Flow Project Flow Versions Save

Computational framework for reinforcement learning in traffic control

v0.4.1

4 years ago
  • introduced support for a new RL library, stable-baselines: https://github.com/hill-a/stable-baselines
  • removed and deprecated all rllab features
  • introduced a multi-lane generic highway network with on- and off-ramps
  • added a tutorial on running the bottleneck experiments
  • added a tutorial for running RLlib experiments on EC2 servers
  • some refactoring to features in grid network
  • increased unit test coverage for renderer features

v0.4.0

4 years ago
  • upgraded to sumo 1.1.0, see PR #293
  • new router class, see PR #524
  • updated documentation for most methods and classes

v0.3.1

4 years ago

This is the final release before we are switching to a more recent version of SUMO. Some changes include:

  • Cleanup to many of the scripts from redundant or unsed pieces of code.
  • Modifications to aimsun functionality and inclusion of an aimsun scripting class.
  • Extended unit test coverage. Our coverage is now about 90%.
  • Tutorials on using OpenStreetMap and Network Template scenarios.
  • Updated behavior to some of the examples and benchmarks to be compatible with the newer version of SUMO.

v0.3.0

5 years ago
  • support for a new traffic simulator: Aimsun
  • simulator specific commands were abstracted and moved from the environments to a kernel class, consisting of simulation, scenario, vehicle, and traffic_light subkernels. The subkernels were implemented for SUMO and Aimsun, and further ones can be created to support a plethora of other traffic simulators

v0.2.1

5 years ago

v0.2.0

5 years ago

0.1.0

6 years ago