  of   automated  vehicles  do  not  result   in  significant  stability   improvements.   •  At  low  market  penetraKon  rates   of  automated  vehicles,   Market  penetra'on  rate     of  connected  vehicles  

Automated,  Connected,  and     Regular  Vehicles  

Stability  Analysis  

Simulation  S egment   –   R ing   Road  

•  200  vehicles  with  40  meters  initial  spacing.   •  To  create  perturbation:   One  vehicle  is  slowed  down  to  v  =  1  m/s   with  maximum  deceleration  (-­‐8  m/s2).   Speed  is  kept  at  1  m/s  for  50  s.    

Stability  Analysis   Ring   Road  A nalysis  

No  Automation   Connected  

No  Automation   Not  Connected  

Self-­‐Driving   Not  Connected  

Stability  Analysis   Ring   Road  A nalysis  

No  Automation   Connected  

Market  Penetration  Rates  of  Connected  Vehicles:     10%                                                          50%                                                                                  90%                                                                                          

Self-­‐Driving   Not  Connected  

Market  Penetration  Rates  of  Autonomous  Vehicles:     10%                                                          50%                                                                                  90%                                                                                          

Stability  Analysis   Simula'on  Results  

A  one-­‐lane  highway  with  an  infinite  length  is  simulated.   String  Stability  as  a  FuncKon  of  ReacKon  Time  and  Platoon  Size  is  invesKgated.    

Regular  

Oscilla'on  Regime   Collision  Regime  

10%  Connected  

90%  Connected  

10%  Automated  

90%  Automated  

Stability  Analysis   Simula'on  Results  

OscillaKon  and  collision  thresholds  increase  as  platoon  size   decreases.     OscillaKon  and  collision  thresholds  increase  as  market   penetraKon  rate  increases.     At  high  market  penetraKon  rates,  Autonomous  vehicles  have   more  posiKve  effect  on  both  oscillaKon  and  collision   thresholds  compared  to  connected  vehicles.   10%  Automated  

90%  Connected  

90%  Automated  

Stability  Analysis   Summary  

The  presented  acceleraKon  framework  is  string  stable.     AnalyKcal  invesKgaKons  show  that  string  stability  can  be  improved  by  the  addiKon  of   connected  and  automated  vehicles.   • 

Improvements  are  observed  at  low  market  penetraKon  rates  of  connected  vehicles  (unlike   automated  vehicles).  

• 

At  high  market  penetraKon  rates,  automated  vehicles  have  more  posiKve  impact  on   stability  compare  to  connected  vehicles.  

  SimulaKon  results  revealed  that    

 

• 

OscillaKon  and  collision  thresholds  increase  as  platoon  size  decreases.  

• 

OscillaKon  and  collision  thresholds  increase  as  market  penetraKon  rate  increases.  

• 

Automated  vehicles  have  more  posiKve  impact  on  stability  compare  to  connected  vehicles.  

Outline Ø  Motivation: Autonomous Vehicles, Connected Systems Ø  Flow Implications: Freeways §  Research Questions §  Simulation Approach: Traffic, Wireless Communication §  Stability Analysis: §  § 

Analytical Approach Simulation Results Trajectory Processor for particle-based simulators

§  Throughput Analysis: Simulation Results §  Lane Changing in Connected Environment: Game Theory

Ø  Flow Implications: Arterials and Urban Street Junctions Ø  Flow Implications: Networks §  NFD §  Fleet Routing – Objective functions, Moving towards SO using empty returns

Ø  Are we there yet?

36  

THROUGHPUT AND FLOW-DENSITY SIMULATION SEGMENT

The average breakdown flow in a series of simulations is considered as the bottleneck capacity.

THROUGHPUT and SPEED-DENSITY RELATION SENSITIVITY ANALYSIS – MIXED ENVIRONMENT

10% R– 0% C – 90% A

10% R– 50% C – 40% A

10% R– 20% C – 70% A

10% R– 70% C – 20% A

10% R– 40% C – 50% A

10% R– 90% C – 0% A

THROUGHPUT SIMULATION RESULTS •  Low market penetration rates of autonomous and connected vehicles do not result in a significant increase in bottleneck capacity. •  Autonomous vehicles have more positive impact on capacity compare to connected vehicles. •  Capacities over 3000 veh/hr/lane can be achieved by using autonomous vehicles. Autonomous, Connected, and Regular Vehicles

Important  Caveat       THERE  ARE  MANY  DIFFERENT  WAYS  OF  IMPLEMENTING  THE   TECHNOLOGIES,  ESPECIALLY  WITH  REGARD  TO  DRIVING  AND  FLOW   CONTROL.       Simulation  testbeds  can  help  evaluate  alternatives  and  examine  implications.    

OUR  PROPOSAL     AGENCY  

TELCO   No  Automa'on   Connected  

No  Automa'on   Not  Connected  

Accel.  Choice   Lane  changing  

Comm.  Protocols   “Rules  of  engagement”  

Accel.  Choice   Lane  changing  

INTEGRATED  TRAFFIC-­‐TELECOM   SIMULATION  PLATFORM  

OEM  

Self-­‐Driving   Not  Connected  

Accel.  Choice   Lane  changing  

Outline Ø  Motivation: Autonomous Vehicles, Connected Systems Ø  Flow Implications: Freeways §  Research Questions §  Simulation Approach: Traffic, Wireless Communication §  Stability Analysis: §  § 

Analytical Approach Simulation Results Trajectory Processor for particle-based simulators

§  Throughput Analysis: Simulation Results §  Lane Changing in Connected Environment: Game Theory

Ø  Flow Implications: Arterials and Urban Street Junctions Ø  Flow Implications: Networks §  NFD §  Fleet Routing – Objective functions, Moving towards SO using empty returns

Ø  Are we there yet?

42  

Lane-­‐Changing  Framework   Talebpour,  Mahmassani  &  Hamdar  (2015).  Modeling  lane-­‐changing  behavior   in  a  connected  environment:  A  game  theory  approach.  Trans.  Res.  Part  C.  

Lane-­‐Changing  Framework  

It  is  assumed  that  V2V  can  provide  information  about  the  nature  of  lane-­‐ changing  maneuvers:    Discretionary  lane-­‐changing  vs.  Mandatory  lane-­‐changing       A  game-­‐theoretical  approach  is  adopted  with  the  following  pure  strategies:   •  • 

Lag  vehicle:  Accelerate,  Decelerate,  Change  Lane   Target  Vehicle  :  Change  Lane,  Do  not  Change  Lane  

Lane-­‐Changing  Framework     Inactive  V 2V  C ommunications  

Without  information,  drivers  are  uncertain  about  the  nature  of  other  drivers’   lane-­‐changing  maneuvers.     Two-­‐person  non-­‐zero-­‐sum  non-­‐cooperative  game  under  incomplete   information.     “Harsanyi  Transformation”  is  used  to  solve  the  game  with  incomplete   information:  

–  “Harsanyi  Transformation”  transforms  the  lag  vehicle’s  incomplete   information  about  the  nature  of  each  lane-­‐changing  maneuver  into  imperfect   information  about  the  move  by  nature.   –  “Nature”  as  a  player  chooses  the  type  of  each  lane-­‐changing  maneuver.   •  Lane-­‐changing  is  mandatory  with  probability  p  and  discretionary  with   probability  (1-­‐p)  

 

Lane-­‐Changing  Framework   Inactive  V 2V  C ommunications  

Discretionary  lane-­‐changing  game  in  normal  form  

Lag Vehicle

ACTION

B1 (Accelerate) B2 (Decelerate) B3 (Change Lane)

Target Vehicle A1 (Change Lane) A2 (Do not Change Lane) ( P11, R11 )

( P12 , R12 )

( P21 , R21 )

( P22 , R22 )

( P31, R31 )

( P32 , R32 )

Mandatory  lane-­‐changing  game  in  normal  form  

Lag Vehicle

ACTION

B1 (Accelerate) B2 (Decelerate) B3 (Change Lane)

Target Vehicle A1 (Change Lane) A2 (Do not Change Lane) ( P11,Q11 )

( P12 ,Q12 )

( P21,Q21 )

( P22 ,Q22 )

( P31,Q31 )

( P32 ,Q32 )

Lane-­‐Changing  Framework   Inactive  V 2V  C ommunications  

ACTION

Lag Vehicle

B1 (Accelerate)

A1M A1D

A1M A2D

( P11 ,

(

pQ11 + (1 − p) R11 )

B2

(Decelerate)

( P21 ,

B3

(Change Lane)

( P31 ,

(

pQ21 + (1 − p) R21 ) pQ31 + (1 − p) R31 )

(

Target Vehicle A2M A1D  

pP11 + (1 − p) P21 , pQ11 + (1 − p) R21 )

(

pP21 + (1 − p) P22 , pQ21 + (1 − p) R22 )

(

pP31 + (1 − p) P32 , pQ31 + (1 − p) R32 )

(

A2M A2D

pP12 + (1 − p) P11 , pQ12 + (1 − p) R11 )

( P12 ,

pP22 + (1 − p) P21 , pQ22 + (1 − p) R21 )

( P22 ,

pP32 + (1 − p) P31 , pQ32 + (1 − p) R31 )

( P32 ,

pQ12 + (1 − p) R12 )

pQ22 + (1 − p) R22 ) pQ32 + (1 − p) R32 )

Lane-­‐Changing  Framework   Active  V 2V  C ommunications  

With  information,  drivers  are  certain  about  the  nature  of  other  drivers’  lane-­‐ changing  maneuvers.    

ACTION Lag Vehicle

 

Two-­‐person  non-­‐zero-­‐sum  non-­‐cooperaKve  game  under  complete  informaKon.  

B1 (Accelerate) B2 (Decelerate) B3 (Change Lane)

Target Vehicle A1 (Change Lane) A2 (Do not Change Lane) ( P11, Q11orR11 )

( P12 , Q12orR12 )

( P21, Q21orR21 )

( P22 , Q22orR22 )

( P31, Q31orR31 )

( P32 , Q32orR32 )

Lane-­‐Changing  Framework   Payoff   F unctions  

Payoff  matrix  of  the  target  vehicle   Target Vehicle

ACTION Lag Vehicle

B1 (Accelerate) B2 (Decelerate) B3 (Change Lane)

 

A1 (Change Lane) η1. AccTC arg et + η 2 .ΔV + ε11 η1. AccTCarg et + η 2 .ΔV + ε 21

η2 .ΔV + ε 31

A2 (Do not Change Lane) 0 + ε 12 0 + ε 22 0 + ε 32

Payoff  matrix  of  the  lag  vehicle  

Lag Vehicle

ACTION

B1 (Accelerate) B2 (Decelerate) B3 (Change Lane)

Target Vehicle A1 (Change Lane) A2 (Do not Change Lane) C C η 3 . AccT arg et + δ11 η3 . AccLead + δ12 η 4 . AccTY arg et + δ 21 η4 .AccYLead + δ 22

η1. AccTCarg et ' + η 2 .ΔV + δ 31

 

AccTCarg et

AccYLead ΔV

:  Acceleration  to  prevent  collision  for  the  lag  vehicle  considering     the  target  vehicle  as  the  leader.   :  -­‐3.05  m/s2   :  Speed  difference  between  the  old  leader  and  the  new  leader    

Lane-­‐Changing  Framework  

Calibration   –   M ethod   o f  S imulated   M oments  

Lane-­‐Changing  Framework   Calibration   Results  

The  NGSIM  US-­‐101  dataset  from  7:50AM  to  8:05AM  is  used  for  calibration.     Discretionary Lane-changing Parameter Calibrated Value   η1 η2 η3 η4

-0.750

β

1.000

Mean Absolute Error (MAE)

0.383

0.875 -0.750 0.125

Mandatory Lane-changing Parameter

Calibrated Value

η1 η2 η3 η4

-0.875

β

1.000

Mean Absolute Error (MAE)

0.059

0.375 -0.625 0.25

Lane-­‐Changing  Framework   Simulation   –   F ictitious   P lay  

   

Lane-­‐Changing  Framework   Simulation  S egment  

3.5 Miles

Lane-­‐Changing  Framework   Simulation   Results  

Simulated  Period  

Loop Detector Data

Gap-Acceptance Model

MOBIL

Game Theory Based Model

Lane-­‐Changing  Framework   Simulation   Results  

Game Theory

Gap-Acceptance

0% MPR

50% MPR

100% MPR

Outline Ø  Motivation: Autonomous Vehicles, Connected Systems Ø  Flow Implications: Freeways §  Research Questions §  Simulation Approach: Traffic, Wireless Communication §  Stability Analysis: §  § 

Analytical Approach Simulation Results Trajectory Processor for particle-based simulators

§  Throughput Analysis: Simulation Results §  Lane Changing in Connected Environment: Game Theory

Ø  Flow Implications: Arterials and Urban Street Junctions Ø  Flow Implications: Networks §  NFD §  Fleet Routing – Objective functions, Moving towards SO using empty returns

Ø  Are we there yet?

55  

Coordination through connectivity and automation: Continuous-flow at-grade intersections

Traffic Flow/System Implications: Arterials and Urban Road Junctions •  What  are  the  implications  of  connectivity  and/or  automated  functions   on  how  we  model  driver  behavior  and  traffic  at  junctions?   •  Will  the  intersection  control  systems  be  as  smart  as  the  vehicles   and  the  communication  systems  would  allow  them  to  be?   •  New  approaches  needed  to  consider  local  logic  vs.  corridor  and   network-­‐level  coordination  (local  vs.  global  intelligence  much  more   relevant)   •  What  about  heterogeneous  traffic–  bicycles,  peds,  transit  vehicles,   delivery  vehicles  in  addition  to  cars  with  varying  degrees  of  autonomy   and  connectivity?   •  May  need  to  revise  notions  of  throughput,  and  performance   indicators.    

•  What  is  the  sensitivity  to  relative  market  penetration  on  impact  on   mixed  traffic  performance?  

Traffic Flow/System Implications: Networks •  What  is  the  effect  of  AV’s  and  CV’s  on  Network  Fundamental  Diagram   (NFD)  and  other  network-­‐level  relations?   •  What  is  an  appropriate  behavioral  notion  for  network  assignment?   -  Does  UE  still  make  sense?   -  Depends  on  mobility  ownership/sharing  model:  How  are  AV  fleet  routed?   -  Would  seek  to  minimize  traveler’s  time  in  transit  when  transporting  someone.   -  What  about  empty  vehicles  (repositioned  for  next  ride)?  Potential  for  SO  routing  

•  Strong  interplay  between  new  mobility  supply  models  and  modeling   network  impacts.    

•  Sensitivity  to  relative  market  penetration  not  only  of  fraction  of  AV’s   but  of  ownership/sharing  models.  

ARE WE THERE YET? WHO IS READY? 1. Technology is here and now; “Big Tech” and “New Tech” is in the lead– ready to market within 3-5 years.

2.  Automotive players– wide range (“waiting on standards”) •  •  •  • 

Connectivity in vehicles here and now; Driver-assist features already in high-end vehicles; Semi-autonomous in 3~5 yrs. Fully-autonomous: Special uses (freight, internal transit) by 2020

3.  System Integrators: more hype than deployment; not quite there yet. 4.  Insurance, Legal: surprisingly nimble 5.  LEAST READY: Public sector agencies; biggest hurdles on system aspects, public sector side. •  • 

More likely to see autonomous vehicles navigating through conventional infrastructure and control systems Limited evidence that agencies will leverage system benefits by transforming control and management systems.

6.  Many challenges ahead, and many more opportunities

We Love Feedback   Ques'ons/Comments   Email:  [email protected]  

  Follow  Me   Twiqer  @b_raKonal  

Connect  with  NUTC  

60  

Acceleration  Model  Parameters  and  Their  Typical  Values    for  car-­‐following  model  of    Talebpour  ,  Mahmassani  and  Hamdar  (2011)    

return  

Acceleration  Model  Parameters  and  Their  Typical   Values  for  IDM  (Treiber  et  al.  (2000)     Parameters          Typical  Value   Free  Acceleration  Exponent    δ=4   Desired  Time  Gap      T=1.5sec   Jam  Distance        s0=2.0m   Maximum  Acceleration    a=1.4m/s2   Desired  Deceleration      b=-­‐2.0m/s2