Autoware Open-Source Autonomous Driving Stack

Autoware is the world's leading open-source software project for autonomous driving, providing a complete autonomous driving software stack built on ROS (Robot Operating System) for self-driving vehicles. Originally developed by Nagoya University's TIER IV team (led by Shinpei Kato) in 2015, Autoware is now managed by the Autoware Foundation (established in 2018, with members including ARM, Autocruise, LG, Plexim, Samsung, and TIER IV). Key features: perception stack including LiDAR-based 3D object detection (Lidar Apollo Instance Segmentation, PointPillars, CenterPoint), camera-based object detection (YOLOv3/v5/v8, TensorRT acceleration), traffic light recognition, lane detection, and sensor fusion (LiDAR-camera-calibration). Localization using Normal Distributions Transform (NDT) matching and LiDAR-based 6DOF pose estimation with EKF (Extended Kalman Filter) fusion of GPS/IMU/wheel odometry. Planning: mission planning (route via lanelet2 maps), behavior planning (lane change, intersection, avoidance), and motion planning (A* search, optimization-based trajectory generation) with velocity planning and obstacle avoidance. Control: pure pursuit, MPC (Model Predictive Control), and LQR (Linear Quadratic Regulator) controllers for steering, throttle, and brake actuation. Simulation: integration with CARLA Simulator, AWSIM (TIER IV's custom simulator), and SVL Simulator for testing without physical vehicles. Autoware.Universe: the latest architecture based on ROS 2 (Foxy, Galactic, Humble, Iron), using apollo-hardware-interface and autoware_auto_msgs. Autoware.AI: the legacy architecture based on ROS 1 (Kinetic, Melodic). Map support: Lanelet2 high-definition maps, Point Cloud Maps, Vector Maps. Vehicle interface: PACMod, Speed/Steer interface for various vehicle platforms (Lexus RX450h, Lincoln MKZ, Jaguar I-PACE). Sensor support: Velodyne, RoboSense, Hesai, Ouster LiDARs; ZED, RealSense cameras. Open source under Apache-2.0.