Speaker
Description
The ROS 2 Navigation Stack (Nav2) has emerged as a foundational framework for autonomous navigation, providing sophisticated path-planning and obstacle avoidance capabilities. However, optimizing its parameters to ensure safe and efficient navigation under dynamic environmental conditions—such as wind—remains a significant challenge. This research develops adaptive navigation strategies for both Unmanned Ground Vehicles (UGVs) and Unmanned Surface Vehicles (USVs) operating in wind-affected environments. We begin with controlled experiments using TurtleBot3 simulations in Gazebo to establish baseline performance metrics and identify key adaptation parameters. These findings inform our subsequent work with USVs, implemented using Virtual RobotX (VRX)—a high-fidelity maritime simulation environment that models complex ocean-atmosphere interactions. We focus on developing an adaptive navigation approach capable of dynamically tuning Nav2 parameters in response to changing wind conditions. To achieve this, we want to explore multiple methods, including Deep Reinforcement Learning (DRL), linear control techniques, and Gaussian Processes. By evaluating these approaches, we aim to enhance the robustness of Nav2 for real-world deployment in autonomous maritime and terrestrial robotics applications.
