ROS workspace for a Self-Driving Polaris GEM e2 golf cart using Dataspeed's drive-by-wire package.
The lane keeping package uses OpenCV that detects white/yellow lanes and publishes Radius of Curvature and offset from middle of lane. Uses HLS color space to threshold and histogram to classify lanes.
rosrun lane_follower lane_tracker.py will launch the lane tracking node and publishes offset on /laneOffset topic.
Has a neural network object detection package based on darkNET's yolo v3. Forked from leggedrobotics
roslaunch darknet_ros/darknet_ros yolo_v3 will launch the object detection node. Current image topic is set to /usb_cam/image_raw and can be configured in config/ros.yaml
roslaunch car_teleop hector_slam_test.launch filename_true:=<bagfile.bag> path:=<folder location> launches a HectorSLAM node without odometry with bag files with LiDAR/pointcloud data.
lane_teleop.launch launches DataSpeed's DbW simulator. Also launches keystroke.py which kills the path planning node and publishes command velocity messages(UlcCmd).
Use roslaunch car_teleop lane_teleop to launch the keyboard teleop node and use W-S-A-D to control the car in Gazebo.
