The development of a HOV driver behavior model under Paramics

Will Recker, UC IrvineShin-Ting Jeng, UC IrvineLianyu Chu, CCIT-UC Berkeley

Existing HOV lane operation Continuous access Buffer-separate

Full-time Part-time

2

HOV Driver Behaviors

HOV drivers select HOV lane or mixed-flow lanes based on Distance to the exiting off-ramp, Familiarity of the traveled freeway, Driving aggressiveness, Travel time saving

Historically experienced or currently perceived or predicted traffic condition along the HOV lanes and mixed-flow lanes.

HOV lanes are utilized more when mixed-flow lanes get congested.

HOV lanes may still be used when the mixed-flow lanes are free-flow due to factors not related to travel time savings,

such as feeling more relaxed or comfortable on HOV lanes.

3

HOV Driver Behaviors Under Buffer-Separate HOV System HOV-eligible drivers may not enter HOV

lanes at the first available ingress point Traffic conditions may not allow them to

change all lanes successfully.

HOV drivers may not exit at the last possible egress point: Don’t know which egress point is the last one; Too congested on mixed-flow lanes Hard to change lanes successfully if they take

the last egress point

4

How to model HOV in Paramics Continuous access HOV

HOV lane Restriction HOV plugin

Buffer-separate HOV Modified version of the HOV plugin under V5.22 Separate links for HOV lane and mixed-flow

lanes HOV modeling is treated as a route choice problem Enable dynamic feed back assignment

5

Purpose

Develop a model to model HOV driver behaviors under buffer-separate HOV

system

Develop a model to have the potential to model HOV driver behaviors under continuous access HOV

system HOT driver behaviors

A research sponsored by Caltrans

6

HOV Driver Behavioral Model Include two models based on discrete choice

modeling: Preferred access choice model:

Examine travel time savings Traffic model

Calculate the acceptable gap to get in/out the HOV lane

Its Theoretical Capability Extended to: Real-world application via providing additional

“implementation” parameters, such as: Ingress/egress points selection control Traffic information update frequency

7

Mathematical Formulations

Preferred Ingress/Egress Choice Model:

Traffic Model:

8

1 1

exp 1

; 1,2, , 1 ; 1 ; 1max ,

car cark k

k k k kk l

k k k k

aL aLv

v vd v k n a b

v v v bv

gk

1 1 1 1

1 1 1 1

1

2 2 2 21

1F H F Hi i k k

F HOV F HOVi i k k

e e e eie

ik

V V V VP

1 1 1 1

1 1 1 1

1

2 2 2 21

1H F H Fi k i k i l i l

H F H Fi k i k i l i l

x x x xkx

i kl

V V V VP

UCI HOV Model Plug-in

Preferred Ingress/Egress Choice Model: Determine the preferences of ingress/egress

points Based on travel time measures

Cost (such as toll) can also be a parameter

Traffic Model: HOV-eligible driver act on the preferences via

checking: Distance between current location and ingress/egress Traffic intensity (density)

9

Model Parameters

Lcar : Average vehicle length Typically, assumed to be on the order of 20 ft.

(alpha): A scalar to reflect the mapping of travel time savings to the utility to enter/exit HOV lane < 0.

a : A parameter to be applied to estimate the spatial gap for a vehicle to change lane, which is a multiplier of a >= 1. b : The tolerable speed decrease fraction when a driver

makes a lane change 0.9 < b <= 1.

10

Implementation Parameters (1) Ingress Priority:

It is used to reflect the extent of the choice set for selecting favorable ingress points and the proportion of the drivers that will follow this rule.

Parameters: Favorable ingress point selection set Proportion of HOV drivers to follow the rule

For example, It could be specified that 80% of HOV drivers will

consider only the first three possible ingress points to enter a HOV lane given a specified route.

11

Implementation Parameters (2) Egress Persistency:

It is used to reflect the persistency of staying in the HOV lane before considering exiting at a favorable egress point, and the proportion of the drivers that will follow the rule.

Parameters Favorable egress point selection set Proportion of HOV drivers to follow the rule

For example, It could be specified that 90% of HOV drivers will

consider only the last two egress points prior to their assigned freeway exit to leave current HOV lane given a specified route.

12

Implementation Parameters (3) Preference Probability Threshold (PV_TH):

The values of the preference probabilities (i.e., Pi

e and Pxi+k) must be greater than PV_TH to be

considered as a candidate ingress/egress point.

Update Time Interval (PV_TIME): The time interval (seconds) to update the

traffic information.

13

Testing network: SR-5714

ParamicsModel

Lambert Rd

Chapman Ave

Sensitivity Analysis

(alpha): Parameter Lcar : 20 ft Parameter : -0.5 ≤ ≤ -10 Parameter a : 3.0 Parameter b : 0.9

a : Parameter Lcar : 20 ft Parameter : -7.0 Parameter a : 3.0 ≤ a ≤ 50 Parameter b : 0.9

15

Implementation Parameter Settings Ingress Priority:

80% HOV vehicles will only consider the first 4 ingress points (i.e. ingress points 1 to 4) to enter the HOV lane.

Egress Persistency: After entering the HOV lane, 80% HOV vehicles will

only consider the last 3 egress points (i.e. egress points 6 to 8) to leave current HOV lane.

Preference Probability Threshold: PV_TH > 0.01

Update Time Interval: PV_TIME = 30 sec

16

Different Values of alpha (1)

17

alpha = -1.5

0102030405060708090

100110120

1 2 3 4 5 6 7 8

Ingress/Egress ID

Num

ber o

f HOV

Veh

icles

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

Ingress Egress Ingress % Egress %

alpha = -3.0

0102030405060708090

100110120

1 2 3 4 5 6 7 8

Ingress/Egress ID

Num

ber o

f HOV

Veh

icles

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

Ingress Egress Ingress % Egress %

alpha = -7.0

0102030405060708090

100110

1 2 3 4 5 6 7 8

Ingress/Egress ID

Num

ber o

f HOV

Veh

icles

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

Ingress Egress Ingress % Egress %

alpha = -10.0

0102030405060708090

100110120130

1 2 3 4 5 6 7 8

Ingress/Egress ID

Num

ber o

f HOV

Veh

icles

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

Ingress Egress Ingress % Egress %

Different Values of alpha (2) It is found that more HOV vehicles take the

first access and the last access as the ingress and egress points respectively with larger settings of parameter.

This is because larger values of parameter imply that the HOV drivers derive greater benefit from any travel time savings associated with travel on the HOV lane.

The HOV drivers enter the HOV lane as soon as possible and stay in HOV lane until as close to their destination as possible.

18

Different Values of a (1)19

a = 20.0

0

20

40

60

80

100

120

1 2 3 4 5 6 7 8

Ingress/Egress ID

Nu

mb

er

of

HO

V V

eh

icle

s

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

80.0%

Ingress Egress Ingress % Egress %

a = 25.0

0

20

40

60

80

100

120

1 2 3 4 5 6 7 8

Ingress/Egress ID

Nu

mb

er

of

HO

V V

eh

icle

s

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

80.0%

Ingress Egress Ingress % Egress %

a = 30.0

0

20

40

60

80

100

120

1 2 3 4 5 6 7 8

Ingress/Egress ID

Nu

mb

er

of

HO

V V

eh

icle

s

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

80.0%

Ingress Egress Ingress % Egress %

a = 50.0

0

20

40

60

80

100

120

1 2 3 4 5 6 7 8

Ingress/Egress ID

Nu

mb

er

of

HO

V V

eh

icle

s

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

80.0%

Ingress Egress Ingress % Egress %

Different Values of a (2)

It is found that high usages of ingress point 1 (> 65%) are observed for all scenarios.

The usage of the egress point 6 increases as the value of parameter increases, due to the setting of the egress persistency implementation parameter.

20

Model Validation

The results collected at ingress point 1 were further investigated:

This location is selected because based on the nature of the proposed model, HOV drivers should prefer to take the first available ingress point to enter the HOV lane as long as They can benefit from the travel time savings Traffic conditions allow the lane changing to

HOV lane.

21

Assumptions

Model parameters alpha = -7.0 and a = 25.0 They are found to accurately replicate the real world

scenario. Only the HOV vehicles that completed their trips

within the simulation time are discussed here. Measures

The HOV lane usage at ingress point 1 are compared against flow and travel time.

Demand scenario Full demand 80% demand (free flow condition).

22

23

The HOV ingress point usage drops at 6:10am as the savings of travel time decreases. As the flow increases from 6:10am to 6:25am, the usage of the ingress point 1 continues to decrease due to insignificant travel time savings.

The freeway travel time starts increasing around 6:25am and reaches the peak around 6:35am, while the 5-min flow starts decreasing after 6:25am and the speed starts decreasing after 6:20am.

Although, as expected, the usage of ingress point 1 also increases while the congestion forms, it drops again after 6:35am, ostensibly:

The speed on the freeway mainline is low and the traffic is very congested

Thus, decreasing the opportunity for the HOV-eligible vehicles to execute the lane-changing maneuvers required to take the first ingress point to enter the HOV lane.

HOV Lane Usage, 5-min Flow, and Travel Time Savings (full demand)

24

HOV Lane Usage, 5-min Flow, and Travel Time Savings (80% demand)

Reasonable HOV lane usage

Usage of ingress point 1 drops as the savings of travel time decreases.

The relationship between ingress point usages and 5-min flow is not strong.

Consistent with the model formulation, with the demand of the mainline freeway scaled down, usage of ingress point 1 is greater than that in the case without applying demand scalar.

Comparing with Paramics Model

Demand: Full demand Dynamic feedback with 2

minutes update interval All HOV-eligible vehicles in the

network were affected by the dynamic feedback control:

The travel time savings on HOV lane may become less significant, since the HOV-eligible vehicles would stick to HOV lane as long as the travel cost is lower on HOV lane.

About 15% HOV-eligible vehicles re-entered the HOV lane after leaving it.

25

Paramics Model

0

20

40

60

80

100

120

140

1 2 3 4 5 6 7 8

Ingress/Egress ID

Num

ber

of H

OV V

ehic

les

ve

h

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

80.0%

90.0%

Ingress Egress Ingress % Egress %

alpha = -7.0

0102030405060708090

100110

1 2 3 4 5 6 7 8

Ingress/Egress ID

Num

ber o

f HOV

Veh

icles

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

Ingress Egress Ingress % Egress %

Comparison of Travel Time Savings

26

Paramics Model: Ingress Point 1

050

100150

200250

300350

5:45

:30 A

M5:

48:0

0 AM

5:50

:30 A

M5:

53:0

0 AM

5:55

:30 A

M5:

58:0

0 AM

6:00

:30 A

M6:

03:0

0 AM

6:05

:30 A

M6:

08:0

0 AM

6:10

:30 A

M6:

13:0

0 AM

6:15

:30 A

M6:

18:0

0 AM

6:20

:30 A

M6:

23:0

0 AM

6:25

:30 A

M6:

28:0

0 AM

6:30

:30 A

M6:

33:0

0 AM

6:35

:30 A

M6:

38:0

0 AM

6:40

:30 A

M6:

43:0

0 AM

6:45

:30 A

M6:

48:0

0 AM

6:50

:30 A

M6:

53:0

0 AM

6:55

:30 A

M6:

58:0

0 AM

Time

Trav

el Tim

e (se

c) (se

-50%

50%

150%

250%

350%

450%

Trav

el Tim

e Sav

ings S

avTravel Time SavingsHOV Section 2Fwy Section 2

UCI Plugin: Ingress Point 1

0

50

100

150

200

250

300

350

5:45

:30 A

M5:

48:0

0 AM

5:50

:30 A

M5:

53:0

0 AM

5:55

:30 A

M5:

58:0

0 AM

6:00

:30 A

M6:

03:0

0 AM

6:05

:30 A

M6:

08:0

0 AM

6:10

:30 A

M6:

13:0

0 AM

6:15

:30 A

M6:

18:0

0 AM

6:20

:30 A

M6:

23:0

0 AM

6:25

:30 A

M6:

28:0

0 AM

6:30

:30 A

M6:

33:0

0 AM

6:35

:30 A

M6:

38:0

0 AM

6:40

:30 A

M6:

43:0

0 AM

6:45

:30 A

M6:

48:0

0 AM

6:50

:30 A

M6:

53:0

0 AM

6:55

:30 A

M6:

58:0

0 AM

Time

Trav

el Tim

e (se

c) (se

-50%

50%

150%

250%

350%

450%

Trav

el Tim

e Sav

ings S

av

Travel Time SavingsHOV Section 2Fwy Section 2

Paramics Model: Ingress Point 2

050

100150200250300350400450500

5:45

:30 A

M5:

48:0

0 AM

5:50

:30 A

M5:

53:0

0 AM

5:55

:30 A

M5:

58:0

0 AM

6:00

:30 A

M6:

03:0

0 AM

6:05

:30 A

M6:

08:0

0 AM

6:10

:30 A

M6:

13:0

0 AM

6:15

:30 A

M6:

18:0

0 AM

6:20

:30 A

M6:

23:0

0 AM

6:25

:30 A

M6:

28:0

0 AM

6:30

:30 A

M6:

33:0

0 AM

6:35

:30 A

M6:

38:0

0 AM

6:40

:30 A

M6:

43:0

0 AM

6:45

:30 A

M6:

48:0

0 AM

6:50

:30 A

M6:

53:0

0 AM

6:55

:30 A

M6:

58:0

0 AM

Time

Trav

el Tim

e (se

c) ss

ss

-50%

50%

150%

250%

350%

450%

Trav

el Tim

e Sav

ings

savTravel Time Savings

HOV Section 3Fwy Section 3

UCI Plugin: Ingress Point 2

050

100150200250300350400450500

5:45

:30 A

M5:

48:0

0 AM

5:50

:30 A

M5:

53:0

0 AM

5:55

:30 A

M5:

58:0

0 AM

6:00

:30 A

M6:

03:0

0 AM

6:05

:30 A

M6:

08:0

0 AM

6:10

:30 A

M6:

13:0

0 AM

6:15

:30 A

M6:

18:0

0 AM

6:20

:30 A

M6:

23:0

0 AM

6:25

:30 A

M6:

28:0

0 AM

6:30

:30 A

M6:

33:0

0 AM

6:35

:30 A

M6:

38:0

0 AM

6:40

:30 A

M6:

43:0

0 AM

6:45

:30 A

M6:

48:0

0 AM

6:50

:30 A

M6:

53:0

0 AM

6:55

:30 A

M6:

58:0

0 AM

Time

Trav

el Tim

e (se

c) se

c

-50%

50%

150%

250%

350%

450%

Trav

el Tim

e Sav

ings s

as

Travel Time SavingsHOV Section 3Fwy Section 3

Speed-Density Relationship27

Paramics Model: Ingress Point 1

0

10

20

30

40

50

60

70

0 50 100 150 200 250

Density (veh/mile/lane)

Spee

d (m

ph)

UCI Plugin: Ingress Point 1

0

10

20

30

40

50

60

70

0 50 100 150 200 250

Density (veh/mile/lane)

Spee

d (m

ph)

Paramics Model: Ingress Point 2

0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160

Density (veh/mile/lane)

Spee

d (m

ph)

UCI Plugin: Ingress Point 2

0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160

Density (veh/mile/lane)

Spee

d (m

ph)

Summary of the Proposed HOV Model

Consider various aspects of HOV driver behaviors Traffic condition dependant Able to take the following parameters for

determining preferred ingress/egress points travel time measures driver aggressiveness, and/or Toll

Able to model continuous access HOV All links are ingress/egress points

Able to model HOT Add toll as part of travel time measure (generalized travel

cost)

28

Conclusion

The HOV model provides a solution to better HOV simulation Results show the reasonableness of the model under various

traffic conditions. The models’ parameters have good physical meanings The HOV model also demonstrated its feasibility and

applicability via setting various calibration parameters and control parameters.

Although promising, additional calibration and validation studies are needed to fully test the models developed. Still on the academic level Need to be easy to apply to general network implementation

29

30

References

Will W. Recker, Shin-Ting (Cindy) Jeng, Lianyu Chu (2010), A Behavioral Model for Ingress/Egress Decision on Buffer-separated HOV Facilities for Microsimulation Model, presented in TRB 2010

Will W. Recker, Shin-Ting (Cindy) Jeng, Lianyu Chu, Microsimulation Modeling of High Occupancy Toll (HOT) Concept in HOV Lanes, Caltrans RTA 65A0235 Final Report, August, 2009