Reinforsment learning (VAE + DDPG) controller for EagleMK4 car.
Demo.
This controller expects EagleMK4 or similar MIT-racecar like ackermann steering robot.
In general, controller can be adapted to any ROS-based robot with monocular camera.
Inputs:
sensor_msgs/Image
from topicimage
.sensor_msgs/Joy
from topicjoy
.
Outputs:
ackermann_msgs/AckermannDriveStamped
from topicnavigation
.
Joystick buttons:
- Triangle - switch tasks (train and test).
- Right bumper - when pressed enables autonomous mode. Unpressing stops the autonomous mode and stops the episode.