ALFA: Advanced LiDAR Framework for Automotive Applications

The automotive industry is rapidly moving towards the deployment of fully autonomous vehicles, which rely on advanced driver-assistance systems (ADAS) and robust perception systems for safe and efficient navigation. Among these technologies, Light Detection and Ranging (LiDAR) sensors have gained prominence for their ability to provide high-resolution 3D representations of the vehicle's surroundings in real-time. This capability significantly improves obstacle detection and object recognition. However, the integration of LiDAR sensors with ADAS may still face several challenges, such as image noise and data output size, slowing their widespread adoption. This article introduces the Advanced LiDAR Framework for Automotive (ALFA), a tool designed for the development and evaluation of LiDAR data processing algorithms in different setups, with a focus on embedded platforms with acceleration capabilities. ALFA enables seamless interfacing with a wide range of devices and the creation and deployment of hardware accelerators for processing point cloud data across diverse applications, facilitating the integration with the computing system of the car. In addition to presenting the framework, this article provides a comprehensive evaluation and benchmarking of ALFA, including point cloud filtering and real-time ground segmentation.