ROS2 Autonomous Navigation

An autonomous navigation system built in C++ and ROS 2, and simulated in Foxglove.

0109.1.mp4

Process

Control Node

• Subscribers: /path, /odom
• Publishers: /cmd_vel
• Function: Starts the journey by reorienting itself towards the final goal. It uses a Pure Pursuit controller, calculating the curvature required to reach a specific lookahead point, and dynamically throttles velocity based on turn sharpness to ensure a stable route.

Costmap Node

• Subscribers: /lidar
• Publishers: /costmap
• Function: Calculates a "cost" for each section of the local area.
  • Low Cost: Unknown or safe to travel.
  • High Cost: Objects present; robot must avoid.
• Logic: It converts Lidar data (polar coordinates) into Cartesian points via trigonometry, marks obstacles, and inflates them so that surrounding cells receive a decaying cost based on distance from the obstacle.

Map Memory Node

• Subscribers: /costmap, /odom
• Publishers: /map
• Function: Transforms local sensor data into the global frame and accumulates it into a consistent, persistent grid (often visualized in red).
• Logic: This ensures sensor data is "remembered" even after the robot moves away, creating a static map for the planner to refer to.

Planner Node

• Subscribers: /map, /goal_point, /odom
• Publishers: /path
• Function: Implements the A (A-Star) algorithm* to find an efficient path to the goal point while considering obstacles.
• Logic: It features a "Lazy Check" loop that continuously validates if the current path is safe as new Lidar data is provided, rather than replanning every single frame.

Acknowledgements

The framework and setup for this project was done by the WATonomous design team. The integration and programming of the modules listed above were done by me.