EVOART-O Autonomous Vehicle System (evoart_ws)

EVOART-O is an end-to-end autonomous vehicle software stack built on ROS 2 (Jazzy/Humble) and Gazebo (Ignition / ros_gz). Designed specifically for autonomous track driving, urban navigation, and robotics competitions (such as RoboTaksi / Teknofest), this repository integrates 3D simulation, multi-sensor localization (EKF), Ackermann steering kinematics, path planning (Nav2), perception, and behavior decision trees into a modular architecture.

---

🏗️ System Architecture & Packages

The workspace is organized into functional ROS 2 packages located inside src/:

Package Name	Description
evoart_bringup	Core Orchestrator: Contains the master launch files (robotaksi.launch.py) that coordinate simulation, visualization (RViz2), sensor fusion (EKF), navigation, and control.
evoart_description	Robot & World Descriptions: Contains the vehicle URDF/Xacro (robotaksi.urdf.xacro), 3D meshes, RViz configurations (robotaksi.rviz), track environments (robotaksi_pist_2024.sdf, teknofest_city.sdf), and custom traffic signage models.
evoart_navigation	Autonomous Navigation & Fusion: Contains Nav2 configurations (nav2_params.yaml), global/local planners (DWBLocalPlanner, NavfnPlanner), behavior trees, and robot_localization EKF settings (ekf.yaml).
evoart_control	Vehicle Kinematics & Control: Converts standard ROS velocity commands (/cmd_vel) into Ackermann steering geometry (cmd_vel_to_ackermann).
evoart_perception	Computer Vision & Detection: Contains deep-learning and OpenCV pipelines for lane detection (lane_detector.py) and traffic sign/object detection (object_detector.py).
evoart_behavior	Decision Making: Implements state machines and behavior trees (behaviortree_cpp) for intersection handling, stop signs, and obstacle avoidance.
evoart_interfaces	Custom Interfaces: Project-specific ROS 2 messages, services, and action definitions.
velodyne	LiDAR Hardware Drivers: Drivers and pointcloud/laserscan converters for physical Velodyne VLP-16 LiDARs.
zed-ros2-wrapper	Stereo Camera Drivers: Drivers and wrapper nodes for physical Stereolabs ZED 2i stereo cameras.

---

📋 Prerequisites & Requirements

• Operating System: Ubuntu 24.04 LTS (for ROS 2 Jazzy) or Ubuntu 22.04 LTS (for ROS 2 Humble)
• ROS 2 Distribution: Jazzy Jalisco (Default) or Humble Hawksbill
• Simulation Engine: Ignition Gazebo (gz-sim / ros_gz_sim & ros_gz_bridge)
• Python Virtual Environment: Python 3.12+ with a dedicated virtual environment (.venv) for machine learning and CV dependencies (ultralytics, opencv-python, pyserial)

---

⚙️ Installation & Build Setup

1. Source ROS 2 & Activate Virtual Environment

Before compiling or running any launch files, always source your ROS 2 environment and activate the project virtual environment:

# Source ROS 2 Jazzy (or change to /opt/ros/humble/setup.bash if using Humble)
source /opt/ros/jazzy/setup.bash

# Activate local Python virtual environment
source /home/leykun/EVOART-O/.venv/bin/activate

2. Build the Workspace

Compile the core simulation and navigation stack using colcon:

cd "/home/leykun/Documents/EVOART-O /evoart_ws"

# Build simulation packages with symlink install for rapid iteration
colcon build --symlink-install --packages-select evoart_interfaces evoart_description evoart_control evoart_navigation evoart_bringup

# Source the generated workspace setup
source install/setup.bash

Note

Hardware driver packages (zed_wrapper, velodyne_driver) depend on physical device SDKs (e.g., Stereolabs ZED SDK or libpcap-dev) and are bypassed during simulation builds.

---

🚀 Running the Simulation (Gazebo + RViz2)

The master simulation pipeline launches Gazebo (Ignition), RViz2, the EKF Sensor Fusion node (ekf_node), the Nav2 Autonomous Navigation Stack, and the Ackermann Steering Controller simultaneously using simulation time (use_sim_time: True and /clock).

1. Launch Master Orchestrator

cd "/home/leykun/Documents/EVOART-O /evoart_ws"
source /opt/ros/jazzy/setup.bash
source /home/leykun/EVOART-O/.venv/bin/activate
source install/setup.bash

ros2 launch evoart_bringup robotaksi.launch.py

2. What Happens on Launch

• Gazebo Track (robotaksi_pist_2024.sdf): Opens the 2024 RoboTaksi competition track equipped with custom 3D road signage (DurakTabelasi, DonelKavsakTabelasi, ParkYeriTabelasi, etc.) and traffic lights.
• Vehicle Spawn: Spawns the evoart_robotaksi Ackermann vehicle at z=0.5m origin.
• ROS 2 Topic Bridge (ros_gz_bridge): Automatically bridges Ignition Gazebo simulation topics to standard ROS 2 topics:
  • /clock ↔ rosgraph_msgs/msg/Clock
  • /cmd_vel ↔ geometry_msgs/msg/Twist
  • /odom ↔ nav_msgs/msg/Odometry
  • /scan ↔ sensor_msgs/msg/LaserScan
  • /joint_states ↔ sensor_msgs/msg/JointState
  • /tf ↔ tf2_msgs/msg/TFMessage
• RViz2 Visualizer: Opens pre-configured with robotaksi.rviz, displaying real-time TF transformations (base_link, lidar_link, zed_camera_link), odometry arrows, and 2D/3D sensor pointclouds.
• Autonomous Navigation (Nav2): Initializes planner_server and controller_server with DWBLocalPlanner and NavfnPlanner, ready to accept navigation goal poses (2D Goal Pose in RViz).

---

🗺️ Simulation Environments

You can easily switch the default Gazebo world in evoart_description/launch/gazebo.launch.py:

1. robotaksi_pist_2024.sdf (Default): Full RoboTaksi race track with realistic road layouts, mandatory turn signs, speed limits, and traffic light intersections.
2. teknofest_city.sdf: Urban city environment for wide-area testing.

All required 3D models are located inside evoart_description/models/ and are automatically exported via GZ_SIM_RESOURCE_PATH and IGN_GAZEBO_RESOURCE_PATH.

---

📡 Key ROS 2 Topics

Topic	Type	Description
/clock	rosgraph_msgs/msg/Clock	Synchronized simulation clock from Gazebo (use_sim_time: True)
/cmd_vel	geometry_msgs/msg/Twist	Target linear and angular velocity commands issued by Nav2 or teleop
/odom	nav_msgs/msg/Odometry	Raw wheel odometry published by Ignition Gazebo Ackermann plugin
/odometry/filtered	nav_msgs/msg/Odometry	Filtered state estimation fused from wheel odometry and IMU via ekf_node
/scan	sensor_msgs/msg/LaserScan	360° 2D laser scan generated from simulated Velodyne VLP-16 GPU LiDAR
/zed/zed_node/rgb/image_rect_color	sensor_msgs/msg/Image	Rectified RGB camera feed from simulated ZED 2i camera
/joint_states	sensor_msgs/msg/JointState	Real-time steering and wheel rotation angles
/tf & /tf_static	tf2_msgs/msg/TFMessage	Complete kinematic transformation tree (odom → base_footprint → base_link → sensors)

---

🛠️ Troubleshooting & Tips

• Spaces in Directory Paths: The workspace launch files utilize robust Python XML/Xacro parsing (xacro.process_file(...).toxml()) and quoted gz_args. If moving the repository, directory paths containing spaces (EVOART-O /evoart_ws) are fully supported.
• Nav2 Plugin Syntax: Configured to use ROS 2 Jazzy double colon namespace syntax (::) inside nav2_params.yaml (nav2_navfn_planner::NavfnPlanner and nav2_behaviors::Spin).
• TF / Time Jumps: When restarting Gazebo simulation, you may see Detected jump back in time. Clearing TF buffer warnings in terminal. This is normal behavior when /clock resets to 0.0s.

evoart