TRAJECTORY PREDICTION OF MULTIPLE OBJECTS AROUND AUTONOMOUS TRAM USING INTERACTING MULTIPLE MODEL KALMAN FILTER ALGORITHM

• Main Author: Nur Fatimah, Rini
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/87018
• Institution: Institut Teknologi Bandung

The increasing number of private vehicles in major cities has led to severe traffic congestion, reduced transportation efficiency, and increased carbon emissions, which negatively impact the environment. As an eco-friendly public transportation solution, autonomous trams powered by electricity offer significant potential to alleviate congestion and emissions while improving transportation efficiency. However, the operation of autonomous trams in mixed-traffic environments requires accurate and efficient multi-object trajectory prediction capabilities to support safe decision-making, including collision avoidance. This study aims to develop a multi-object trajectory prediction method based on the Interacting Multiple Model Kalman Filter (IMM-KF) algorithm, capable of addressing the complexity of urban traffic. IMM-KF enables the combination of two motion models, namely Constant Velocity and Constant Acceleration, allowing it to capture diverse object dynamics in real-time. The research is validated through simulations using the inD dataset, implementation on a testbed vehicle, and testing on autonomous trams operating in the Jalan Slamet Riyadi area, Surakarta. The algorithm's evaluation focuses on measuring trajectory prediction accuracy, computational efficiency, and its ability to support decision-making, such as activating the Collision Avoidance (CA) system. Testing scenarios involve mixed- traffic environments, with data obtained from the autonomous tram’s LiDAR sensors. The results of this study demonstrate that the IMM-KF algorithm is capable of predicting multi-object trajectories with high accuracy over short to medium prediction time, as well as achieving computational efficiency that scales linearly with the number of objects, meeting real-time requirements. The system has also proven reliable in supporting decision-making processes, such as activating Collision Avoidance to prevent potential collisions.