Lidar Parser

The purpose of the lidar parser is to construct a point cloud from lidar packets. Although the lidar manufacturer provides a library, I needed custom functionality, and since integrating the functionality with the packet parsing process would result in zero-cost implementation, I decided to implement the driver myself. In addition, the official library integrated not only packet parsing but also the network stack, and since packet structures varied by lidar model, it was inconvenient to use. The library I have implemented provides a unified interface for accessing packet information and shows about 3x better performance than the official library.

To improve performance, the followings are considered:

• Using the memory space allocated to the packet as is in the network stack.
• Initializing Polymorphic Objects on the stack.
• Caching unit vectors to construct point cloud.
• Data alignment for memory localization.
• Utilizing SIMD.