1. Oversaturated Freeway Flow Algorithm
Next Generation Simulation
Oversaturated Freeway Flow Algorithm
Alexander Skabardonis, Hwasoo Yeo
University of California, Berkeley
NGSIM Team Meeting
Washington DC
January 21, 2007

2. Outline Problem Statement Oversaturated freeway flow model
Car-following model
Lane changing model
Estimation of Parameters
Car-following parameters
Lane changing parameters
Implementation
Algorithm implementation
Verification
Validation

3. Problem Statement Existing Simulation Approaches
Do not accurately model oversaturated traffic conditions such as repeated stops and starts, increased lane changes to position onto perceived “moving” lanes, or in the presence of tall vehicles, and large vehicle headways
Use additional rules and parameters to the basic “desired headway” and “gap-acceptance” based car-following and lane changing models
Introduce a large number of parameters that generally cannot be readily observed in the field

4. Project Overview Objectives:
Develop an improved model for oversaturated freeway flow
Focus on both car-following and lane-changing behaviors during congested conditions
Evaluation of Model for “non-oversaturated” Flow
Predict when & how non-oversaturated conditions breakdown
Coordination
Completed and ongoing NGSIM algorithms
Software Developers
Stakeholder participation in expert panels

5. Car-following: Concept
Vehicle will change speed according to the current spacing
sjam s v
1/t
speed
spacing vf v
sjam t
Alternative Distance form

6. Car-following base model concept -acceleration case
xCF(t+dt)=xLeader(t+dt- t)-sjam
Leader
v, speed v2 vf CF s2 v1 v2 s1 v1
Acceleration
s, spacing s1 s2

7. Car-following base model concept -deceleration case
xCF(t+dt)=xLeader(t+dt- t)-sjam
v, speed vf v2
Deceleration s2 v1 s1 v1 v2 s2 s1
s, spacing

8. Car-following Model Newell’s simplified CF model Maximum Acc
Jam Spacing
Wave travel time
Newell’s simplified CF model
Maximum Acc
Free Flow speed
Safety Constraint
Maximum Dec

9. Lane Changing Model Lane changing choice model Gap acceptance model
Mandatory Lane changing
Entry, exit
Discretionary lane changing
Gap acceptance model
Lane changing mechanism
Lane changing
Lane changing with cooperation
Emergency Lane changing
Lead gap
Lane changing
Vehicle n
Lag gap
n+1′
cooperation vehicles

10. Lane changing choice model (1) Mandatory lane changing
Applied to turning and exit traffic
Emergency Lane changing
lExit T0
Exit location En
Desired location in exit lane
Number of lane changes needed Tn
Stopping distance

11. Lane changing choice model(2) Discretionary Lane Changing
Lane changing probability is the function of speed difference
Probability that vehicle n at time t moves from lane l to lane l’
Ф : sensitivity to speed change

12. Gap acceptance conditions
Lead gap condition
Lead gap
(Minimum gap condition) n
Lag gap
(Safety condition)
Lag gap condition
(Minimum gap condition)
n+1′
(Safety condition)

13. Lane Changing Mechanism (1)
Car-following rule for Lane changing pending
x=MIN( xCF(Leader) , xCF(Veh Down) )
x= xCF(Leader)
Try to pass and find next gap
upstream vehicle
downstream vehicle
Veh up
Veh down LC
Leader
Request Cooperation
Conflict point

14. Lane Changing Mechanism (2)
Car-following rule for cooperating vehicle
x=MIN( xCF(Leader) , xCF(LC) )
End cooperation
upstream vehicle
Veh up
Leader
Coop LC
Request Cooperation
Conflict point

15. Lane Changing conflict
If there exist conflict in lane changing
x=MIN( xCF(Leader) , xCF(Veh Down) )
LC1 will yield to LC2
else x= xCF(Leader)
LC2 will pass LC1
upstream vehicle LC
downstream vehicle
Veh up
LC2
LC1
Leader
Request Cooperation

16. On-Ramp Model – conflict zone

17. On-Ramp Model Conflict Zone x=MIN( xCF(Leader) , xCF(Veh Down) )
Find a downstream vehicle
Veh down
Leader
Conflict point

18. Short Gap Mode (optional)
Lane changing cooperation vehicles keep short gap before and after lane change
Applied to Lane Changing or Cooperation vehicles
Ends when the leader vehicle starts acceleration
v, speed
Leader
Short gap mode
Short gap mode
s, spacing

19. Short gap mode

20. Effects of Short Gap (Relaxation)
Merging area

21. Parameter Estimation Car-following Parameters Lane Changing Parameters
vf : free-flow speed
gjam : jam gap
t: wave travel time
au, al : Maximum acceleration and deceleration
Lane Changing Parameters
Ф: sensitivity to speed difference
T : slope of lane changing for exit lane change
E: target location in the exit lane for exit lane change
mn: perceived HOV activation time
Find starting times of vehicles exiting HOV

22. Free Flow Speed Extract from detector data 5min detector data
30sec detector data

23. Jam gap g= xn-1-ln-1-xn Jam gap Minimum gap at jam condition
vehicle speed <3km/hr

24. Wave Travel Time
t: time between the action of leader vehicle and the following vehicle
150
200
250
300
350
400
450
500 -2 2 4 6 8 10 12 14 t
time(sec)
Speed(ft/sec)
Acc(ft/sec2)

25. Wave Speed Wave speed = sjam / t = (ㅣ+gjam) / t US101 I-80
Mean=18.07km/hr,11.22miles/hr
I-80
Mean=19.59km/hr,12.18miles/hr

26. Max Acceleration & Deceleration
Extracted from NGSIM trajectories data US101
Passenger cars (n=4163)
Mean=4.516(m/s2), std=
mean= (m/s2), std=0.827

27. Sensitivity to speed difference(φ) in discretionary lane changing
Assume constant φ
Probability of LC

28. Exit Lane Changing Parameters (E and T)
Exit location

29. Parameters - Car-following (1)

30. Parameters - Car-following (2)

31. Parameters – Lane Changing

32. Algorithm Implementation (1)
Mode?
Car-following
Lane Changing
Cooperating
Change Mode
Car-following for Cooperating
Gap Acceptable?
Yes No
Change lane
Lane Changing?
Emergency Lane Changing?
Yes
Yes
Car-following rule No
Maximum deceleration No
Change Mode to CF
Find cooperating vehicle and set mode
Car-following rule
Change mode of the cooperating vehicle to CF
Car-following for lane changing pending
End

33. Algorithm Implementation (2)
AIMSUN SDK
Replacing AIMSUN car-following and lane changing model
Programming language: C++
A2VehicleBehavioralModel
A2Vehicle
A2VehicleModelTestCreator
A2VehicleBehavioralModelTest
A2VehicleTest
dll entry point for new model
Car-following and lane changing Logic
Vehicle behaviors

34. Algorithm Verification – US101

35. Algorithm Verification: Exit Lane Changing

36. Model Testing I-80 (1)
Simulation 2:30PM-3:00PM
Detector 7

37. Model Testing I-80 (2)

38. Characteristics of the Proposed Algorithm
Consistent with kinematic wave theory
Parameters
Physical meaning
observable
Microscopic: wave travel time, jam gap, free flow speed, Max Acc, Dec
Macroscopic:
wave travel time, jam density, free flow speed
Can be used for model calibration
Mechanism Oriented approach
Integrated Algorithm with Car-following and Lane changing

39. Next Steps Complete Model Validation Software Code & Documentation
Trajectory data
Aggregate data
Software Code & Documentation
Final Report

40. Possible Future Enhancements
Relaxation Process
Partially implemented
Caused by driver characteristic change
Sampling interval for speed change
Jam gap, reaction time
Traffic Hysteresis
Caused by asymmetric behavior
Oscillations in speed-spacing
Not considered
Asymmetry model

41. Components of the suggested algorithm
Oversaturated Freeway flow
algorithm
Car-following
Lane changing
LC Choice model
Base Car-following
Coop Car-following
Car-following
for Conflict
Cooperation choice
Forced LC choice
CF model before LC
LC pending
Car-following
LC Target choice
Gap Acceptance
LC Car-following
Apply LC model
After LC
Car-following
CF model after LC
Emergency LC
Car-following `