Project Abstract

In an era where autonomous systems are pivoting from theoretical constructs to practical, real-world applications, this work arrives at a critical juncture to address the challenges of real-time decision-making and object detection in autonomous driving systems. The motivation stems from the imperative need for enhanced safety, efficiency, and reliability as these systems inch closer to widespread adoption, impacting transportation and, subsequently, societal norms on a grand scale.This research posits a novel integration of Convolutional Neural Networks (CNN) and Curriculum Reinforcement Learning (SAC) tailored for the autonomous vehicular context. It orbits around the central aim to evaluate and enhance the precision and adaptability of such systems within a dynamic real-world environment. The core research question probes the extent to which these advanced learning algorithms can optimize autonomous driving performance metrics.The methodological backbone of the study is an empirical approach leveraging a mixed-methods design that juxtaposes quantitative performance data with qualitative behavioral analysis. The study harnesses the computational prowess of CNN for object detection and SAC for learning efficient driving policies, benchmarked across diverse track configurations under varying environmental conditions.Key findings reveal a marked progression in the SAC agent?��s capability to modulate speed and garner rewards, illustrating a substantive leap in learning efficiency and strategy optimization. Notably, the agent exhibited a consistent ability to generalize its driving strategy, maneuvering adeptly through complex track layouts, and demonstrating a high degree of detection accuracy.The conclusion of this research accentuates a dual contribution: It delineates a robust framework for autonomous system training and provides empirical evidence supporting the application of SAC agents in real-world autonomous driving scenarios. This study illuminates the pathway for future innovations in the domain, ultimately expanding the horizons of artificial intelligence in practical applications.

Keywords: Artificial Intelligence, Reinforcement Learning, Convolutional Neural Networks

 Conference Details

Session: Poster Session B at Poster Stand 6

Location: Sir Stanley Clarke Auditorium at Wednesday 8th 09:00 – 12:30

Markers: Matt Roach, Lu Zhang

Course: BSc Computer Science, 3rd Year

Future Plans: I’m undecided