Route planning for automated driving1. Extend a global path by adding extra drivable links - Path extension is designed for the vehicle to move onto the side links in a much wider range2. Create detour route - Detour route helps to avoid obstacles ahead - The driving system plans for the vehicle to take the side link and then rejoin the main route3. Manage sharing links in a route - A global route is not guaranteed to have non-repeating links. The same link may appear more than once in a global route, depending on the specific driving plan - The driving system keeps track of the next and previous links for each shared link It ensures that the vehicle follows the driving plan4. Ensure the stability of the current link - The current link plays a vital role in self-driving If the current link is not found or even not stable, the driving system may deviate from its route5. Build a variety of simulated environments - Develop a Python client code for CARLA, a simulation tool for testing automated driving systems on different driving scenarios - Construct the testing environment using the Unreal Engine editor

1. 2D/3D Lidar Viewer Program (Video: https://youtu.be/lI8I-UYhpHg) - Design a C# software program to visualise the distance data coming through a lidar on both 2D and 3D planes - Convert 2D and 3D distance data based on the FoV and detection range of the lidar sensor - Combine two 8-bit binary numbers, assuming little-endian and unsigned format, into a single decimal distance measurement - Embed an unmanaged DLL written in C into C#, using Marshaling array elements for C-style arrays - Optimise the code to speed up the conversion time of raw data into 2D and 3D point clouds by 50%2. ROS Driver for 2D/3D Lidar and Robot - Develop a C++ ROS driver to plot 2D and 3D distance data on RVIZ using LaserScan and PointCloud - Create a C++ ROS driver to visualise the real-time robot pose on RVIZ and control the robot (Video: https://youtu.be/l3xZ3a1hWG0)3. Shortest path finding system - Dijkstra finds a path in a network of the nodes which is created from 3D lidar data by the centroid-clustering algorithm4. Object Detection using 3D distance data (Video : https://youtu.be/73bk0Ffn5AU) - Prevent a vehicle from colliding with both static and dynamic objects by analysing the distribution of the rate of change in frames

1. BigData analysis - Grabbed rawdata on a daily basis via API - Organised data and found out abnormal patterns of the data - Splitted the dataset accummulated on a monthly basis into each weekly report per publisher* Python was used to speed up the cumbersome task from 8 hours to a few seconds2. Communication with global based companies (publishers)

1. Scheduling2. Outsourcing management3. English translation to Korean