Cognitive Automobile Laboratory

This is a course project from MRT of Karlsruhe Institute of Technology, focusing on implementation of algorithms for autonomous driving.

Team KITAF

• Perception Team: Kuangye Chen, Esteban Rivera
• Control Team: Weiyun Chen, Tianjing Chen

Important Packages from Team KITAF

kitaf_detector_ros_tool

Detection of traffic signs and other vehicles based on TensorFlow and SSD neural network.
Our result is capable of detecting traffic signs in real time with an accuracy above 96%, and detecting vehicles with low false negative rate.

kitaf_lateral_control_ros_tool

Controlling the vehicle based on stanley controller.
We first use kinematic model for its easy implementation, and then move to dynamic model for better performance.

map_creation_ros_tool, kitaf_navigation_ros_tool

• Generating an occupancy grid map based on point cloud from kinect2. With local grid maps from each time step, we maintain a global grid map with bayes updating rule.
• Providing path when driving along the road. Use potential field method based on occupancy grid to plan a path for going throungh the unknown environment with pylons.

kitaf_following_ros_tool

Control our vehicle to follow another vehicle based on detection results and point clouds.

Final Competition

We won the first place in the final! 🏆