-
Notifications
You must be signed in to change notification settings - Fork 1
A combination of reinforcement and evolutionary learning are used in an attempt to allow autonomous vehicles to learn behaviors necessary for the navigation of an intersection in a simple 2D traffic simulator. The open source AIM4 simulator developed by the Learning Agents Research Group at The University of Texas was extended to implement a spe…
License
chawk073uofc/aim4sim-shout-ahead
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
This README is for a version of the AIM4 traffic simulator that was modified in order to carry out a research project in machine learning. For the README for the original traffic simulator, see AIM4_README.txt. A demo video is available at https://youtu.be/DuWz5eTghcs. Downloading and Running the Program: 1) Clone this repository. 2) Run aim4.Main Using Eclipse: 1) Open Eclipse 2) Select "Checkout Projects form Git" from the Welcome Screen 3) Select "Clone URI" and click "next" 4) Enter https://github.com/chawk073uofc/aim4sim-shout-ahead.git as the URI. No user name or password is required. Click "next". 5) Select a destination for your local Git repository and click "next" 6) Select "import existing Eclipse projects" and click "next" 7) Click "Finish" 8) Select aim4.Main in the project explorer 9) Click run (green arrow) Using the Simulator: This traffic simulation program includes 4 modes that can be used to evaluate 4 different ways of routing traffic through an intersection. These include the three that were included by the authors of the original simulator 1) Traffic Lights 2) Stop Signs 3) Autonomous Intersection Management (AIM) Protocol and one that was added to facilitate machine learning of necessary driving behaviours through a learning method that is a hybrid of reenforcement and evolutionary learning 4) Shout Ahead. To be able to access all simulation modes set aim4.config.Debug.SHOUT_AHEAD_FAST_DEBUG_MODE to FALSE, run the program, select one of the four simulation modes, choose simulation parameters, and click "run". To start full machine learning run, where all simulation and learning parameters are taken from aim4.ShoutAheadAI.ShoutAheadSimSetup, set the same flag to TRUE and run the program. Multiple simulations are automatically run one after the other until the last simulation of the last generation is complete. Understanding the code: To see how the learning works see aim4.ShoutAheadAI.LearningRun and to see how an individual simulation works see aim4.ShoutAheadAI.ShoutAheadSimulator.step(double). Collecting experimental results: Set aim4.ShoutAheadAI.LearningRun.LEARNING_RECORDS_PATH to the directory in which you would like to store the results of learning runs that you run. This will create a directory for each learning run (having a name indicating the time and date it was performed). In this directory you will find two files: 1) learningLog a csv containing key metrics for each simulation that can be easily imported into an Excel spreadsheet 2) parameters a text file containing the simulation and learning parameters that were used for this learning run
About
A combination of reinforcement and evolutionary learning are used in an attempt to allow autonomous vehicles to learn behaviors necessary for the navigation of an intersection in a simple 2D traffic simulator. The open source AIM4 simulator developed by the Learning Agents Research Group at The University of Texas was extended to implement a spe…
Topics
Resources
License
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published