SP2001::ISG::Green Light District / News: Recent posts

i-Atracos, which stands for "Intelligent Argentinean TRAffic COntrol System", is an expanded and improved version of GLD. It was made by Gastn Daniel Escobar, Marcelo Pastorino, Gustavo Brey, and Marisa Espinosa.

i-Atracos supports these functions:
* Dynamic incident generation: Users can declare roads unusable with a mouse click.
* Flow Traffic Schedule: It is possible to run a simulation with changing SpawnFrequencies by specifying intervals in an XML file.
* Simulation Data Sender Interface: Simulation data can be exported in XML format.
* Traffic interval sender Interface: An interface to send the SignConfig intervals.
* Green-wave support: In the new SignConfig Configuration format, you can send the node's offsets to produce a green wave.

Green Light District binaries, source, docs and techdocs. We got it all. Grab it while you can, and see how well our self-learning algorithms switch those trafficlights to green, optimizing trafficflows.
Thanks to the very open architecture, it will be very easy for any intrested Java coder to implement new TrafficLightController-algo's, and we most certainly invite you all to do so.

This will be the 'final' release for now, as this project has come to an end. Enthousiastic as we are, there might be some sudden new releases in the future, but none are guaranteed.... read more

The release of version 0.7 of GLD means a huge step forward in our project. Internally a whole lot has changed, and also the eye will notice many differences to older releases.
The Editor has been hugely expanded, and there's been a lot of improvement on the interface design and userfriendlyness.
Roadusers now actually really cross an intersection or junction, however they arent drawn while doing so, this adds hugely to the realism of testing the TLC-algorithms.
Extended statistical views and tracking windows provide very strong means of testing out performance of these algorithms. Possibilities of exporting to standard user-defined seperation character files provide means of comparing the performance of algorithms in full-flexed statistical programs.
Algorithmically a lot has changed, many new algorithms, many working much better now, though the best is still to come..... read more

The first public release of both the source and binaries of the GLD program is a fact now.
The program already boasts near all functionality as being specified in the requirements and product specifications.

Two self-learning algorithms, TC-1 and TC-3, developed by Marco Wiering of the Utrecht University are already in working (and learning;) state and can be seen in action in the simulator.

Keep an eye on the releases as from now on every one or two weeks there'll be another major GLD release with more functionality, better graphics and GUI, more AI-algorithms (genetic, neural, reinforcement learning) and maybe even a real logo.

Finally the specifications of our project are in a state we may call workable. During the runningtime of our project we shall continue to keep these up-to-date with our implementing-choices and reasons.

aloha