Trajectory prediction of dynamic obstacles in fleet management systems

• Main Author: Quintero, Dann Marko Gayanes
• Other Authors: Su Rong
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Trajectory prediction
• Online Access: https://hdl.handle.net/10356/176559
• Institution: Nanyang Technological University

Fleet management systems play a pivotal role in enhancing the operational efficiency of logistics, manufacturing, and transportation sectors. This thesis investigates the improvement of trajectory prediction for Automated Guided Vehicles (AGVs) within these systems, specifically tailored for industrial contexts. While conventional approaches encounter challenges in dynamic environments, recent advancements, particularly in deep learning, offer promising avenues for resolution. Through an exhaustive exploration of deep learning techniques and their applicability, this study endeavours to augment trajectory prediction accuracy for AGVs, while concurrently addressing pertinent real-world challenges such as interpretability, computational efficiency, and industry-specific usability. The primary objective of this research is to identify the optimal trajectory prediction model for AGVs within fleet management systems. This pursuit focuses on various methodologies, including the possibility of integrating multiple algorithms to achieve superior performance. The findings of this research contribute significantly to the ongoing efforts aimed at optimizing trajectory prediction methods within fleet management systems. It was observed during the course of this research that most literature review lacks comprehensive coverage that combines a thorough understanding of Trajectory Prediction with practical utilization of reviewed algorithms. Existing literature predominantly showcases algorithmic superiority without delving into practical usability aspects. Furthermore, the rapidly evolving technological landscape underscores the imperative of future-proofing research endeavours, prompting a conscientious effort to ensure the enduring relevance and utility of this thesis. As Artificial Intelligence continues to evolve at an unprecedented pace, the potential for enhancing trajectory prediction for AGVs within fleet management systems becomes increasingly promising. With ongoing advancements in deep learning techniques and the proliferation of big data analytics, the trajectory prediction models of tomorrow hold immense potential to revolutionize industrial operations. By utilizing AI-powered solutions, accuracy, adaptability, and efficiency in trajectory prediction can be enhanced, leading to advancements in fleet management optimization. This study contributes to the fleet management sector by offering advanced trajectory prediction techniques, leading to improved usage of AGVs in industrial environments. As such, this study not only enhances our understanding of current trajectory prediction techniques but also provides a platform for driving innovation and progress in fleet management industries worldwide.