Unmanned surface vehicles (USVs) are increasingly utilized in complex maritime tasks such as harbor surveillance, coastal monitoring, and port patrols. Safe navigation in congested and dynamic waters requires a unified framework for perception, mapping, planning, and control. This paper presents a fully integrated navigation system for a catamaran-style USV operating in crowded harbor environments. The system forms a closed loop from multi-sensor perception to motion execution. It fuses data from GPS, IMU, 3D LiDAR, and cameras for real-time SLAM and obstacle detection. Robust localization is achieved with a LiDAR–inertial odometry algorithm (LIO-SAM) enhanced by GPS/IMU fusion via an Extended Kalman Filter. Path planning combines a global A* planner on occupancy maps with a local Dynamic Window Approach (DWA) for dynamic obstacle avoidance. Motion execution is handled by a dual-loop PID controller regulating linear and angular velocities. The framework is validated in high-fidelity Gazebo harbor simulations, demonstrating safe navigation toward waypoints while avoiding static and moving obstacles. Results show that integrating SLAM, A*-based global planning, and DWA local refinement significantly improves autonomy. The architecture emphasizes efficiency, modularity, and extensibility, supporting future vision-based perception for enhanced safety in dynamic maritime environments.