Paramics vs Vissim on Network Coding and Simulation Speed

Paramics vs Vissim

 

Introduction

This software review compares two microscopic traffic simulation packages that are commonly used by transportation professionals; Paramics version 6.8.0 and Vissim version 5.3-00. Paramics is developed by Quadstone Paramics Ltd, a Scottish company under Pitney Bowes Business Insight, and Vissim is developed by Planung Transport Verkehr (PTV), a German based company. Paramics and VISSIM both aim to accurately simulate vehicular traffic while presenting a friendly and intuitive user interface.

This software review evaluated Paramics and Vissim from two perspectives:

  1. Network Coding Time
  2. Simulation Calculation Time

To assess the first perspective, we coded the network and timed how long each coding task took. The second perspective was measured by actually timing the simulation speed using the same computer.

 

Study Area and Network Details

The study area we modelled was a 2.5 mile section of Interstate 5 in Southern California. The exact location is between south of the Avenida Vista Hermosa overpass and north of the Camino De Estrella overpass, as shown in the image below. This study area includes:

  • 4 and 5 mainline lanes, assuming the innermost lane is an HOV lane
  • 4 Highway off ramps
  • 5 Highway on ramps
  • 4 Signalized intersections at ramp terminals. Two of them are full actuated signals and another two are actuated coordinated signals. The signal timing data were obtained from California Department of Transportation (Caltrans).
  • 20 vehicle detection stations along the freeway

 

Map of the study area I-5 in southern California

 

Network Coding Time

 

Modeling preparation

The fundamental details of model development, as shown in Table below, are prepared before the development of the models.

 

Model details

Time Period

6-7AM

Background image

Obtained from Google Earth

Vehicle Type

SOV, HOV, Truck, Bus

Road Facilities

65 mph for freeway and 45 mph for arterials and ramps

Number of Zones

7

Demands

estimated using traffic data gathered from PeMS (https://pems.eecs.berkeley.edu) and observed turning count data

 

Network Coding

Two expert coders were assigned to code the study network under Paramics and Vissim, respectively. Both coders first discussed the coding requirements in both software and then summarized a list of tasks/steps that need to be followed to code a network. It is noted that the actual procedure to code a network in both software may vary.

Then, the coders were asked to recap or re-learn various modules (if needed) before each coding task, which was not counted in the coding time. We only counted the time taken to perform the actual coding tasks.

The tasks to code the network, the technical terms and coding details used by different software, and the time taken for each task are listed below. 

 

Category

Task

Paramics

Paramics Time (min)

Vissim

Vissim Time (min)

Network setup

1

Vehicle type and classes, Road Category, and Simulation time setup

2

 Vehicle type and classes, Road speed distributions,  Simulation time setup

9

Network Coding

2

Load overlay

5

 Load overlay

10

3

Skeleton

47

 Skeleton

54

4

Network Adjustments: curve, kerbs and stoplines

65

 Network Adjustments: Add shape points to better match background, reduced speed area

26

5

Speed limit (category defines speed limit)

0

 Add desired speed areas

12

6

Lane attributes (HOV lane restrictions)

 

3

 Lane closure (HOV lane restrictions)

8

7

Movement/Priority (nextlane, barred turns)

 

14

 Link-link connectivity

0

8

Signal Control: 2 fixed-time and 2 ASC, including detectors for signal

30

 Signal Control: RBC module for all 4 signals, added controllers, signal heads, detectors,  conflict areas, priorities

72

9

9. Infrastructure: link based detectors

14

9. Lane based detectors

30

Travel Demand & Route Choice

10

 Route choice(change configuration)

1

Add nodes to the network

 

9

11

Add zones

2

Add parking lots and zones

9

12

Load Demand: Use Demand editor to add demands, profile, matrix files

4

Load Demand: Manually generate multiple fma files to reflect demand variation using excel and 3rd-party text editor

7

Network Fine-tuning and Simulation

13

13. Network adjustment, local parameters, lane choices

 

40

 

 Network adjustment, local parameters, dynamic assignment simulation

 

30

 

14

Measurement (only collect general data)

1

Evaluation (only collect overall performance)

1

15

3D vehicle setup via PMX

2

3D vehicle setup

2

Total

 

 

230

 

279

 

It should be noted that for a beginner just learning the software, the time it takes for each task will be greatly increased. For example, one of colleague who knows very little about Paramics couldn't finish coding network in Paramics when other two expert coders coded their networks.

Here are a few notes about these tasks that deserve to be mentioned here:

  • The Network Setup in Vissim needed more manual work than Paramics
  • The Network Coding, specifically to Load the overlay, Vissim's background module is worse than Paramics. Slower to use.
  • Paramics has more details than Vissim and thus it takes longer time for network detail adjustment. Paramics has a curve link function but Vissim uses shape points along a straight link to mimic the curvature of a curve link. It is faster to code network skeleton and its details in Vissim.
  • Vissim doesn't have speed limit input but uses speed distributions. Modelers need to add desired speed areas to any links/connectors with a new speed limit setting in the real world.
  • In Vissim, a detector is lane based and thus it takes more time to add detectors in Vissim.
  • Vissim has more powerful actuated signal control module (called RBC) than Paramics. Paramics'  actuated signal control only implements a set of basic actuated signal control functions. Paramics also needs a better time-of-day signal control function.
  • Vissim is a path based simulator while Paramics is a link based simulator with some path features.
  • Vissim is more complex to setup for an OD based simulation, which is what we tried to do here.

The final coded network in Paramics can be seen here:

 

Screenshot of coded network in Paramics

 

In VISSIM, the coded network is shown below:

 

Screenshot of coded network in Vissim

 

Network Coding Time Comparison

 

Comparing the cumulative network coding time for each task between these two software's, it was found that the tasks take relatively the same amount of time up to Task 7: Movement/Priority (Paramics) & Link-link connectivity (Vissim). Among tasks 1 through 7, it was found that Paramics takes a shorter amount of time overall to prepare the network then Vissim, Paramics took more than twice as long, compared to Vissim, to modify the network to correctly match the overlay by making curves, kerbs and stopline adjustments.

For Tasks 8 and 9, however,

  • Vissim takes an additional 42 minutes than Paramics in coding signals in Task 8.
  • Vissim takes an additional 16 minutes to code detectors, due to Vissim requiring loop detectors to be entered in each lane rather than the whole facility as can be done in Paramics.

Graphing the cumulative time spent on each task, we can graphically see how these two software's compare for each task.

 

Paramics & Vissim Individual task time comparison

 

Simulation Time Comparison

 

After coding networks, we spent some time to roughly calibrate the networks in order for them to be in good shape to be simulated. We didn't compare the time taken for model calibration. We think some additional work is still needed to make models to be in better shape. For example, Paramics network is a little bit more congested than the Vissim network. One possible reason is that Vissim removed stuck vehicles but Paramics does not (the vehicle removal tool of Paramics can remove stuck vehicles during model development phase but it is not suggested being used during model production runs). 

The same computer was used to simulate Paramics and Vissim simulation models. The computer has an Intel i7 920 CPU running at 2.67 GHz. It had 8GB of RAM and a Nvidia Geforce 9800 GTX+ video card. The operating system was Windows Vista 64 Home Premium Service pack 2. During simulation runs, all cores of CPU are used to maximize the simulation speed.

The times taken to simulate Paramics and Vissim networks are shown in the graph below. We can see the actual simulation time is comparable for the simulation runs without displaying vehicles. However, when we rdisplay vehicles during simulation, Vissim takes much longer time compared to Paramics. This difference is most likely attributed to Paramics utilizing video card's hardware acceleration (i.e. OpenGL).

 

Paramics and Vissim Simulation calculation time

 

In summary VISSIM and Paramics both provide benefits and drawbacks. A more refined comparison will be conducted to tell you more about them.

 

Update:

PTV America sent us an email on 11/6/2012, saying that Vissim 6.0 or higher has much faster simulation speed with vehicle visualization. However, we didn't have the version of the software and thus won't be able to provide an update on how fast the simulation is under V6. Also, PTV America mentioned that V5 modelers need to avoid using the Aero theme.