It depends on what you have to do and/or for which purpose. There is an interesting article that you can read (http://www.massey.ac.nz/~kahawick/cstn/095/cstn-095.pdf) with six software compared. Also, by experience the not free software every year come up with new and interesting features.
Another free software is SUMO (http://sumo-sim.org/)
Based on my knowledge the most used licensed software are:
Macroscale - VISUM, CUBE (ex TRIPS), SATURN (UK),
Microscale VISSIM, AIMSUN, PARAMICS.
AIMSUN has recently launched the first available integrated Macro-Meso-Micro suite with the version 8.
I agree with the Fabio. It is based on what you need, or what is your purpose. First of all do you want to make it at macroscopic or microscopic level? Do you need visual effects or 3D views? Or just want to compare a model for a specific purposes. Answer of these questions determined your needs about simulation software. After that I can recommend some.
I have worked with Corsim quite a bit and would recommend it for Micro simulations. As mentioned by Sazi Murat, you need to figure out if you want to perform macro, meso or micro simulation.
Hope the following link helps you decide on simulation software.
Simulation software is only as good as the programmer who designed it, so use common sense when viewing your results. I recently saw a Synchro simulation that made absolutely no sense in terms of what drivers would or would not do, at least Michigan drivers.
Before deciding, it is important to consider the mode in which you will be using the simulation. Is it for research, teaching, demos, pre-commercial pilots?
You will find that seemingly expensive simulation packages can be offered for a very low price if it is only for research or teaching, and even for free. I have found that it is best to call the simulation company and discuss your need. It also helps if you can team up with your colleagues and if you can indicate that you are considering establishing a new, long-term relationship with the company.
This is important because, instead of starting with the question, "What is free?", you may start with the question, "What is best for me?"
You can use PTV VISUM. Its quite user friendly and user can easily import GIS shapefiles, databases etc. The company also offers the licensed software for free to students for their thesis work.
There is no single true answer to this question. I would suggest you look into the suggested open source frameworks and look:
which one is best suited for your purpose (e.g. micro scale vs meso scale)
which one solves your problem with the least effort
Both communities--MATSim and SUMO--are very supportive towards new users and both frameworks are excellently maintained. So I would suggest looking into these two first.
If you are not happy with what you find you can always switch to a commercial software. But keep in mind that 'commercial' does not necessarily mean better. However, at least you can buy the needed support to setup your model much easier for commercial software packages than for an open source solution.
We distribute our package for free. Some limitations apply. Download the free version, ask for authentication and activation and if you have special needs we can add modules.
The Best Transportation Simulation software is Transmodeler from Caliper Corporation (maker of the popular GIS software Maptitude, Maptitude for Redistricting and TransCAD). Transmodeler does micro-meso-macro all 3 simulations and allows all GIS functionality as usual on top of simulation. The SE version (for few 100 nodes simulation) is ~$1,000 and the full version is ~ $10,000 but worth every penny ...
I would suggest SUMO, CORSIM and VISSIM. You can select any one depending on your environment. SUMO is freely available, CORSIM and VISSIM are paid software.
I found PTV Viswalk a suitable social force based model to study pedestrian crowd. PTV vissim is also suitable to apply area-based and multiple road users simulations. Whatever simulation you used, I found it imortant to calibrate and validate it to a local context as the results (e.g. capacity, delay time etc) could be significantly different from analysis with default network settings.