Reinforcement autonomous driving project of Kovan Lab.
In this project we have built an environment which subclasses OpenAI's gym environment.
-
autonomouscar contains the gym environment and interacts with ROS.
-
sensormessages handles sensor data coming from ROS gazebo simulation.
In order to update any functionality related to autonomouscar, please refer to autonomouscar.
Action space consists of 3 inputs.
- Throttle must be in the range of [0.0, 1.0]
- Brake must be in the range of [0.0, 1.0]
- Steer must be in the range of [-1.0, 1.0]
In order to estimate the state of the agent you should utilize and combine different data from sensors. You can choose which sensors to use by modifying the autonomouscar's observation space.