ROS2-based indoor autonomous navigation for moving obstacle avoidance

With the rapid advancement of automation technology, enabling mobile robots to navigate autonomously in dynamic and uncertain indoor environments has become increasingly critical. This paper presents a high-fidelity simulation for indoor autonomous navigation, developed using ROS2, with a focus on integrating a Livox MID360 LiDAR sensor for dynamic pedestrian obstacle avoidance. The MID360 LiDAR sensor is modeled in URDF/XACRO based on manufacturer specifications, configured with a Gazebo Ray plugin supporting a maximum range of 40 m, and enhanced with a Gaussian noise model to simulate real-world range inaccuracies. The sensor’s point cloud output is validated in Rviz2 to ensure consistency and accuracy. For dynamic obstacle simulation, a pedestrian actor system is implemented in Gazebo, where each pedestrian follows predefined paths governed by a state machine that emulates walking, turning, and pausing behaviors, leveraging the actor element for realistic motion animations. These components are integrated into a unified simulation environment to evaluate navigation performance using A∗ strategy for global path planning and the Dynamic Window Bundle (DWB) for local obstacle avoidance. The final results demonstrate the platform’s ability to achieve smooth, collision-free path tracking and real-time obstacle avoidance in scenarios with dynamic pedestrian obstacles. This research offers a reproducible and scalable testbed for iterative development and rigorous evaluation of autonomous navigation algorithms in complex indoor settings.