Autonomous navigation for Unmanned Surface Vessels (USVs) is essential for expanding their applications in maritime environments, where situational awareness and collision avoidance are critical. However, these environments present unique challenges for multi-object tracking (MOT) such as a wide diversity of vessel sizes, a high dynamic environment due to waves and current, and low visibility. Although neural network-based models could facilitate MOT in such scenarios, there is a lack of publicly available datasets in Maritime environments. Therefore, this work presents a probabilistic, point-based MOT framework specifically designed for short-range tracking (≤ 100 m), utilizing weather-resilient sensors to ensure robust operation in diverse conditions. The framework integrates LiDAR and Automatic Identification System (AIS) data through a late fusion approach, improving state estimation by combining the dynamic tracking abilities of LiDAR with AIS’s vessel identification capabilities. Key methods include an Interacting Multiple Model (IMM) for adaptive maneuver handling and Joint Probabilistic Data Association (JPDA) for data association. Validation in a simulated environment highlights significant limitations, showing that while the framework can manage basic tracking tasks, it remains far from optimal for the full scope of nearshore and coastal applications. This thesis underscores the need for further research to meet the demands of maritime MOT, particularly in handling the unique challenges posed by large vessels.