Towards a Real-Time End-to-End Pipeline for Autonomous Vehicles
Aulia Widyaputra, Bambang Riyanto Trilaksono · 2025
This paper provides an approach to deploying a soft real-time end-to-end system for autonomous vehicles. There are several components that will be included in the system. Most importantly is a 3D object detection model. Our system uses LiDAR to identify objects in a 3D environment. A tracking algorithm will then be used to derive each object's identity and velocity between frames. This will then be fed into a decisionmaking algorithm to determine an appropriate action for the vehicle. The system developed will also include a 2D camerabased detection model to identify object and signs. Lastly, the system will include a rendering system to visualize the results in a graphical user interface or GUI.